Bellingham, Washington USA

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Library of Congress Cataloging-in-Publication Data

Palmer, James M.
Art of radiometry / James M. Palmer and Barbara G. Grant.
p. cm. -- (Press monograph ; 184)
Includes bibliographical references and index.
ISBN 978-0-8194-7245-8
1. Radiation--Measurement. I. Grant, Barbara G. (Barbara Geri), 1957- II.
Title.
QD117.R3P35 2009
539.7'7--dc22
2009038491

Published by

SPIE
P.O. Box 10
Bellingham, Washington 98227-0010 USA
Phone: +1 360.676.3290
Fax: +1 360.647.1445
Email: Books@spie.org
Web: http://spie.org

Copyright © 2010 Society of Photo-Optical Instrumentation Engineers (SPIE)

All rights reserved. No part of this publication may be reproduced or distributed

in any form or by any means without written permission of the publisher.

The content of this book reflects the work and thought of the author(s).
Every effort has been made to publish reliable and accurate information herein,
but the publisher is not responsible for the validity of the information or for any
outcomes resulting from reliance thereon.

Printed in the United States of America.

On the cover: A Crooke radiometer and the equation of radiative transfer.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Contents
Foreword ................................................................................................... xi
Preface .................................................................................................... xiii

Chapter 1 Introduction to Radiometry / 1

1.1 Definitions............................................................................................ 1
1.2 Why Measure Light?............................................................................ 2
1.3 Historical Background.......................................................................... 4
1.4 Radiometric Measurement Process .................................................... 5
1.5 Radiometry Applications...................................................................... 7
References ................................................................................................ 9

Chapter 2 Propagation of Optical Radiation / 11

2.1 Basic Definitions ................................................................................ 11
2.1.1 Rays and angles.................................................................... 11
2.1.2 System parameters ............................................................... 19
2.1.3 Optical definitions .................................................................. 23
2.2 Fundamental Radiometric Quantities ................................................ 24
2.2.1 Radiance ............................................................................... 24
2.2.2 Radiant exitance.................................................................... 26
2.2.3 Irradiance .............................................................................. 28
2.2.4 Radiant intensity .................................................................... 29
2.3 Radiometric Approximations.............................................................. 30
2.3.1 Inverse square law ................................................................ 30
2.3.2 Cosine3 law ........................................................................... 31
2.3.3 Lambertian approximation ..................................................... 32
2.3.4 Cosine4 law ........................................................................... 33
2.4 Equation of Radiative Transfer .......................................................... 36
2.5 Configuration Factors ........................................................................ 38
2.6 Effect of Lenses on Power Transfer .................................................. 40
2.7 Common Radiative Transfer Configurations ..................................... 42
2.7.1 On-axis radiation from a circular Lambertian disc ................. 42
2.7.2 On-axis radiation from a non-Lambertian disc ...................... 43
2.7.3 On-axis radiation from a spherical Lambertian source .......... 44
2.8 Integrating Sphere ............................................................................. 46
2.9 Radiometric Calculation Examples.................................................... 48
2.9.1 Intensities of a distant star and the sun ................................. 48

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 vi Contents

2.9.2 Lunar constant....................................................................... 50

2.9.2.1 Calculation ..................................................................... 50
2.9.2.2 Moon–sun comparisons ................................................. 51
2.9.3 “Solar furnace”....................................................................... 52
2.9.4 Image irradiance for finite conjugates ................................... 53
2.9.5 Irradiance of the overcast sky ............................................... 55
2.9.6 Near extended source ........................................................... 55
2.9.7 Projection system .................................................................. 56
2.10 Generalized Expressions for Image-Plane Irradiance ..................... 57
2.10.1 Extended source ................................................................... 57
2.10.2 Point source .......................................................................... 58
2.11 Summary of Some Key Concepts ................................................... 58
For Further Reading ................................................................................ 59
References .............................................................................................. 59

Chapter 3 Radiometric Properties of Materials / 61

3.1 Introduction and Terminology ............................................................ 61
3.2 Transmission ..................................................................................... 62
3.3 Reflection .......................................................................................... 63
3.4 Absorption ......................................................................................... 69
3.5 Relationship Between Reflectance, Transmittance, and
Absorptance ...................................................................................... 69
3.6 Directional Characteristics ................................................................. 69
3.6.1 Specular transmittance and reflectance ................................ 69
3.6.2 Diffuse transmittance and reflectance ................................... 73
3.7 Emission ............................................................................................ 76
3.8 Spectral Characteristics .................................................................... 77
3.9 Optical Materials Checklist ................................................................ 79
For Further Reading ................................................................................ 80
References .............................................................................................. 80

Chapter 4 Generation of Optical Radiation / 83

4.1 Introduction ........................................................................................ 83
4.2 Radiation Laws .................................................................................. 84
4.2.1 Planck’s law........................................................................... 84
4.2.2 Wien displacement law.......................................................... 86
4.2.3 Stefan-Boltzmann law ........................................................... 89
4.2.4 Laws in photons .................................................................... 89
4.2.5 Rayleigh-Jeans law ............................................................... 92
4.2.6 Wien approximation ............................................................... 93
4.2.7 More on the Planck equation................................................. 93
4.2.8 Kirchhoff’s law ....................................................................... 97
4.3 Emitter Types and Properties .......................................................... 102
4.3.1 Metals .................................................................................. 102

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Contents vii

4.3.2 Dielectrics ............................................................................ 102

4.3.3 Gases .................................................................................. 103
4.4 Practical Sources of Radiant Energy............................................... 104
4.4.1 Two major categories .......................................................... 104
4.4.2 Thermal sources.................................................................. 105
4.4.2.1 Tungsten and tungsten-halogen lamps ........................ 105
4.4.2.2 Other metallic sources.................................................. 108
4.4.2.3 Dielectric thermal sources ............................................ 108
4.4.2.4 Optical elements........................................................... 109
4.4.2.5 Miscellaneous thermal sources .................................... 109
4.4.3 Luminescent sources .......................................................... 110
4.4.3.1 General principles ........................................................ 110
4.4.3.2 Fluorescent lamps ........................................................ 115
4.4.3.3 Electroluminescent sources ......................................... 117
4.4.3.4 LED sources ................................................................. 117
4.4.3.5 Lasers .......................................................................... 118
4.4.4 Natural sources ................................................................... 119
4.4.4.1 Sunlight ........................................................................ 119
4.4.4.2 Skylight, planetary, and astronomical sources ............. 120
4.4.4.3 Application: energy balance of the earth ...................... 121
4.5 Radiation Source Selection Criteria................................................. 121
4.6 Source Safety Considerations ......................................................... 123
4.7 Summary of Some Key Concepts ................................................... 123
For Further Reading .............................................................................. 123
References ............................................................................................ 124

Chapter 5 Detectors of Optical Radiation / 127

5.1 Introduction...................................................................................... 127
5.2 Definitions ........................................................................................ 128
5.3 Figures of Merit ............................................................................... 131
5.4 #N$O%&I*S@E~^ ........................................................................... 133
5.4.1 Introduction to noise concepts............................................. 133
5.4.2 Effective noise bandwidth.................................................... 136
5.4.3 Catalog of most unpleasant noises ..................................... 137
5.4.3.1 Johnson noise .............................................................. 137
5.4.3.2 Shot noise .................................................................... 139
5.4.3.3 1/f noise ........................................................................ 139
5.4.3.4 Generation-recombination noise .................................. 140
5.4.3.5 Temperature fluctuation noise ...................................... 141
5.4.3.6 Photon noise ................................................................ 141
5.4.3.7 Microphonic noise ........................................................ 142
5.4.3.8 Triboelectric noise ........................................................ 142
5.4.3.9 CCD noises .................................................................. 142
5.4.3.10 Amplifier noise .............................................................. 143
5.4.3.11 Quantization noise........................................................ 143

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 viii Contents

5.4.4 Noise factor, noise figure, and noise temperature ............... 143
5.4.5 Some noise examples ......................................................... 144
5.4.6 Computer simulation of Gaussian noise.............................. 147
5.5 Thermal Detectors ........................................................................... 147
5.5.1 Thermal circuit ..................................................................... 147
5.5.2 Thermoelectric detectors ..................................................... 150
5.5.2.1 Basic principles ............................................................ 150
5.5.2.2 Combinations and configurations ................................. 153
5.5.3 Thermoresistive detector: bolometer ................................... 155
5.5.4 Pyroelectric detectors .......................................................... 157
5.5.4.1 Basic principles ............................................................ 157
5.5.4.2 Pyroelectric materials ................................................... 160
5.5.4.3 Operational characteristics of pyroelectric detectors ... 162
5.5.4.4 Applications of pyroelectric detectors........................... 162
5.5.5 Other thermal detectors....................................................... 163
5.6 Photon Detectors ............................................................................. 164
5.6.1 Detector materials ............................................................... 164
5.6.2 Photoconductive detectors .................................................. 169
5.6.2.1 Basic principles ............................................................ 169
5.6.2.2 Noises in photoconductive detectors ........................... 173
5.6.2.3 Characteristics of photoconductive detectors .............. 174
5.6.2.4 Applications of photoconductive detectors ................... 175
5.6.3 Photoemissive detectors ..................................................... 175
5.6.3.1 Basic principles ............................................................ 175
5.6.3.2 Classes of emitters....................................................... 176
5.6.3.3 Dark current ................................................................. 181
5.6.3.4 Noises in photoemissive detectors ............................... 182
5.6.3.5 Photoemissive detector types ...................................... 183
5.6.4 Photovoltaic detectors ......................................................... 185
5.6.4.1 Basic principles ............................................................ 185
5.6.4.2 Responsivity and quantum efficiency ........................... 195
5.6.4.3 Noises in photovoltaic detectors .................................. 196
5.6.4.4 Photovoltaic detector materials and configurations ...... 198
5.7 Imaging Arrays ................................................................................ 199
5.7.1 Introduction.......................................................................... 199
5.7.2 Photographic film................................................................. 199
5.7.2.1 History .......................................................................... 199
5.7.2.2 Physical characteristics ................................................ 201
5.7.2.3 Spectral sensitivity ....................................................... 201
5.7.2.4 Radiometric calibration................................................. 201
5.7.2.5 Spatial resolution.......................................................... 202
5.7.2.6 Summary ...................................................................... 202
5.7.3 Electronic detector arrays.................................................... 203
5.7.3.1 History .......................................................................... 203
5.7.3.2 Device architecture description and tradeoffs .............. 203

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Contents ix

5.7.3.3 Readout mechanisms .................................................. 204

5.7.3.4 Comparison .................................................................. 207
5.7.4 Three-color CCDs ............................................................... 207
5.7.5 Ultraviolet photon-detector arrays ....................................... 208
5.7.6 Infrared photodetector arrays .............................................. 209
5.7.7 Uncooled thermal imagers .................................................. 210
5.7.8 Summary ............................................................................. 211
For Further Reading .............................................................................. 211
References ............................................................................................ 213

Chapter 6 Radiometric Instrumentation / 215

6.1 Introduction...................................................................................... 215
6.2 Instrumentation Requirements ........................................................ 215
6.2.1 Ideal radiometer .................................................................. 215
6.2.2 Specification sheet .............................................................. 215
6.2.3 Spectral considerations ....................................................... 216
6.2.4 Spatial considerations ......................................................... 217
6.2.5 Temporal considerations ..................................................... 217
6.2.6 Make or buy?....................................................................... 218
6.3 Radiometer Optics........................................................................... 218
6.3.1 Introduction.......................................................................... 218
6.3.2 Review of stops and pupils.................................................. 218
6.3.3 The simplest radiometer: bare detector ............................... 219
6.3.4 Added aperture.................................................................... 219
6.3.5 Basic radiometer ................................................................. 221
6.3.6 Improved radiometer ........................................................... 223
6.3.7 Other methods for defining the field of view ........................ 224
6.3.8 Viewing methods ................................................................. 224
6.3.9 Reference sources .............................................................. 226
6.3.10 Choppers ............................................................................. 226
6.3.11 Stray light ............................................................................ 227
6.3.12 Summing up ........................................................................ 228
6.4 Spectral Instruments ....................................................................... 228
6.4.1 Introduction.......................................................................... 228
6.4.2 Prisms and gratings............................................................. 230
6.4.3 Monochromator configurations ............................................ 231
6.4.4 Spectrometers ..................................................................... 234
6.4.5 Additive versus subtractive dispersion ................................ 235
6.4.6 Arrays .................................................................................. 236
6.4.7 Multiple slit systems ............................................................ 236
6.4.8 Filters................................................................................... 236
6.4.9 Interferometers .................................................................... 237
6.4.10 Fourier transform infrared.................................................... 237
6.4.11 Fabry-Perot ......................................................................... 238

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 x Contents

For Further Reading .............................................................................. 240

References ............................................................................................ 240

Chapter 7 Radiometric Measurement and Calibration / 241

7.1 Introduction ...................................................................................... 241
7.2 Measurement Types ........................................................................ 241
7.3 Errors in Measurements, Effects of Noise, and Signal-to-Noise Ratio
in Measurements ............................................................................. 241
7.4 Measurement and Range Equations ............................................... 250
7.5 Introduction to the Philosophy of Calibration ................................... 253
7.6 Radiometric Calibration Configurations ........................................... 257
7.6.1 Introduction.......................................................................... 257
7.6.2 Distant small source ............................................................ 258
7.6.3 Distant extended source...................................................... 260
7.6.4 Near extended source ......................................................... 261
7.6.5 Near small source ............................................................... 262
7.6.6 Direct method ...................................................................... 262
7.6.7 Conclusion........................................................................... 263
7.7 Example Calculations: Satellite Electro-optical System .................. 263
7.8 Final Thoughts ................................................................................. 267
For Further Reading .............................................................................. 268
References ............................................................................................ 268

Table of Appendices / 269

Appendix A: Système Internationale (SI) Units for Radiometry and
Photometry ....................................................................... 271
Appendix B: Physical Constants, Conversion Factors, and Other Useful
Quantities.......................................................................... 275
Appendix C: Antiquarian’s Garden of Sane and Outrageous
Terminology ...................................................................... 277
Appendix D: Solid-Angle Relationships ................................................. 283
Appendix E: Glossary............................................................................ 285
Appendix F: Effective Noise Bandwidth of Analog RC Filters and the
Selection of Filter Parameters to Optimize Signal-to-Noise
Ratio ................................................................................. 297
Appendix G: Bandwidth Normalization by Moments ............................. 305
Appendix H: Jones Near-Small-Source Calibration Configuration ........ 309
Appendix I: Is Sunglint Observable in the Thermal Infrared? .............. 313
Appendix J: Documentary Standards for Radiometry and Photometry 321
Appendix K: Radiometry and Photometry Bibliography ........................ 341
Appendix L: Reference List for Noise and Postdetection Signal
Processing ........................................................................ 357

Index / 361

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Foreword
The material for this book grew out of a first-year graduate-level course,
“Radiometry, Sources, Materials, and Detectors,” that Jim Palmer created and
taught at the University of Arizona College of Optical Sciences for many years.
The book is organized by topic in a similar manner, with the first five chapters
presenting radiation propagation and system building blocks, and the final two
chapters focusing on instruments and their uses. Chapter 1 provides an overview
and history of the subject, and Chapter 2 presents basic concepts of radiometry,
including terminology, laws, and approximations. It also includes examples that
will allow the reader to see how key equations may be used to address problems
in radiation propagation. Chapter 3 introduces radiometric properties of materials
such as reflection and absorption, and Chapter 4 extends that discussion via a
detailed consideration of sources. Point and area detectors of optical radiation are
considered in Chapter 5, which also includes thermal and photon detection
mechanisms, imaging arrays, and a discussion about film.
In Chapter 6, the focus shifts to instrumentation. Concepts introduced in
Chapter 2 are here applied to instrument design. Practical considerations relating
to radiometer selection are detailed, and a “Make or Buy?” decision is explored.
Several monochromator configurations are examined, and spectral instruments
are discussed. Proceeding from instruments to their uses, Chapter 7 details types
of measurements that may be made with radiometric systems and provides a
discussion of measurement error. The philosophy of calibration is introduced,
and several radiometric calibration configurations are considered.
The material in the appendices covers a variety of topics, including
terminology, standards, and discussions of specific issues such as Jones source
calibration and consideration of solar glint. Due to Jim’s attention to detail and
the length of time over which he accumulated material, the long lists he provided
here may be viewed as comprehensive, if not current by today’s standards.
The level of discussion of the material is suitable for a class taught to
advanced undergraduate students or graduate students. The book will also be
useful to the many professionals currently practicing in fields in which
radiometry plays a part: optical engineering, electro-optical engineering, imagery
analysis, and many others.

In 2006, Jim Palmer was told that he was terminally ill, and he asked me to
complete this work. I was humbled and honored by the request. I’d met Jim as a
graduate student in optical sciences in the late 1980s, and he had served on my
thesis committee. My career after graduation had focused on systems engineering
and analysis, two areas in which radiometry plays a significant role. For nearly
the last ten years of Jim’s life, I’d been able to receive mentoring from the master
simply by showing up at Jim’s office door with a question or topic for discussion,
but I never anticipated that our discussions would one day come to an end. Upon
Jim’s death, I sought to weave his collection of resources and narrative together
xi

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 xii Forward

with newer material and discussion in a manner I hope will be both informative
to read and valuable to reference. The preface that follows was written by Jim
before he died and has been left as he wrote it.
I am grateful for the assistance of many. First is William L. Wolfe, Jim’s
professor and mentor, who offered helpful comments on each chapter and
adapted Chapter 6 on radiometric instrumentation. Others for whose help I am
grateful, all from or associated with the University of Arizona College of Optical
Sciences, are Bob Schowengerdt, who contributed the narrative on film; Anurag
Gupta of Optical Research Associates, Tucson, Arizona, who adapted the
appendix material; and L. Stephen Bell, Jim’s close friend and colleague, who
revised the signal processing discussion that appears in that section and provided
a complete bibliography on the subject. A special note of thanks goes to Eustace
Dereniak, who provided office space for me, helpful discussions, and hearty
doses of encouragement. I also wish to thank John Reagan, Kurt Thome (NASA
Goddard Spaceflight Center, Greenbelt, Maryland), Mike Nofziger, and Arvind
Marathay for review, discussion, and helpful insights. In addition, I am grateful
for the assistance of Anne Palmer, Jim’s beloved sister, and University of
Arizona College of Optical Sciences staff members Trish Pettijohn and Ashley
Bidegain. Gwen Weerts and Tim Lamkins of SPIE Press have my gratitude for
the special assistance they provided to this project. I also gratefully acknowledge
Philip N. Slater, my professor in optical sciences, who selected me as a graduate
student and trained me in remote sensing and absolute radiometric calibration
from 1987 to 1989, and Michael W. Munn, formerly Chief Scientist at Lockheed
Martin Corporation, who instilled the value of a systems perspective in the
approach to technical problems.
Finally, I am grateful to my family for providing financial support; to Ralph
Gonzales, Arizona Department of Transportation, and Sylvia Rogers Gibbons for
providing professional contacts; and my friends at Calvary Chapel, Tucson,
Arizona, whose donations and prayers sustained me as I worked to complete this
book.

Barbara G. Grant
Cupertino, California
October 2009

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Preface
This volume is the result of nearly twenty years of frustration in locating suitable
material for teaching the subject of radiometry and its allied arts. This is not to
say that there is not a lot of good stuff out there—it’s just not all in one place,
consistent in usage of units, and applicable as both a teaching tool and as a
reference. I intend this book to be all things to all people interested in radiometry.
The material comes from teaching both undergraduate and graduate-level courses
at the Optical Sciences Center of the University of Arizona, and from courses
developed for SPIE and for industrial clients. I have unabashedly borrowed the
tenor of the title from the superb text The Art of Electronics by Paul Horowitz in
the hope that this volume will be as useful to the inquisitive reader.
I gratefully acknowledge the contributions of my mentor, William L. Wolfe,
Jr., and the hundreds of students whose constant criticism and occasional faint
praise have helped immeasurably.
This book is dedicated to the memory of my mother, Candace W. Palmer
(1904–1996) and my father, James A. Palmer (1905–1990). She was all one
could wish for in a Mom, and he showed me the path to engineering.

James M. Palmer
1937–2007

xiii

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Chapter 1
Introduction to Radiometry
1.1 Definitions
Consider the following definitions a starting point for our study of radiometry:

radio- [<L. radius] a combining form meaning ray, raylike

-metry [Gr. -metria < metron] a terminal combining form meaning the process,
art, or science of measuring

radiometer (rä'dë-óm´í-ter) n. [radio- + -meter], an instrument for detecting

and measuring the intensity of radiant energy, by exposing to sunlight a set of
vanes blackened on one side and suspended on an axis in a vacuum and
measuring their speed of rotation (i.e., the mechanical energy into which the
radiant energy has been converted) (See Fig. 1.1)

radiometry ( rä'dë·óm´í·tri), n. the use of the radiometer: the measurement of

radiation1

These definitions are taken from Webster's New World Dictionary, and may
be satisfactory for the general nonscientist. The definitions are not satisfactory,
however, for scientists and engineers pursuing the art of radiometry. So let’s get
technical:

radiometry (rä'dë·óm´í·tri) n. the measurement of optical radiant energy

The practical electromagnetic spectrum extends from dc to frequencies

greater than 1020 Hz. The optical portion of the spectrum covers the five-decade
frequency range from 3 × 1011 to 3 × 1016 Hz, corresponding to the wavelength
range from 10 nm to 1000 μm, as shown in Fig. 1.2. This range includes the
ultraviolet, visible, and infrared regions. Shorter wavelengths are called x rays
and gamma rays, while longer wavelengths are microwave and millimeter radio
waves.
1

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 2 Chapter 1

Figure 1.1 Classic vane radiometer, commonly called the Crooke radiometer.1 [Reprinted
by permission from Webster’s Third New International® Dictionary, Unabridged ©1993 by
Merriam-Webster, Incorporated (www.Merriam-Webster.com)].

The optical radiation spectrum will be treated in this text, including the
ultraviolet, visible, and infrared regions. The visible portion of the optical
spectrum covers a rather narrow band of wavelengths between 380 nm and
760 nm; the radiation between these limits, perceivable by the unaided normal
human eye, is termed “light.” Measurements within this region may be called
“photometric” if the instruments used incorporate the response of the eye. The
short wavelength (ultraviolet) limit of radiometric coverage is about 200 nm,
approximately the shortest wavelength that our atmosphere will transmit. The
longest wavelength (infrared) treated in this book is about 100 μm. This
wavelength range includes 99% of the energy (95% of the photons) from a
thermal radiator at 0° C (273.16 K).

1.2 Why Measure Light?

But why measure light in the visible, ultraviolet, or even infrared region? What
are these measurements good for? Let's look at some historical perspectives:

I often say that when you can measure what you are speaking
about, and express it in numbers, you know something about it;
but when you cannot measure it, when you cannot express it in
numbers, your knowledge is of a meager and unsatisfactory
kind; it may be the beginning of knowledge, but you have
scarcely, in your thoughts, advanced to the stage of science,
whatever the matter may be.

Lord Kelvin

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Introduction to Radiometry

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

Figure 1.2 The electromagnetic spectrum.2 [Reprinted by permission of author from Optical Radiation Measurement series,
Vol. 1, F. Grum and R. J. Becherer, Radiometry, p. 1 (1979)].
3
 4 Chapter 1

...nobody will object to an ardent experimentalist boasting of his

measurements and rather looking down on the “paper and ink”
physics of his theoretical friend, who on his own part is proud of
his lofty ideas and despises the dirty fingers of the other.
Max Born

If you are really doing optics, you get photons under your
fingernails.

James M. Palmer

Measurement is the point at which the rubber meets the road. Hypotheses,
uncorroborated by measurement, cannot fulfill the same function. And if rubber
doesn’t meet the road, the car cannot move.
The measurement of light is often critical in transitioning from theory to the
development of systems and techniques. Although instrument and system design
may be based on theory, performance evaluation and system improvement
require that accurate radiometric measurements be applied. When calibrated
measurements are needed, that is, when field or laboratory measurements must
be correlated with specific values presenting the relationship between measured
phenomena and an absolute standard, radiometric measurements take on even
greater significance.

1.3 Historical Background

Scientists and engineers have been involved in the measurement of light since the
early experiments and instruments described by P. Bouguer in 1729 and by J. H.
Lambert in 1760. Exploration into other spectral regions began with the
discovery of the infrared region by W. Herschel in 1800 and the ultraviolet
region by J. W. Ritter the following year. Table 1.1 shows some of the significant
events in the history of radiometry and photometry.

Table 1.1 Some significant events in radiometry.

Year Event Principal investigator

? ... and then there was light! from Genesis
1666 Investigation of the visible spectrum Newton
1729 Inverse square law Bouguer
1760 Cosine law, exponential absorption Lambert
1800 Discovery of the infrared region Herschel
1801 Discovery of the ultraviolet region Ritter
1830 Radiation thermopile (first practical Nobile, Melloni
detector)
1837 Calorimetric detector Pouillet
1839 Photoelectric effect Becquerel
1859 Relation between absorption and emission Kirchhoff

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Introduction to Radiometry 5

Table 1.1 (Continued.)

Year Event Principal investigator

1860 Standard lamp fueld by sperm whale oil British Metropolitan
Gas Act
1879 Incandescent lamp (carbon filament) Edison
1880 Bolometer Langley
1892 Integrating sphere (theory) Sumpner
1893 Absolute thermal detector (pyrheliometer) Angstrom, Kurlbaum
1900 Blackbody radiation theory Planck
1900 Integrating sphere (reduction to practice) Ulbricht
1905 Photoelectric effect Einstein
1910 Tungsten lamp Collidge
1931 Adoption of colorimetric standards International
Commission on
Illumination (CIE)
1936 Photomultiplier tube (multistage) Zworykin, Morton &
Malter
1938 Pyroelectric detector (theoretical) Ta
1948 Adoption of platinum blackbody for Consultative
standard candela Committee on
Photometry and
Radiometry (CCPM)
1954 Silicon photodiode Chapin, Fuller &
Pearson
1955 Pyroelectric detector (reduction to practice) Chenowyth
1960 Invention of light amplification by Maiman
stimulated emission of radiation
(LASER)
1961 Tungsten-halogen lamp Zubler & Mosby
1970 Laser calorimetry West
1975 Fourier transform spectrometer Vanesse
1977 Photometry relegated to subset of CCPM
radiometry
1980 Self-calibrated silicon detector Zalewski & Geist
1983 Trap detector Zalewski & Duda
1984 Definitive measurement of the Stefan- Quinn & Martin
Boltzmann constant
1985 Cryogenic absolute radiometer Martin, Fox, & Key;
Foukal

1.4 Radiometric Measurement Process

This book describes the many facets of optical radiation measurement, from
radiation sources to detectors and signal processing. To fully understand and
appreciate a radiometric measurement, we must understand the processes of

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 6 Chapter 1

generation, transmission, and detection of optical radiation. In addition, we must

possess a firm grasp of the underlying mathematics and what is loosely called
“measurement science.”
A generic radiometric configuration is shown in Fig. 1.3. A target, or object
of measurement interest, can be either active, emitting radiation by virtue of its
temperature or some form of atomic excitation, or passive, reflecting radiation
from a different active or passive illuminator. Examples of active sources include
the sun, tungsten or fluorescent lamps, lasers, and any nontransparent object with
a temperature greater than 0 K. Passive sources include the entire range of natural
and artificial reflective surfaces that make up our environment. An additional
source of optical radiation can be classified as background, the radiation that may
be in our instrumental field of view along with the target. Also included is the
intervening medium, the atmosphere, which includes both radiation sources and
sinks, acting through the mechanisms of absorption, emission, and scattering.
The myriad of small arrows in Fig. 1.3 represent scattered, absorbed, emitted,
and reflected radiation.
After traversing the atmosphere, the rays from the target (and possibly the
background as well) reach our instrument, whose parameters define the ranges
over which the spatial, spectral, temporal, and radiometric characteristics of
incoming radiation will be accepted. This is accomplished through the use of
lenses and mirrors, choppers and apertures, prisms, gratings, filters, attenuators,

Figure 1.3 Generic radiometric configuration.2 [Reprinted by permission of author from

Optical Radiation Measurement series, Vol. 1, F. Grum and R. J. Becherer, Radiometry, p.
7 (1979)].

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Introduction to Radiometry 7

polarizers, optical fibers, etc. The optical radiation transmitted through the
instrument is finally incident on one or an array of detectors, transducers which
convert the incident optical radiation to a more tractable form of energy.
Detectors may be thermal (thermoelectric, bolometric, and pyroelectric) or
photon (photoemissive, photoconductive, and photovoltaic) in character; other
viable detectors include the human eye and photographic film.
The final block in the diagram involves signal processing. Most of the
detectors in common use generate electrical signals. Postdetector processing may
include filtering, linearization, and background suppression before the processed
result is recorded and displayed. Even the eye and film detectors include
processing steps, such as the filtering and interpretation of information by the
brain and the photographic development process for film.

1.5 Radiometry Applications

The fields in which radiometric instruments and techniques are used are very
diverse. Table 1.2 lists some of the more common applications.

Table 1.2 Common applications of radiometry.

Appearance measurement
Astrophysics
Atmospheric physics
Clinical medicine
Colorimetry
Diagnostic medicine
Remote-sensing satellites
Electro-optics
Illumination engineering
Laser measurements
Materials science
Meteorology
Military systems
Night-vision devices
Optoelectronics
Photobiology
Photochemistry
Photometry
Radiative heat transfer
Solar energy
Television systems
Visual displays
Vision research

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 8 Chapter 1

Most books on radiometry begin with a vast and often confusing array of
terms, definitions, etc. In this work, detailed listing of terminology is relegated to
the glossary in Appendix E, and radiometric terms will be introduced as they are
needed.
Radiometry and photometry are applied to a variety of phenomena whose
output occurs over many orders of magnitude. Tables 1.3 through 1.5 illustrate
the ranges of illumination encountered. “Luminance” is power per unit area and
unit solid angle weighted by the spectral response of the eye; its units are lumens
per square meter per steradian (lm/m2/sr), or candelas per square meter (cd/m2).
“Illuminance” is power per unit area weighted by the same function; its units are
lumens per meter squared (lm/m2).*

Table 1.3 Luminances of astronomical sources.

Source Luminance (cd/m2)

Night sky, cloudy, no moon 10–4
Darkest sky 4 × 10–4
Night sky, clear, no moon 10–3
Night sky, full moon 10–2
Clear sky 0.5 hr after sunset 0.1
Clear sky 0.25 hr after sunset 1
Cloudy sky at sunset 10
Gray sky at noon 102
Cloudy sky at noon 103
Moon 2.5 × 103
Average clear sky 8 × 103
Clear sky at noon 104
Solar disk 1.6 × 109
Lightning 8 × 1010

*
For more thorough discussions of photometry, see J. T. Walsh, Photometry, Dover, New York
(1958); P. Moon, The Scientific Basis of Illuminating Engineering, Dover, New York (1961); and
R. McCluney, Introduction to Radiometry and Photometry, Artech House, Boston (1994).

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Introduction to Radiometry 9

Table 1.4 Luminances of practical sources.

Source Luminance (cd/m2)

Minimum visible level (human) 3 × 10–6
Scotopic vision valid (human) < 0.003
Photopic vision valid (human) >3
Green electroluminescent 25
T8 cool white fluorescent 104
Acetylene burner 105
60 W inside frosted lamp 1.2 × 105
Candle 6 × 105
Sodium vapor lamp 7 × 105
High-pressure Hg vapor lamp 1.5 × 106
Tungsten lamp filament 8 × 106
Plain carbon arc crater 1.6 × 108
Cored carbon arc crater 109
Atomic fusion bomb 1012

Table 1.5 Illuminances of various sources.

Source Illuminance (lm/m2)

Absolute minimum Mv = 8 1.6 × 10–9
Typical minimum Mv = 6 10–8
0 Mv star outside atmosphere 2.54 × 10–6
Venus 1.3 × 10-4
Full moonlight 1
Street lighting 10
Recommended reading 102
Workspace lighting 102 to 103
Lighting for surgery 104

References

1. Webster’s Third New International® Dictionary, Unabridged, Merriam-

Webster, Inc. (1993).
2. F. Grum and R. J. Becherer, Radiometry, Vol. 1 in Optical Radiation
Measurements, F. Grum, Ed., Academic Press, New York (1979).

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Chapter 2
Propagation of Optical
Radiation
2.1 Basic Definitions

2.1.1 Rays and angles

A ray is represented by a vector: a straight line indicating the magnitude and
direction of propagation. A wavefront is a notional surface locally normal to a
ray. Thus, a wavefront could be a plane (all rays parallel, as if from infinity) or a
curved surface (indicating diverging rays, as if emanating from a point). Figure
2.1 illustrates rays and wavefronts for both cases.
A beam can be defined by two separated elements of area as shown in Fig.
2.2. It is thus the locus of possible rays that pass through the two areas separated
by distance d. As dA1 and dA2 approach zero, the beam approaches a single ray.
Since there is a small, but nevertheless finite, cross-sectional area associated with
a beam, it is capable of carrying power without the flux density (power per unit
area) approaching infinity.

(a) (b)

Figure 2.1 (a) Plane and (b) diverging wavefronts, with arrows indicating the direction of
radiation propagation.

11

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 12 Chapter 2

dA1 dA2
Figure 2.2 A beam between two area elements.

The speed of light is the rate at which light propagates through a vacuum. It
is represented by c and is a constant equal to about 3.00 × 108 ms–1, a faster rate
of travel than in any other medium. The index of refraction of a medium is the
ratio of the speed of light in vacuum to the speed of light in the medium:

c
n= , (2.1)
v
where

n = index of refraction of the medium, and

v = velocity of light in the medium.
Because c must be greater than or equal to v in Eq. (2.1), a medium’s
classical index of refraction is constrained to values equal to or greater than one.
Examples of nominal indices of refraction of various media are shown in Table
2.1.
Snell’s law allows one to calculate the angle change in propagation direction
upon refraction:
n1 sin θ1 = n2 sin θ2 . (2.2)

Note from Fig. 2.3 that the refracted ray is closer to the normal to the
boundary between media when the index of refraction is higher; that is, when
n2 > n1. Also note that a cone of light becomes narrower in a higher-index
medium.

Table 2.1 Indices of refraction of various media.

Medium n
Vacuum 1
Air 1.0003
Water 1.33
Quartz 1.45
Glasses 1.5–1.9
CaF2 1.42
Al2O3 1.75
ZnSe 2.4
Si 3.4
Ge 4.0

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Propagation of Optical Radiation 13

θ1
θ2
n1=1 n2=2

Figure 2.3 Illustration of Snell’s law.

Projected area is defined as the rectilinear projection of a surface of any

shape onto a plane normal to the surface’s unit vector. The differential form is
dAp = cosθ dA, where θ is the angle between the line of observation and the local
surface normal n. Integrating over the surface area we obtain

Ap =  cos θ dA . (2.3)
A

Some common examples of projected area are shown in Table 2.2. Figure 2.4
depicts the relationships between surface and projected area for a circle and a
sphere.
Plane angles and solid angles are both derived units in the SI system. Figure
2.5 depicts the plane angle θ, with l the arc length and r the circle radius. The
solid angle ω extends the plane-angle concept to three dimensions. It is the ratio
of the element of spherical area dAsph to the square of the sphere radius r2. Figure
2.6 illustrates a solid angle.
The unit of the plane angle is the radian, defined as:
The radian is the plane angle between two radii of a circle that cut off on
the circumference an arc equal in length to the radius.1
Table 2.2 Shapes and corresponding surface- and projected-area formulas.

Shape Surface area Projected area

Rectangle A = LW Ap= LW cosθ

Flat circle A = πr2 = [π(d2/4)] Ap = πr2 cosθ = [π(d2/4)] cosθ

Sphere A = 4πr2 = πd2 Ap= A/4 = πr2

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 14 Chapter 2

θ Ap (a)

Asph (b)
Ap

Figure 2.4 Surface- and projected-area relationships for a (a) circle and (b) sphere.

θ θ = l/r

Figure 2.5 Planar-angle relationships.

dAsph
r

ω = dAsph
r2

Figure 2.6 Solid-angle relationships.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Propagation of Optical Radiation 15

rsinθdφ
dφ rdθ

dAsph
dθ
θ

φ
rsinθdφ
rsinθ
Figure 2.7 An element of solid angle in spherical coordinates.2

In other words, l = r defines one radian. Since there are 2π radians in a circle, the
conversion between degrees and radians is 1 rad = (360/2π) = (180/π) degrees.
The steradian (sr) is defined in an analogous manner:

The steradian is the solid angle that, having its vertex in the center of a
sphere, cuts off an area of the surface of the sphere equal to that of a
square with sides of length equal to the radius of the sphere.1

Dividing the entire surface area of a sphere by the square of its radius, we
find that there are 4π sr of solid angle in a sphere, and 2π sr in a hemisphere.
Figure 2.7 depicts the solid angle in terms of the planar angle θ and the
rotational (azimuthal) angle φ, where r is sphere radius. The small element of
area dAsph lies on the surface of the sphere. The element of solid angle subtended
by dAsph is expressed as:

dAsph
dω = = sin θd θd φ . (2.4)
r2

To determine the value of the solid angle, integrate over θ and φ:

ω =   sin θd θd φ . (2.5)
φ θ

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 16 Chapter 2

Θ1/2

Figure 2.8 Right circular cone, with Θ1/2 the cone half angle.

In the most general case, a solid angle can subtend a surface of any shape.
However, in optical systems, which typically have circular apertures, the
envelope of the solid angle is a right circular cone, as shown in Fig. 2.8.
In Fig. 2.8, Θ1/2 is the cone half angle. It is the plane angle between the
centerline of the cone and anywhere on the edge of the cone. It is related to the
solid angle of the cone ω by:

2π Θ1/ 2
ω=  dφ
0
 sin θd θ ,
0
(2.6)

which results in

ω = 2π(1 − cos Θ1/ 2 ) . (2.7)

Application of a simple trigonometric identity provides the equivalent

expression
Θ
ω = 4π sin 2 1/2 . (2.8)
2

A feel for the magnitudes of various solid angles can be obtained by inspecting
Table 2.3. Table 2.4 facilitates conversion from steradians to other units.
Table 2.3 Some objects and corresponding solid angles.

Object Linear subtense Solid angle ω

Dime at 1 km 0.065 arcsec 3.1 × 10–13 sr
Jupiter (mean) 0.65 arcsec 3.1 × 10–11 sr
Cone in human eye 1 arcmin 2.7 × 10–7 sr
Sun, moon at earth’s surface 32 arcmin = 0.53 deg 6.8 × 10–5 sr
Tennis ball at 1 m 3.7 deg 0.0033 sr
Plane — 2π
Sphere (from the inside) — 4π

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Propagation of Optical Radiation 17

Table 2.4 Steradian conversions.

1 sr = 1 rad2
1 sr = 3282.8 deg2
1 sr = 1.1818 × 107 arcmin2
1 sr = 4.2545 × 1010 arcsec2

Both plane angles and solid angles are dimensionless quantities, and their use
can lead to confusion when attempting dimensional analysis. For example, the
simple inverse square law of irradiance (to be discussed in detail in Sec. 2.3.1),
E = I/d 2, appears dimensionally inconsistent. The left side has units W/m2, while
the right side has W/m2sr. It has been suggested that this equation be written E =
I Ωo/d2, where Ωo is the unit solid angle, 1 sr. Inclusion of the term Ωo will render
the equation dimensionally correct, but Ωo will far too often be considered a free
variable rather than a constant equal to 1, which leads to erroneous results.
Current practice suggests that another type of solid angle, the projected (or
weighted) solid angle, is more useful. The symbol for a projected solid angle is
Ω, and the units are also steradians. It is defined as the solid angle ω projected
onto the plane of the observer, as shown by the defining equation:

d Ω = d ω cos θ = sin θ cos θd θd φ . (2.9)

It involves an additional cosine term. This is depicted graphically in Fig. 2.9.

Figure 2.9 Projected solid-angle relationships. [Adapted from Radiometric Calibration:

3
Theory and Methods, C. Wyatt, p. 21 (1978)].

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 18 Chapter 2

To determine the projected solid angle, integrate over θ and φ:

Ω =   sin θ cos θ d θ d φ . (2.10)

φ θ

As before, the example most relevant to optical systems is the right circular
cone depicted in Fig. 2.8, for which the integral may be expressed as

2π Θ1/ 2
Ω =  d φ sin θ cos θ d θ , (2.11)
0 0

resulting in

Ω = π sin 2 Θ1/ 2 , (2.12)

where Θ1/2 is the cone half angle, as before. The plane area subtended by the cone
is Ωr2, and the spherical surface area subtended is ωr2. A couple of special cases
are worthy of mention. For a hemisphere, Ω = π sr, while an entire sphere
subtends 2π sr of a projected solid angle.
In addition to the dimensional concern raised above, there is another good
reason to employ two definitions of a solid angle. For many radiometric
problems, the emitter or receiver is flat, and the projected solid angle Ω is the
proper choice as it requires the inclusion of the cosθ term relating to the
projected area of the surface. In other cases, the emitter or receiver approximates
a point, emitting uniformly in all directions or responding to incoming radiation
equally from all directions. The solid angle ω is appropriate under these
conditions.
For a right circular cone at a half angle Θ1/2 of 90 deg, the projected solid
angle is π according to Eq. (2.12), and ω is 2π according to Eq. (2.7). At the
other extreme, when Θ1/2 is equal to 0 deg, both ω and Ω are zero. For small
angles, the solid angle and projected solid angle differ only by a cosine, are
nearly identical, and are therefore interchangeable in numeric value, if not in
concept. The error incurred by using ω rather than Ω is given in Table 2.5 (note
that ω > Ω for angles greater than 0 deg).
The definitions and symbols presented here have not been universally
applied in the past. One must be very cautious when reading the literature, as
different investigators use the terms and symbols solid angle ω and projected
solid angle Ω interchangeably, incurring predictable confusion and potentially
incorrect results.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Propagation of Optical Radiation 19

Table 2.5 Percentage error when not using a projected solid angle.

Θ1/2 (in deg) Error using ω rather than Ω

10 < 1%
16 < 2%
25 < 5%
35 < 10%
48 < 20%

2.1.2 System parameters

So far, we have defined only the terms necessary to describe the angles important
to a study of propagation. Now, let’s define quantities specific to optical systems.
The f-number (f/#) of an optical element or system is most often seen as the
ratio of the effective focal length f (object at infinity) of the optical element or
system to the diameter D of the entrance pupil:*

f
f #= . (2.12)
D

Good optical systems fulfill the Abbe sine condition with a spherical
wavefront converging to the focal point, and the preferred definition of f/# is

1
f #= . (2.13)
2sin Θ1/2

The numerical aperture (NA) of a system is defined as the sine of the vertex
angle (half angle) of the largest cone of meridional rays that can enter or leave an
optical system or element, multiplied by the refractive index of the medium in
which the vertex of the cone is located. It is generally measured with respect to
an object or image point, and will vary as that point is moved. The defining
equation is

NA = n sin Θ1/2 . (2.14)

Both f/# and numerical aperture can be related to projected solid angle:

*
For an informative discussion of f/#, see D. Goodman, “The f-word in optics,” Optics &
Photonics News, p. 38, April (1993).

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 20 Chapter 2

π
Ω= , (2.15)
4( f #) 2

and
π(NA) 2
Ω= , (2.16)
n2

and may be related to one another by

n
f /# = . (2.17)
2NA

Throughput, also known as geometrical extent or étendue, is the product of

the cross-sectional area of a beam and its projected solid angle

T = AΩ . (2.18)

Figure 2.10 shows the interaction between two differential area elements dA1
and dA2 a distance d apart. The area dA1 subtends a solid angle ω1, having apex at
dA2, while dA2 subtends a solid angle ω2, whose apex is at dA1. The surface
normals are shown in the figure by boldface n.
The throughput relationships become

dA1 cos θ1dA2 cos θ 2

T1→ 2 = d ω1dA2 cos θ 2 = 2
,
d
dA2 cos θ 2 dA1 cos θ1 (2.19)
T2→1 = d ω2 dA1 cos θ1 = 2
.
d

By inspection, T1→2 = T2→1 = T . The throughput is invariant.

ω1 ω2
dA1
dA2

d
n1 n2

Figure 2.10 Area- and solid-angle relationships used to define throughput. (Adapted from
a figure courtesy of William L. Wolfe.)

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Propagation of Optical Radiation 21

Figure 2.11 Invariance of throughput for a case in which the source image fills the
detector.

So far, our discussion has focused on theoretical constructs. Now, let’s

introduce some system elements as we proceed with demonstrating the
invariance of throughput in an optical system.
Consider Fig. 2.11. It represents a special case in which the image of the
source exactly fills the detector. This configuration consists of a source, a lens,
and a detector. In the figure, As is the area of the source, Ao the area of the optics,
and Ad the area of the detector. The projected solid angles are defined thus: Ωos is
the angle the optics subtend at the source, Ωso the angle the source subtends at the
optics, Ωdo the angle the detector subtends at the optics, and Ωod the angle the
optics subtend at the detector. For this case, the following equalities hold:

As Ωos = Ao Ω so = Ao Ω do = Ad Ωod . (2.20)

Since these pairings are equal, any of the above pairs can be chosen for
calculation purposes. If the image of the source does not exactly fill the detector
area, care must be taken to determine the proper ΑΩ product to use. The author’s
(Palmer) personal preference is the most often-used pair AoΩdo, inasmuch as the
entrance aperture size Ao and the field of view of the system Ωdo are determinable
characteristics of a radiometer. The next most often used pair is AdΩod, as the
detector size and the f/# of the optics are also measurable characteristics of the
radiometer.
Basic throughput is the name given to the quantity conserved across a
lossless boundary between two media having different indices of refraction. It
can be written as

n 2T = n 2 AΩ , (2.21)

and the relationship between medium 1 and medium 2 is

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 22 Chapter 2

n12T1 = n2 2T2 , (2.22)

where the subscripts denote the respective quantities in media 1 and 2. Like
throughput, basic throughput is invariant. Most optical systems have both the
object and the image located in the same medium, typically air, so basic
throughput is not often used.
Figure 2.12 depicts the appropriate and inappropriate area- and solid-angle
pairings used to define throughput. The correct area–solid angle pair is shown in
Fig. 2.12(a), the incorrect angle pair is shown in Fig. 2.12(b). Because the
definition of throughput includes two (projected) areas and the distance between
them, a correct pairing has the apex of its solid angle located at the correct area.
The incorrect pairing uses one area twice and ignores the other. The maxim “no
ice cream cones” should be applied.
Let’s look at some examples of the AΩ product. First, a spectrometer: Fig.
2.13 shows the area- and solid-angle product of the entrance slit (typically 1 × 10
mm) and the projected solid angle the collimating lens subtends at the slit, Ωls.
The AΩ product of a spectrometer is usually very small, and narrow spectral
bandwidths are typically employed. Therefore, it is difficult to get much light
through a spectrometer.
A different example of throughput may be found in the common camera. It is
related to the f/# of the lens and the size of the film. In this case, the detector
(film) size is not the overall dimension of the image, but the size of an individual
film grain. The smaller the f/#, the “faster” the camera and the greater the
throughput. Similarly, “fast” film has a larger grain area, permitting a higher
throughput and a shorter exposure time than “slow” film with a smaller grain
area.

(a) (b)
Figure 2.12 Right and wrong area–solid angle combinations for throughput determination.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Propagation of Optical Radiation 23

Slit, Lens
Area A

Ωls

Figure 2.13 Example of the correct area (A) and solid angle (Ωls) product used to
determine throughput in a spectrometer. (Adapted from a figure courtesy of William L.
Wolfe.)

2.1.3 Optical definitions

Some optical quantities are relevant to a study of radiometry, and they are
defined here. For more detailed treatment, the reader is advised to consult a text
on geometrical optics.†
Figure 2.14 depicts the location of object and image planes, along with some
key rays. In Fig. 2.15, the chief ray in an optical system originates at the edge of
the object and passes through the center of the entrance pupil (NP). It passes
through the center of the aperture stop (AS), the edge of the field stop (FS), and
the center of the exit pupil (XP), to the edge of the image, defining the image size
(height) and the lateral (transverse) magnification. There may be several
intermediate pupil planes in a complex optical system.

Image
Marginal Ray

Object

Chief Ray

Figure 2.14 Chief and marginal rays in an optical system, shown schematically.

†
Several excellent texts exist; a recent one is E. L. Dereniak and T. D. Dereniak, Geometrical and
Trigonometric Optics, Cambridge University Press, Cambridge (2008).

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 24 Chapter 2

NP
CHIEF RAY XP

MARGINAL RAY
AS FS
Figure 2.15 Optical system stops and pupils.

The marginal (rim) ray in an optical system is the ray from the object that
originates at the intersection of the object and the optical axis and passes through
the edge of the entrance pupil. It touches the edge (rim) of the aperture stop,
proceeds through the center of the field stop, and the edge of the exit pupil. The
marginal ray intersects the optical axis at the center of the image, defining the
location of the image and the longitudinal magnification. There may be several
intermediate image planes in a complex optical system.
The entrance pupil is the image of the aperture stop in object space (as seen
from the object), while the exit pupil is the image of the aperture stop in image
space (as seen from the image).
Of particular importance to radiometry are the aperture and field stops. The
aperture stop determines how much light may enter the system, while the field
stop determines the system’s angular field of view. In a simple system consisting
of a lens and detector at the rear focal point, the lens is the aperture stop, and
both the entrance and exit pupils are located at the lens, with the same size as the
lens. In more complicated systems, the stops may be internal and separated from
the pupils, as shown in Fig. 2.15.

2.2 Fundamental Radiometric Quantities

2.2.1 Radiance
The study of radiometry begins with fundamental units. Radiant energy has the
symbol Q and has as its unit the joule (J). Radiant power, also known as radiant
flux, is energy per unit time (dQ/dt), has the symbol Φ, and is measured in watts
(W). These definitions give no indication of the spatial distribution of power in
terms of area or direction. Radiance is the elemental quantity of radiometry,
power per unit area, and per unit projected solid angle. It is a directional quantity;
it can come from many points on a surface that is either real or virtual; and
because it is a field quantity, it can exist anywhere. The symbol for radiance is L
and the units are W/m2sr.
The defining equations are

 ΔΦ  d 2Φ d 2Φ d 2Φ
L = lim   = = = , (2.23)
ΔAs , Δω→ 0 ΔA cos θΔω
 s  dAs cos θd ω dAs d Ω dAp d ω

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Propagation of Optical Radiation 25

where θ is the angle between the normal to the source element and the direction
of observation as shown in Fig. 2.16.
Radiance is also associated with a source, either active (thermal or
luminescent) or passive (reflective), as discussed further in Chapter 4. Because
radiance may be evaluated at any point along a beam, it is associated with
specific locations within an optical system, including image planes and pupils.
Other radiometric units may be derived from radiance by integrating over
area and/or solid angle. Integration over solid angle yields irradiance (arriving at
a location, such as a sensor) or radiant exitance (leaving a location, such as a
source), both of which are expressed in W/m2. Integration over area yields
radiant intensity expressed in W/sr. Integration over both area and solid angle
yields radiant power in watts.
If rays are traced across a lossless boundary between two materials having
different indices of refraction as shown in Fig. 2.3, the solid angle changes
according to Snell’s law. Taking this change into account, the quantity Ln–2 is
seen to be invariant across the boundary. This quantity is called basic radiance. It
is useful for calculations when an object and its image are located in spaces with
different indices of refraction.
In the absence of sources or sinks along the path of a beam, power along a
beam is conserved. Since it was previously demonstrated that throughput is
conserved in an optical system, the radiance must also be invariant in order for
conservation of power (energy per unit time) to be obeyed. The results of this
invariance of radiance are significant:
(1) The radiance of the image at the detector plane of a camera (film or array
device) is the same as the radiance of the scene if there are no
transmission losses due to atmosphere and optics; and
(2) The radiance at the focal plane of a radiometer (imaging or point) is the
same as that of the target, if there are no transmission losses due to
atmosphere and optics.
Note that the transmission of atmosphere and optics is not likely to be unity
(perfect transmission); however, results (1) and (2) greatly simplify radiometric
calculations.

dω

θ
n
Figure 2.16 Radiance from area element dA, tilted at angle θ from surface normal, n.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 26 Chapter 2

The defining equation for radiance can be inverted and integrated over area
(and in the most general case, a projected solid angle) to determine the power in
an optical system:

Φ =  L dAs d ωs cos θ =  L dAs d Ω . (2.24)

2.2.2 Radiant exitance

Radiant exitance is radiation that exits a source. It is defined as power per unit
area radiated into a hemisphere (dΦ/dAs). The symbol for radiant exitance is M
and the units are W/m2. Its defining equation is

 ΔΦ  d Φ
M = lim  = . (2.25)
ΔAs → 0 ΔA
 s  dAs

The radiant exitance of a source is obtained by integrating radiance over the

projected solid angle of a hemisphere:

M =  L dΩ . (2.26)
π

Integrating radiant exitance in W/m2 over area results in radiant power Φ in

watts:

Φ =  M dAs . (2.27)
A

Figure 2.17 further illustrates the concept of radiant exitance. The relationship
between radiant exitance and radiance is complex, depending on the angular
distribution of radiance L(θ,φ). This is illustrated in Fig. 2.18 and described
mathematically in the following equations:

L ( θ, φ ) dA1 cos θ dA2

d2Φ = , (2.28)
r2

where r is the radius of the hemisphere, as shown in Fig. 2.18, dA1 is an element
of the emitting area, dA2 an element of the receiving area, and θ is the angle
between the surface normal of dA1 and the direction of propagation.
Making a substitution for dA2 using polar coordinates,

dA2 = r 2 sin θd θd φ
results in
d 2 Φ = L(θ, φ)dA1 sin θ cos θd θd φ . (2.29)

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Propagation of Optical Radiation 27

Figure 2.17 Radiant exitance from a source.

Applying limits for a hemisphere and constructing an integral results in

2π π /2
Φ =  dφ  L ( θ, φ ) A1 sin θ cos θd θ . (2.30)
0 0

Assuming that L is independent of direction (that is, the source at A1 is

Lambertian), then

π2
 sin 2 θ 
Φ = 2πLA1   . (2.31)
 2 0

Substituting the equality M = Φ/A and solving, we obtain

M = πL . (2.32)

The error most often committed in radiometry is in this conversion. Don’t

make the frequent mistake of forgetting the sine and using 2π!

r2sinθdθdφ=dA2

r
θ

dA1

Figure 2.18 Relationships between emitting and receiving areas used in deriving the
relationship between radiance and radiant exitance.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 28 Chapter 2

2.2.3 Irradiance
Irradiance (radiant incidence) is the opposite of radiant exitance, in that it is the
power per unit area that is incident on a surface. Its symbol is E and its units are
again W/m2:

 ΔΦ  d Φ
E = lim  = . (2.33)
ΔAs → 0 ΔA
 s  dAs

Irradiance on a surface is obtained by integrating radiance over the projected

solid angle of a hemisphere:

E =  Ld Ω . (2.34)
π

Integrating irradiance over area results in incident power

Φ =  EdA . (2.35)
A

This is shown conceptually in Fig. 2.19.

The most common example of irradiance is the solar “constant.” It is the
irradiance from the sun incident upon the earth’s atmosphere at the mean earth-
sun distance. Its numeric value is nominally Eo = 1368 W/m2, but it varies
slightly over the years, primarily due to sunspots. Also, the number 1368 is not
exact, and other values are often quoted, usually within 0.1% or so. The solar
irradiance at the earth’s surface, on the other hand, varies widely due to
atmospheric effects; a nominal value of 1000 W/m2 is often used.

Figure 2.19 Irradiance upon a surface.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Propagation of Optical Radiation 29

2.2.4 Radiant intensity

Radiant intensity is power emitted per unit solid angle in a specific direction.
Radiant intensity is frequently used when describing the radiation of an isotropic
(same in all directions) point source. Its symbol is I and its units are W/sr. Figure
2.20 conceptually illustrates the radiant intensity of a point source and a surface
illuminated by it. Equation (2.36) defines radiant intensity mathematically:

 ΔΦ  d Φ
I = lim  = . (2.36)
Δωs → 0 Δω d ωs
 s 

Intensity is derived from radiance by integrating over area

I =  LdA , (2.37)
A

and radiant power can be derived from intensity by integrating over solid angle
ω. (Note that the definition of intensity does not include an area term; therefore,
the integration is performed over solid angle, without a cosine projection.)

Φ =  Id ω . (2.38)
ω

In the case of the isotropic point source, the radiant power is 4π times the
intensity. Conversely, intensity may be found by dividing the total emitted power
Φ by 4π.
A word of warning: intensity is the most problematic radiometric quantity,
because it means different things to different people in different but related
fields. Laser physicists are prone to use it as the square of the electric field
strength, with units W/m2. Atmospheric scientists and heat transfer engineers use
the term “specific intensity,” frequently omitting the “specific,” to mean
W/m2sr. Some scientists and engineers equate intensity simply to power, and a
few use it to describe spectroscopic line strengths. Which usage is correct?

Figure 2.20 Radiant intensity from a point source.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 30 Chapter 2

Figure 2.21 Illustration of the inverse square law.

In answering the question, we might invoke precedent. Atmospheric scientists’

usage goes back to Chandrasekhar’s Radiative Transfer (1950)4. The physics
community usage goes back as far as Drude’s Lehrbuch der Optik (1900).5 Usage
in photometry and radiometry dates back to at least 1760 (Bouguer). But such
citing leads only to pointless argument. There is one compelling reason to
standardize our usage: intensity is an SI base quantity. Any other use of the term
intensity is incompatible with the SI and should be deprecated and abandoned.6,7,8

2.3 Radiometric Approximations

2.3.1 Inverse square law

There are a number of approximations in radiometry, most of them having to do
with radiation geometry. The inverse square law of irradiance is perhaps the best
known approximation. It states that the irradiance from an isotropic point source
varies inversely with the square of the distance from the source.
As shown Fig. 2.21, the rays are straight lines diverging from a source
having intensity I. At distance d, they fill area A. At a distance 2d, the length of
each side has doubled and the area subtended has increased by a factor of four.
Assuming a lossless medium, the amount of radiant power at the second surface
is the same as that at the first surface, due to the conservation of power. At the
second surface, power is spread over a larger area, and irradiance decreases. The
relationship between irradiance and distance is given by

Φ I
E= = . (2.39)
A d2

The relationship expressed by Eq. (2.39) presumes that the area shown in
Fig. 2.21 lies normal to the optical axis. If this is not the case and the surface is
tilted at an angle θ to the optical axis, a cosine θ factor must be included in the
equation:
I cos θ
E= . (2.40)
d2

This situation is illustrated in Fig. 2.22.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Propagation of Optical Radiation 31

I A

d θ
n
Figure 2.22 Area A tilted at angle θ from the axis.

Inverse square laws pervade all of physics, beginning with Newton’s

universal law of gravitation. Our radiometric application works well for small
sources observed at great distances. But how well may a real source approximate
an isotropic point source?
To avoid the question entirely, one might measure the source only in a well-
defined, specified direction. This approach will allow application of the inverse
square law, but results may not be repeatable if measurements are made in a
different direction. (In photometric terminology, source output under this
condition is termed “directional candlepower.”)
Another approach involves use of a “small” source rather than a point
source—once we determine what “small” is. If the maximum dimension of the
source is less than d/10, i.e., one-tenth of the distance from the source (equivalent
to about a 3-deg half angle), the inverse square law may be applied with an error
of less than 1%. If this is not good enough, a stricter criterion may be applied. If
the maximum source dimension is less than d/20 (1.5-deg half angle), the error in
applying the inverse square law is less than 0.1%.

2.3.2 Cosine3 law

The cosine3 law is applied to calculate the irradiance on a plane surface from an
isotropic point source, for example, the distribution of irradiance over a floor
from a bare light bulb suspended above. It is often convenient to describe the law
in terms of θ rather than distances along the surface, as the expression is
somewhat simpler.
Referring to Fig. 2.23, the irradiance at a point x directly below the source
with intensity I may be calculated according to the inverse square law as

I
Ex = .
D2

At position y, the distance from the source has increased such that d=D/cosθ.
If the target at y is perpendicular to the vector between source and target, the
irradiance at y is:

I I cos 2 θ
Ey = = .
d2 D2

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 32 Chapter 2

Is

θ
d D

y x

Figure 2.23 The cosine3 law.

If the target is now rotated so that it is parallel to surface x-y (the floor), an
additional cosθ term is introduced, resulting in

I cos3 θ
Ey = ,
D2

or more generally,
I cos3 θ
E= , (2.41)
D2

which is the cosine3 law. Note that two cosines were introduced to account for
the increased distance from the source; the third accrued from the projected area
of the target.

2.3.3 Lambertian approximation

Before discussing the final important approximation, another look at Lambertian
radiation is needed. A Lambertian source is one whose radiance L is independent
of direction:

L (θ,φ) = constant.

Thus, in the equation

d 2 Φ = L(θ, φ)dA cos θd Ω ,

the power received from a Lambertian radiator is proportional only to cosθ, the
angle of observation, and by extension, the projected area of the source.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Propagation of Optical Radiation 33

The Lambertian approximation is frequently used to describe the angular

distribution of radiation from a source. It is powerful because it significantly
reduces the complexity of the mathematical solutions to radiative transfer
problems, allowing them to be simplified so that calculation focuses only on
radiation geometry.
Fortunately, the Lambertian approximation is surprisingly good. Most
conventional blackbody radiation simulators (often used as infrared sources) and
integrating spheres (often used as visible sources) are very nearly Lambertian
over a wide range of angles. Reflective sources, also called “flat” or matte
reflectors (flat paint, matte white paper, etc.) are quite Lambertian out to nearly
grazing angles. Most natural surfaces, with the exception of still water, are highly
Lambertian. Surprisingly, polished metal surfaces, which are highly polarized in
emission at angles away from normal, are Lambertian out to 50 deg or so when
the total radiation is taken into account.
By contrast, specular reflecting surfaces, such as glossy paints, mirrors, still
water, etc., do not lend themselves to the Lambertian approximation.
It is common practice to utilize the Lambertian approximation, but make sure
to test the assumption with measurement whenever possible, and to keep track of
its effect on radiometric calculations.

2.3.4 Cosine4 law

The cosine4 law of irradiance is similar to the cosine3 law, except that a small
Lambertian source with radiance L replaces the isotropic point source. Such a
source might be the combination of a single lamp with a flat diffuser panel
oriented parallel to the ground, as shown in Fig. 2.24.
Since the basic geometrical layout is the same as that in the previous
example, start with the cosine3 law:
I cos3 θ
E= .
D2

Ls

θ
d
D

y x

Figure 2.24 The cosine4 law.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 34 Chapter 2

Recall that the projection of the target onto the floor added the third cosθ
term; the projection of the source area in the direction of point y adds the fourth.
For a Lambertian source, the relationship between intensity (point source)
and radiance (extended source) is I = LAproj. Substituting,

LAp cos 4 θ
E= , (2.42)
D2

which is the cosine4 law.

This fourth-power reduction in irradiance (falloff) applies to many, but not
all, sources that are extended and Lambertian. The effect is a noticeable dimming
of brightness near the image edges, often seen in images from projection systems
and wide-angle cameras. The effects of cosine3 and cosine4 falloff are shown
numerically in Table 2.6 for a variety of angles.
There are some cases in which a source may be “too extended” for the
cosine4 law to correctly predict irradiance. Consider the configuration shown in
Fig. 2.25, in which a circular source (shown edge on) radiates downward to the
observation plane. The angle θ is the angle of incidence to the surface, and angle
Θ1/2 is the half angle of the system viewing the source, related to the system’s f/#
by Eq.(2.13). Values of Θ1/2 for a variety of common f/#s are shown in Table 2.7.
Table 2.6 Fractional irradiance at observation point as a function of cosine factor.

Cosine Angle θ (degrees)

factor 0 10 20 30 40 50 60 70 80
cos θ 1 0.98 0.94 0.87 0.77 0.64 0.50 0.34 0.17
cos3θ 1 0.96 0.83 0.65 0.45 0.27 0.13 0.040 0.005
cos4θ 1 0.94 0.78 0.56 0.34 0.17 0.063 0.014 0.001

Circular source viewed edge on

Θ 1/2

Figure 2.25 Foote’s formula geometry.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Propagation of Optical Radiation 35

Table 2.7 System half angles as a function of f/#.

f/# Φ1/2 (deg.)

16 1.8
8 3.6
4 7.2
2.8 10.3
2 14.5
1.4 20.9
1 30

An expression for the irradiance at the sensor for this particular geometry has
been derived by Foote (1915).9,10 Foote’s formula is shown below, and the
corresponding plot using f/# as a parameter is shown in Fig. 2.26. For this
particular geometry, irradiance is least likely to follow the cosine4 law in lower
f/# systems. Note that at normal incidence (θ = 0), the equation reduces to
E = πLsin2θ.
 1 + tan 2 θ − tan 2 Θ1 2 
πL  
E= 1 − 1/ 2 
.
2   tan θ + 2 tan θ(1 − tan Θ1 2 ) + sec Θ1 2  
4 2 2 4
 

Figure 2.26 Foote’s formula results compared to Lambertian.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 36 Chapter 2

2.4 Equation of Radiative Transfer

The equation of radiative transfer is the most important equation in radiometry.
All axioms discussed in the previous section depend on it. It is, in differential
form:

L(θ, φ)dA1 cos θ1dA2 cos θ2

d 2 Φ1→2 = . (2.43)
d2

The physical situation described by the equation is shown in Fig. 2.27, where
dA1 and dA2 are the differential area elements, d is the distance between them,
and θ1 and θ2 are the angles between the normals to the area elements and the
optical axis.
As the equation indicates, the radiant power received at surface 2 depends
directly on the radiance L(θ,φ), the area elements, the angles between area
normals and the optical axis, and inversely on the square of the distance between
the surfaces. If the distance squared is much larger than the largest area, i.e.,
d 2 >> (A1 or A2), the differential areas may be replaced with the actual areas, and
Eq. (2.44) can be applied to calculate power:

L(θ, φ) A1 cos θ1 A2 cos θ2

Φ1→2 = . (2.44)
d2

Because d 2 is much greater than the size of any area element, the variations
in θ and φ are small, and the term L(θ,φ) need be evaluated at only one particular
set of angles. If either area is appreciable in relation to d 2, Eq. (2.44) cannot be
used.
The integral form of the equation of radiative transfer looks much more
formidable:

L(θ, φ) cos θ1 cos θ2

Φ1− 2 = 
A2 A1
d2
dA1dA2 . (2.45)

θ1 d

θ2
dA2
dA1

Figure 2.27 Radiation transfer geometry.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Propagation of Optical Radiation 37

To utilize this integral form, the following factors must be taken into
account:

(1) The angles θ1 and θ2 may vary from one part of area A1 or area A2 to
another.
(2) The distance d may also vary from one part of area A1 or area A2 to
another.
(3) The angular variation in radiance L(θ,φ) may be significant.

An additional implied assumption is that the principle of superposition is

applicable when adding up the small elemental contributions to radiant power; in
other words, the source providing the radiation is incoherent and interference
effects do not occur in the beam.
If the source is Lambertian, the radiance is independent of θ and φ, and the
equation of transfer becomes much simpler. It is

cos θ1 cos θ2
Φ1− 2 = L   dA1dA2 . (2.46)
A2 A1
d2

In this case, the transfer equation can be thought of as the product of a

radiance term L and the geometric term expressed by the double integral.
The transfer equation can be simplified even further if certain assumptions
can be made regarding the radiation geometry: first, that the square of the
distance d 2 is much larger than either area A1 or A2, and second, that both areas
are on axis, θ1 and θ2 are zero, and their cosines are therefore unity. If these
assumptions can be made in conjunction with the Lambertian approximation, the
result is the ultimate simplification:

Φ1→2 = LAΩ . (2.47)

A corollary expression to calculate flux density, power-per-unit area in a

system, is
E = LΩ . (2.48)

These simple equations are the logical starting points for all radiometric
engineering calculations, as they provide first-cut, back-of-the-envelope answers.
In many instances, they are all we need. In order to fully understand a particular
application, assumptions must be tested and the errors incurred by their use
assessed.
The Lambertian approximation is relatively simple to verify if we possess
hard data about the directionality of the source. The on-axis assumption is also
easy to verify. As discussed in Sec. 2.3, the distance between the two area
elements must be at least 10× the maximum linear dimension of the largest

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 38 Chapter 2

element in order for the inverse square law to be good to 1%. If the distance
between the two is increased to 20× the maximum dimension, the uncertainty in
applying the law reduces to 0.1% or less.

2.5 Configuration Factors

The concepts of radiative transfer are used not only in optics, but also in related
fields, such as thermal and illumination engineering. Equation (2.46) shows the
separation of the equation of radiative transfer into a radiance term and a
geometry term, assuming the Lambertian approximation may be applied. Other
names for the geometry term are view factor, interchange factor, shape factor,
form factor, and the term we will utilize, configuration factor. The symbol F is
used to denote this quantity, defined by Eq. (2.49):

Φ1→ 2
F= , (2.49)
Φ1

where Φ1 is the power leaving surface 1 and Φ1→ 2 is the power reaching surface
2 from surface 1. Both power terms are dimensionless. The radiant power terms
are further defined as
Φ1 = M 1 A1 ,

where M1 is the radiant exitance in W/m2 leaving surface 1, and A1 its area; and

M 1 cos θ1 cos θ2
π 
Φ1→ 2 = dA1dA2 ,
d2

where the radiance term outside the integral is obtained from Eq. (2.32), itself
dependent upon the Lambertian approximation. The fraction of radiant power
leaving surface 1 that arrives at surface 2 is

1 cos θ1 cos θ2
F12 =
πA1  d2
dA1dA2 ,

which is the configuration factor. The power transferred from surface 1 to surface
2 then becomes

Φ1→ 2 = Φ1 F12
= M 1 A1 F12
= πL1 A1 F12 .

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Propagation of Optical Radiation 39

Because the radiant power incident on surface 2 equals L1A1Ω21 by Eq.

(2.47),
Ω 21 = πF12 ,
and
Φ1→2 = L1 A1Ω 21 .

To restate: in engineering calculations, using the assumptions given above,

Φ1→2 = M 1 A1 F12 = L1 A1Ω 21 , (2.50)

and
E2 = L1Ω12 . (2.51)

The advantage of using configuration factors is that numerous solved

geometries appear throughout the literature. Some relevant to optics are shown in
Fig. 2.28.11 Further information on configuration factors may be found in Refs.
12, 13, and 14.

1 r 
2
Fd 1−2 = Fd 1− 2 = 
(h r )
2
+1 h

A2 A2
r
r

h
h

dA1

dA1
(a) (b)

Figure 2.28 Configuration factor examples: (a) Planar element parallel to circular disk,
and (b) planar element to sphere. (Adapted from Ref. 11.)

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 40 Chapter 2

2.6 Effect of Lenses on Power Transfer

Radiometer configurations will be discussed in detail in Chapter 6, but the effects
of lenses on power transferred to a detector will be introduced here. Look first at
Fig. 2.29. Two configurations are shown: A1 and A2 are the areas of the stops; D
is the distance from source to stop at A1; and S is the distance from A1 to A2 (at
the detector d). The difference between the two configurations is solely the
presence of the lens at A1 in the second. Expressions for power at the detector in
each case, for both point and extended sources, will be formulated. In both cases,
it will be assumed that the transmittance of the atmosphere between the source
and the detector is unity.
For the point source case, the irradiance at the detector in Fig. 2.29 is
expressed by the inverse square law, Eq. (2.39):

Is
Ed = ,
d2

where d is the distance between source and detector. Without a lens, [Fig.
2.29(a)], assuming no transmission losses in the intervening medium, the power
at the detector is simply the irradiance multiplied by the available sensitive area:

Φ d = Ed A2 . (2.52)

Expressed in terms of intensity, the power is

I s A2
Φd = , (2.53)
( D + S )2

where (D + S) is the source–detector distance. Note that in this configuration, A2

acts as the aperture stop, defining how much flux is collected, while A1, the field
stop, defines the detector’s field of view.

(a) (b)

Figure 2.29 Configurations (a) without and (b) with a lens.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Propagation of Optical Radiation 41

Adding a lens, as in Fig. 2.29(b), yields a different set of equations. In this

case, the power at the detector may be expressed as

Φ d = Ed A1 , (2.54)

or, more specifically, as

Φ d = τlens E A1 A1 , (2.55)

where τlens is the lens transmission.

In terms of intensity, it is:

τlens I s A1
Φd = . (2.56)
D2

Note that in this case, the aperture that limits the flux into the system is A1,
the aperture stop, and that A2 is the field stop, limiting the detector’s field of
view. The size of A2 is unimportant as long as it does not vignette the source’s
image at the detector.
The difference in received power between the two expressions in terms of
irradiance is expressed as

τlens A1
G= , (2.57)
A2

where G is the gain of power on the detector. To maximize G, make the lens
transmission and the area ratio as large as possible, while not vignetting the
source image.
To determine the effect of a lens on the same instrument configuration with
an extended source, begin with Eq. (2.47):

Φ d = Ls AΩ .

At area A = A2, the appropriate solid angle is subtended by area A1. This solid
angle is expressed as in Eq. (2.12) by

Ω12 = π sin 2 Θ1/ 2 ,

where Θ1/2 is the cone half angle. Assuming that it is small, the approximation
A1/S2 may be used for Ω12, so that

A1 A2
AΩ = , (2.58)
S2

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 42 Chapter 2

and
Ls A1 A2
Φd = . (2.59)
S2

The same approach can be pursued with the other area–solid angle
combination, that is, with A1 and Ω21. In that case, the solid angle is
approximately A2/S2, and Φd is again obtained by Eq. (2.59). Inserting a lens at A1
limits the power by the transmission of the lens, so that

τlens Ls A1 A2
Φd = . (2.60)
S2

The radiance-area–solid angle relationship holds true regardless of whether

the first or second area–solid angle combination is used to calculate throughput.
Inserting a lens yields no net gain in detector power for an extended source.
Rather, the power is less due to the nonunity (in the real world) transmission of
the lens.

2.7 Common Radiative Transfer Configurations

2.7.1 On-axis radiation from a circular Lambertian disc

This case is shown in Fig. 2.30. Assuming a lossless optical system, the flux
transferred from source to detector is given by Eq. (2.47), where L is the
Lambertian disc radiance. The area–solid angle pair we will use in this case is the
area of the detector Ad (in the figure) and the solid angle the source subtends at
the detector Ωsd, which may also be expressed by Eqs. (2.12) and (2.15) as

π
Ω sd = π sin 2 Θ1/2 = , (2.61)
4( f #) 2

where f/# is defined in Eq. (2.13). Considering the geometry in the figure,

a2
sin 2 Θ1/ 2 = .
(a 2 + b 2 )

We can now substitute in Eq. (2.48) to provide several equivalent

expressions for the irradiance at the detector:

a2 πL
Ed = πL sin 2 Θ1/ 2 = LΩ sd = πL = = πL( NA) 2 . (2.62)
(a + b ) 4( f #) 2
2 2

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Propagation of Optical Radiation 43

a Θ1/2
Ad
b
Ls

Figure 2.30 On-axis Lambertian disc, irradiance measured at detector of area Ad.

If the distance b is far greater than 2a, the linear dimension of the source,
then the inverse square law holds and Ed may be approximated as πLa2b–2. The
error incurred using a2b–2 rather than a2(a2 + b2)–1 is less than 1% when the
diameter-to-distance ratio is less than 0.1. Under these conditions, source
intensity Is may be substituted for πLa2 (the radiance times the area of the source)
so that

Is
I s = LAs  Ed = .
b2

Table 2.8 summarizes the relationships between source–detector distance and

irradiance at the detector for a variety of cases.

Table 2.8 Irradiance at detector as a function of source distance for a Lambertian disc.

Distance Half-angle Θ1/2 (deg) Irradiance Ed

b >> 2a very small πLa2b–2=Ib–2
b = 2a 26.5 πL/5
b=a 45 πL/2
b=0 90 πL

To determine the irradiance on the detector from an annulus (ring) rather

than a disc, calculate irradiances from discs having both outer and inner radii,
and subtract the latter from the former.

2.7.2 On-axis radiation from a non-Lambertian disc

In this case, source radiance is not independent of observation direction, and an
integration must be performed. The source’s radiant exitance may be obtained by

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 44 Chapter 2

a
Θ1/2
Ad

Ls

Figure 2.31 On-axis Lambertian sphere, irradiance measured from surface.

integrating Eq. (2.26) over a hemisphere, and the irradiance at the detector
calculated as a function of the half angle. To illustrate, take the relatively simple
example in which Ls = Locosθ. In this case, the radiant exitance is

M s = Lo  cos θd Ω , (2.63)

where dΩ is taken from Eq. (2.9). The resulting integral is then

2π π /2
M s = Lo  d φ sin θ cos 2 θd θ , (2.64)
0 0

which results in
2πLo
Ms = . (2.65)
3

Assuming a lossless medium, the irradiance at the detector is

2πLo
Ed =
3
(1 − cos3 Θ1/ 2 ) . (2.66)

In general, closed-form solutions are not readily available, and numerical

methods must be employed.

2.7.3 On-axis radiation from a spherical Lambertian source

If the disc is replaced with a Lambertian sphere of the same radius, as in Fig.
2.31, Eq. (2.62) may still be used, except that the sine squared of the half angle
now becomes

a2 a2
sin 2 Θ1/ 2 = = , (2.67)
(a + b) 2 a 2 + 2ab + b 2

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Propagation of Optical Radiation 45

Table 2.9 Irradiance at the detector as a function of source distance for a Lambertian
sphere, measured from a surface.

Distance Half-angle Θ1/2 (deg) Irradiance Ed

b >> 2a Very small πLa2b–2=Ib–2
b = 2a 19.5 πL/9
b=a 30 πL/4
b=0 90 πL

and the expression for irradiance at the detector is

a2
Ed = πL . (2.68)
a 2 + 2ab + b 2

Table 2.9 summarizes irradiance at the detector for a variety of cases. Note
that when b >> 2a, the inverse square law applies and the irradiance from the
sphere is the same as that from the disc, above.
If the source–detector distance is measured from the center of the sphere, as
shown in Fig. 2.32, the sine of the half angle is always a/b. The irradiance at the
detector is therefore

a 2 LAs I s
Ed = πL = 2 = 2. (2.69)
b2 b b

Equation (2.69) reveals an interesting result: the inverse square law holds for any
sphere and at any source–detector distance, as long as the surface is Lambertian
and the distance is measured from the center of the sphere. This counterintuitive
result simplifies calculation; results are shown in Table 2.10.

a
Θ1/2
Ad

Ls

Figure 2.32 On-axis Lambertian sphere, irradiance measured from the center.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 46 Chapter 2

Table 2.10 Irradiance at the detector as a function of source distance for a Lambertian
sphere, measured from the center.

Distance Half-angle Θ1/2 (deg) Irradiance Ed

b >> 2a very small πLa2b–2=Ib–2
b = 2a 30 πL/4
b=a 90 πL
b=0 — —

2.8 Integrating Sphere

The integrating sphere, invented by British scientist W. E. Sumpner in 1892, and
fully described by German scientist R. Ulbricht a few years later, is a device that
provides a spatially uniform source of radiance. It is depicted in Fig. 2.33, with
two elements of area inside the sphere labeled dA1 and dA2, the linear distance
between them d, and the sphere radius R.
To analyze the sphere’s behavior, we begin with the differential form of the
equation of radiative transfer, Eq. (2.43):

L(θ, φ)dA1 cos θ1dA2 cos θ2

d 2 Φ1→2 = .
d2

By inspection, θ1 = θ2 = θ and cosθ = d(2R)–1. Also assume that dA1 = dA2 =

dA. If the interior of the sphere is Lambertian, i.e., coated with material having
Lambertian properties, then

d Φ LdA
dE = = . (2.70)
dA 4 R 2

dA1

d
θ1

R θ2 dA2
R

Figure 2.33 The integrating sphere. (Adapted from Ref. 15 with permission from John
Wiley & Sons, Inc.)

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Propagation of Optical Radiation 47

This result means that irradiance within the sphere, for any area element dA,
is independent of position θ within the sphere and is dependent only on sphere
radius and radiance L. In other words, irradiance is constant over the sphere. This
fact makes the integrating sphere useful as a uniform radiance source.
If a source with power Φ is placed into the sphere (through a “port” in the
sphere), the radiance of the sphere wall L (assumed to be Lambertian) can be
determined as

Eρ
L= , (2.71)
π

where ρ is a property of the sphere coating material called its reflectance (to be
discussed in detail in Chapter 3). Combining Eqs. (2.70) and (2.71) and solving
for dE, we obtain

EρdA
dE = . (2.72)
4πR 2

This is the irradiance on an infinitesimal element of sphere area dA.

Integrating over the area to produce sphere irradiance is complex, as it must take
into account multiple reflections within the sphere. The bottom line is

ρΦ
E= . (2.73)
4πR 2 (1 − ρ)

This result is interesting because as the reflectance approaches unity, the

irradiance approaches infinity, as all the input power remains in the sphere. Real
sphere coatings are nonideal, however, with non-Lambertian surfaces and
reflectances less than one. Real spheres are fitted with ports and baffles, the
purpose of the latter to prevent “first pass” (unreflected) radiation from reaching
the detector. A useful equation for the radiance in a practical sphere is16

ρΦ
L= , (2.74)
πAsph [1 − ρ(1 − f ) ]

in which f is the ratio of the total port area to that of the sphere. (A sphere may
have several ports.) Thus, real spheres are not particularly efficient unless
reflectance is high and the total port area is kept small. Table 2.11 details some of
the many uses of integrating spheres.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 48 Chapter 2

Table 2.11 Some integrating sphere applications.

Uniform light sources Uniform detection systems

Measurement of transmission Measurement of reflectance
Depolarization Cosine receiver
Light source mixing Color mixing

2.9 Radiometric Calculation Examples

2.9.1 Intensities of a distant star and the sun

Figure 2.34 depicts a simplified configuration in which a distant star is viewed by
a telescope. Assuming that there are no atmospheric or optical system
transmission losses (i.e., that the mirror reflects 100% of the incident radiation),
that all power collected by the mirror is relayed to the detector, and that 10–6 W
are incident on the detector, we can determine the irradiance at the detector. We
can then use detector irradiance to calculate the star’s intensity.
As seen in Fig. 2.34, the system is f/2 with a focal length of 1 m. The mirror
diameter D is f/(f/#), or 0.5 m. The area of the mirror (assumed circular) is πD2/4.
The irradiance on this (perfectly reflecting) mirror is

Φ m 4 × 10−6
Em = = = 5 × 10−6 W/m 2 .
Am πD 2

The inverse square law applies due to the source distance, and as the source
is on axis, no cosine term is required. Inverting Eq. (2.39) to calculate intensity,
we have

I star = Em d 2 = 5 × 10−6 Wm −2 × (1012 m) 2 = 5 × 1020 W/sr .

I
DET

F/2,
f/2,
12
f =f=1m
1
d= 1012
d=10

Figure 2.34 Hypothetical distant star and system used to measure irradiance.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Propagation of Optical Radiation 49

Now, the intensity of our sun can be approximated by a spherical blackbody

source at 5750 K. Its radiance is given by the following expression (which will
be discussed in detail in Chapter 4):

σT 4
L= , (2.75)
π

and its value is 2 × 107 W/m2sr. Consider the geometry in Fig. 2.35, where Ap is
the projected area of the sun. The sun’s diameter is 1.4 × 109 m, so according to
Table 2.2, its projected area is (πdsun2)/4, or 1.54 × 1018 m2. At earth-sun distance
d = 1.5 × 1011 m, the solid angle subtended by the sun at the detector is Ωsd = 6.8
× 10–5 sr. Noting also that the sun subtends approximately 32 minutes of arc
(arcmin), Ωsd may also be calculated as πsin2(16 arcmin), which produces the
same result. The irradiance at the detector, area A in the figure (assumed to be
placed at the top of the atmosphere, therefore no atmospheric transmission loss),
is

Ed = LΩ sd = 1360 W/m 2 . (2.76)

Note that the diameter-to-distance ratio is substantially less than 0.1, and the
inverse square law may be applied. Calculating intensity as in the example
above:

I sun = Ed d 2 = 3.06 × 1025 W/sr . (2.77)

Ap
A

Figure 2.35 Source–detector geometry for solar irradiance calculation.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 50 Chapter 2

Irradiance may also be obtained in another way. Consider that the power
delivered to the detector with area A may be represented by

LAp A
Φ d = LAp Ω ds = , (2.78)
d2

where Ωds is the solid angle subtended by the detector, and that

Φ d LAp
Ed = = 2 = 1368 W/m 2 . (2.79)
A d

Applying the inverse square law as in Eq. (2.77), we obtain Isun = 3.08 × 1025
W/sr. Note that though the numbers are not identical, they are very close. The
value of 1368 W/m2 is referred to as the solar constant, and is specifically defined
as the irradiance falling upon a 1 m2 unit surface (hypothetical surface) at the
mean earth–sun distance. The solar constant has wide application in fields
including remote sensing. It is often given the symbol Eo.
Note that the total intensity of the sun has to do with the power radiated into
4π sr, the solid angle of a sphere as referenced in Table 2.3. Table 2.12 provides
relevant calculations related to solar power and intensity.

2.9.2 Lunar constant

2.9.2.1 Calculation
This concept is analogous to that of the solar constant, whose 1368 W/m2 are
incident upon the moon as well. If the moon is assumed to be Lambertian, with a
reflectance of 0.2, its radiance is

Eo ρ
Lmoon = = 87 W/m 2sr . (2.80)
π

Table 2.12 Solar quantities and their values.

Quantity Formula Value

Solar area Asun = 4Ap 6.16 × 1018 m2
Total solar radiant exitance M = πL 6.28 × 107 W/m2
Total solar power Φ = MA 3.87 × 1026 W
Total solar power (alternative) Φ = 4πI 3.87 × 1026 W
Intensity I = Φ/4π 3.08 × 1025 W/sr
Intensity (alternative) I = Eod2 3.08 × 1025 W/sr

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Propagation of Optical Radiation 51

At the earth’s surface, the angular subtenses of the moon and the sun are the
same, approximately 32 arcmin. This means that ΩME, the solid angle subtended
by the moon at the earth, is equal to Ωsd, above, with a value of 6.8 × 10–5 sr.
Neglecting the relatively minimal distance between the top of earth’s atmosphere
and its surface, the irradiance produced by the moon at the top of earth’s
atmosphere is

Φ M →E
Emoon , TOA = = Lmoon Ω ME = 5.9 × 10−3 W/m 2 . (2.81)
ApE

Note that in the above equation, ApE is the projected area of the earth,
analogous to the projected area of the sun discussed earlier.

2.9.2.2 Moon–sun comparisons

Comparing irradiances from the sun and moon, we have

Esun  Eo   1368 W/m 2 

=   2 
= 2.3 × 105 .
Emoon  Eo , moon  5.9 × 10 −3
W/m 

Also, comparing radiances we find

Lsun 2 × 107 W/m 2sr

= 2
= 2.3 × 105 .
Lmoon 87 W/m sr

The numbers are the same because the solid angles subtended are the same.
Assuming an atmospheric transmission of 0.75, the solar irradiance at the
earth’s surface is this factor multiplied by the solar constant

Eearth = τEo = 1026 W/m 2 ,

and assuming an earth reflectance of 0.2 along with the Lambertian

approximation, the earth’s radiance is

Eearth ρ
Learth = = 65 W/m 2sr .
π

Applying an atmospheric transmittance of 0.75 to the moon’s radiance at the

top of the atmosphere, we obtain the moon’s apparent radiance; that is, its
radiance when viewed from the ground:

L′moon = τLmoon = 65 W/m 2 sr .

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 52 Chapter 2

This interesting result means that the radiance of an “average” sunlit scene‡ is
the same as the apparent radiance of the moon. It also means that in photography,
the same exposures can be used to photograph both. Exposure parameters should
be set during the daytime and applied to night photography. If the moon is
photographed through a telescope, exposures should be increased to compensate
for transmission losses within the instrument. In addition, given the factor of
2.3 × 105 difference between solar and lunar irradiances, photographing a
moonlit scene requires significantly longer exposures than are needed for
daylight illumination.
A point about assumptions should be made, specifically, that the moon is not
a strict Lambertian surface. It is somewhat retroreflective, as though covered
with Scotchlite.™ Simple measurements made by Palmer with a silicon detector
indicate that the apparent intensity of the full moon is more than ten times that of
the quarter moon. When viewed with a telescope or binoculars, the edge appears
a bit brighter than the rest. The lunar surface is dusted with small glassy
spheroids, ejecta from meteorite collisions. Its reflectance is approximately 0.08,
somewhat less at shorter wavelengths and somewhat more at longer wavelengths.

2.9.3 “Solar furnace”

This example concerns a “solar furnace” operated in space, delivering power to a
collector just outside the earth’s atmosphere, but the equations are valid for any
source located at a large distance from a collector. What is the irradiance
delivered to the target Et?
Consider Fig. 2.36, in which the sun is represented by the vertical bar at the
left-hand side with radiance [Eq. (2.75)] of 2 × 107 W/m2sr. The power from the
sun to the collector is Φsc = LsAcΩsc, using the area of the collector and the solid
angle the sun subtends at the collector. The irradiance at the collector is then

Φ s →c
Ec = = Ls Ω sc ,
Ac

where Ωsc is the solid angle the sun subtends at the collector, 6.8 × 10–5 sr.
Choosing a target diameter (or linear dimension) and system focal length f so that
Ωsc = Ωtc, we have

At
Ωtc = and At = Ωtc f 2 .
f2

‡
Eastman Kodak has shown through extensive research that the reflectance of an average scene is
18%; all exposure meters are calibrated using this assumption (J. M. Palmer, 2005).

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Propagation of Optical Radiation 53

Ac Ωtc

Ls = 2 x 107 Wm–2sr–1
At
Ωsc=Ωtc
f
Figure 2.36 The “solar furnace.”

Assuming no transmission loss between collector and target, the irradiance at

the target Et is expressed as

Φ c →t Ec Ac E A
Et = = = c c2 .
At At Ωtc f

Therefore, the target irradiance is the product of the source radiance and solid
angle the collector subtends at the target. That solid angle may also be
characterized [Eq. (2.15)] as

π
Ω ct = ,
4( f #) 2

so that

πLs
Et = , (2.82)
4( f #) 2

which provides a way of characterizing target irradiance in terms of both source

radiance and the f/# system parameter, for a configuration such as this one.

2.9.4 Image irradiance for finite conjugates

The definition of f/# presented earlier was for an object at infinity; however,
many systems operate at finite conjugates. Figure 2.37 depicts such a system, in
which neither image nor object is at infinity. In such cases, a “working f/#,” often
symbolized as f/#′, is used.17

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 54 Chapter 2

As Ac At

Figure 2.37 Finite conjugates.

A working f/# is defined as

 m 
f # ′ = f # 1 +  , (2.83)
 m
 p 

where magnification m is the ratio of image height to object height, and has
values between 0 and infinity. The term mp is pupil magnification, the ratio of the
diameter of the exit pupil to the diameter of the entrance pupil, and has values
between 0.5 and 2. For a single lens or mirror, it is always 1. Substituting f/#′ for
f/# in Eq. (2.82) we obtain

πL
Et = 2
. (2.84)
 m 
4 ( f # ) 1 +
2
 mp 
 

If mp = 1, as it frequently does, Eq. (2.84) becomes the camera equation

πL
Et = . (2.85)
4 ( f # ) (1 + m )
2 2

Table 2.13 shows two important cases.

Table 2.13 Target irradiance using the camera equation.

Case I – Object at infinity Case II – Equal conjugates

m=0 m=1
πL πL πL
Et = Et = =
4 ( f #) 4 ( f # ) (1 + 1) 16 ( f # )
2 2 2 2

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Propagation of Optical Radiation 55

The expressions for irradiance at the target show us that image irradiance
decreases as the in-focus object is moved closer to the camera. In order to
maintain focus, the detector must be moved backward, which decreases the solid
angle of the lens as seen from the detector Ωct by a factor of four.

2.9.5 Irradiance of the overcast sky

A reasonable value for the radiance of the overcast sky is 50 W/m2sr, somewhat
less than the 65 W/m2sr calculated above for a typical sunlit scene. Assuming
that the sky radiance is constant, with no brightening at the horizon, the
irradiance from the sky at the earth’s surface is

Φ sky →earth 2π π/ 2
Eearth = = Lsky Ω sky −earth = Lsky  d φ  sin θ cos θd θ
Aearth 0 0
2
Eearth = πLsky sin (90 deg)
Eearth = 157 W/m 2 .

This is a factor of 6 or 6.5 less than the irradiance received from the sun on a
clear day (1000 W/m2), which explains why flat-plate solar collectors continue to
function well on a cloudy day (provided that the clouds are “conservative”
scatterers.) By comparison, on a clear day, the diffuse solar irradiance (excluding
the direct beam) can be as high as 50 to 100 W/m2 due to scattering.

2.9.6 Near extended source

A near extended source such as the one shown in Fig. 2.38 may be found in the
laboratory. It provides a nice way to calibrate a radiometer, because:

(1) If the image of the extended source overfills the field of view of the
detector with area Ad, the distance d is unimportant;
(2) If the source is Lambertian, the angle between source and optical axis is
unimportant; and
(3) If the detector or radiometer with area Ad is not placed exactly at the
focal distance f, it doesn’t matter.

The power Φd on area Ad is calculated according to

Ao πDo 2
Φ d = Ls Ad Ωod = Ls Ad 2
= Ls Ad ,
f 4f2
which equates to

πLs Ad
Φd = .
4( f #) 2

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 56 Chapter 2

Ls
θ
Ad

d f
Figure 2.38 Near extended source.

2.9.7 Projection system

Figure 2.39 depicts two different designs for projection systems. The Abbe
projector was invented first, and has significant disadvantages. As can be seen
from the diagram, the source is imaged onto the slide, which is then imaged onto
the screen. Hot spots can occur at the slide, resulting in smoke.
The Koehler system is superior. The source is imaged into the projection
lens, a pupil location rather than an image location. The slide is positioned in an
area of relatively uniform brightness, allowing for a more uniform image on the
screen.

Figure 2.39 Two projection systems.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Propagation of Optical Radiation 57

The equation for illuminance on the screen resembles the camera equation,
Eq. (2.85), with the addition of a cos4θ term to account for the off-axis angle to
the screen as seen from the projector:

τo πLv cos 4 θ
Ev = , (2.86)
4 ( f # ) (1 + m )
2 2

where Ev is illuminance. The transmission of the optical system το appears also.

To maximize irradiance for a given magnification, there are only two
possibilities: minimize the f/# or maximize the radiance of the source. Candidate
sources with high radiance values include tungsten lamps, tungsten-halogen
lamps, carbon arcs, xenon arcs, metal-halide lamps, and high-brightness
phosphor screens.

2.10 Generalized Expressions for Image-Plane Irradiance

2.10.1 Extended source

To provide a more general expression for the irradiance at the image plane from
an extended source, several factors must be added to the expression in Eq. (2.85).
First is a cosn term, accounting for the reduction in irradiance as we look off axis.
Its value is typically 4 to account for projections of the source and target areas,
but good optical designers can reduce this factor to 3.18 Next, losses in the optical
system due to transmission, reflection, and scattering may be combined into the
general term τo as discussed above. A term to account for vignetting, fv, the
reduction in the cross-sectional area of the beam as the off-axis angle is
increased, applies as well. Finally, to account for the presence of a central
obscuration in a system such as a Cassegrain, the factor (1 – A2) is applied, where
A is the ratio of the diameter of the central obscuration to the diameter of the
primary mirror. (If there is no central obscuration, this factor can be eliminated.)
Considering the above terms and using the most general expression for
source radiance, the expression for image-plane irradiance from an extended
source becomes

τo πf v (1 − A2 ) L(θ, φ) cos n θ
Et = . (2.87)
4 ( f # ) (1 + m )
2 2

Note that this expression does not take into account the spectral nature of the
radiation, to be discussed in greater detail in Chapter 4, nor does it account for
the transmission of the atmosphere.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 58 Chapter 2

2.10.2 Point source

If we assume an isotropic source, apply the factors for vignetting, etc., mentioned
above, and begin with Eq. (2.41) cosine3 law, we obtain

τo f v (1 − A2 ) I cos n θ
Et = , (2.88)
d2

where n is 3 for an isotropic source.

2.11 Summary of Some Key Concepts

This chapter has presented a number of concepts fundamental to an
understanding of radiometry; a short summary of some appears below.
First, the basic equation of radiative transfer in differential form:

L(θ, φ)dA1 cos θ1dA2 cos θ2

d 2 Φ1→2 = .
d2

Second, the integral form of this equation:

L(θ, φ)cos θ1 cos θ2

Φ1→2 = 
A2 A1
d2
dA1dA2 .

Handy simplifications can be utilized, provided that their underlying

assumptions are met. These are

Φ = LAΩ ,
and
E = LΩ .

Further, the Lambertian approximation is okay for most emitters, but angles
must be considered when applying it to metals. It is alright for matte reflectors,
and no good at all for specular reflectors.
The choice of solid angle for radiometric calculations is an important one.
When the source or receiver is isotropic, solid angle ω may be used. When the
source or receiver is Lambertian, projected solid angle Ω is the correct choice.
A final comment on sources deserves mention. These “notes” (lyrics by Jon
Geist and Ed Zalewski, NIST, ca. 1982) are to be sung to the theme song from
the ancient television series, “Mr. Ed:”

A source is a source, of course, of course

And no one can make a point of a source
Unless, of course, it’s the sort of source
That only exists in your head!

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Propagation of Optical Radiation 59

For Further Reading

F. C. Grum and R. J. Becherer, Radiometry, Vol. 1 in Optical Radiation
Measurements series, F. Grum, Ed., Academic Press, New York (1979).
R. McCluney, Introduction to Radiometry and Photometry, Artech House,
Boston (1994).
W. L. Wolfe, Introduction to Radiometry, SPIE Press, Bellingham, Washington
(1998).
E. F. Zalewski, “Radiometry and Photometry,” Chapter 24 in Handbook of
Optics Vol. 2: Devices, Measurements, and Properties, Second Edition, M.
Bass, Ed., Optical Society of America, Washington, D.C. (1994).

References

1. Guide for the Use of the International System of Units (SI), NIST SP811,
U.S. Government Printing Office (1995).
2. F. E. Nicodemus, Self-Study Manual on Optical Radiation Measurements:
Part I–Concepts, NBS Technical Note 910-1, p. 68, NBS, U.S. Government
Printing Office, Washington, D.C. (1976).
3. C. Wyatt, Radiometric Calibration: Theory and Methods, Academic Press,
New York (1978).
4. S. Chandrasekhar, Radiative Transfer, Clarendon Press, Oxford (1950),
reprinted Dover (1960).
5. P. Drude, Lerhbuch der Optik (1900), translated into English and reprinted as
The Theory of Optics, Longmans, Green, New York (1902).
6. B. N. Taylor, The International System of Units (SI), NIST SP330, U.S.
Government Printing Office, Washington D.C. (1991).
7. J. M. Palmer, “Getting intense on intensity,” in Metrologia 30(4), pp. 371–
372 (1993).
8. J. M. Palmer, “Intensity,” in Optics & Photonics News, p. 6, February
(1995).
9. P. D. Foote, “Illumination from a radiating disc,” Bulletin of the Bureau of
Standards, NBS, 12, p. 583 (1915).
10. R. Kingslake, Optical System Design, Academic Press, New York (1983).
11. B. T. Chung and P. S. Sumitra, “Radiation shape factors from plane point
sources,” J. of Heat Transfer 94(3), pp. 328–330 (1972).
12. E. M. Sparrow and R. D. Cess, Radiation Heat Transfer, Brooks/Cole,
Belmont, California (1970).
13. P. Moon, The Scientific Basis of Illuminating Engineering, McGraw-Hill,
Dover, New York (1936).

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 60 Chapter 2

14. M. Donabedian, “Cooling systems,” Chapter 15 in The Infrared Handbook,

W. L. Wolfe and G. L. Zissis, Eds., U.S. Government, Washington, D.C.
(1978).
15. R. W. Boyd, Radiometry and the Detection of Optical Radiation, Wiley &
Sons, New York (1983).
16. “A guide to integrating sphere theory and applications,” Labsphere Inc., at
http://www.labsphere.com/tecdocs.aspx (2006).
17. R. Kingslake, Optical System Design, Academic Press, New York (1983).
18. P. N. Slater, Remote Sensing: Optics and Optical Systems, Addison-Wesley,
Reading, Massachusetts (1980).

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Chapter 3
Radiometric Properties of
Materials
3.1 Introduction and Terminology
When radiant flux is incident upon a surface or medium, three processes occur:
reflection, absorption, and transmission. A fraction of the beam is reflected,
another fraction is absorbed, and the remainder is transmitted. Transmittance τ is
the ratio of transmitted power to incident power. Reflectance ρ is the ratio of
reflected power to incident power. Absorptance α is the ratio of absorbed power
to incident power.
Figure 3.1 shows an ideal geometric case, where the transmitted and reflected
components are either specular (regular, in the mirror direction) or diffuse
(scattered into the hemisphere). Figure 3.2 shows the transmission and reflection
for real surfaces. Both spectral and directional properties are important.

RETROREFLECTION
DIFFUSE
INCIDENT TRANSMISSION
BEAM
INCIDENT
BEAM

DIFFUSE
REFLECTION

SPECULAR SPECULAR (REGULAR)

REFLECTION TRANSMISSION

Figure 3.1 Idealized reflection and transmission.

61

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 62 Chapter 3

STRONG DIFFUSE STRONG SPECULAR STRONG RETRO-

REFLECTION REFLECTION REFLECTION

REGULAR TRANSMISSION DIFFUSE TRANSMISSION

Figure 3.2 Generalized reflection and transmission.

A continuing dialog over terminology has taken place, particularly over the
suffixes -ance and –ivity. 1,2,3,4,5 The usage here reserves terms ending with -ivity
(such as transmissivity, absorptivity, and reflectivity) for properties of a pure
material, while the suffix –ance is used when the characteristics of a specific
sample are described. One can then distinguish between the reflectivity of pure
aluminum (as calculated from the complex index of refraction n and κ) and the
reflectance of a particular specimen of 6061 aluminum with surface structure
associated with rolling or machining and with a natural oxide layer.
The adjective spectral refers to a characteristic at a particular wavelength and
is indicated as a function of wavelength λ, i.e., τ(λ), ρ(λ) or α(λ). For example,
spectral transmittance τ(λ) is often plotted against wavelength λ for a colored
filter. The absence of “spectral” implies integration over all wavelengths,
weighted by a source function.

3.2 Transmission
Transmission is the process by which incident radiant flux leaves a surface or
medium from a side other than the incident side (usually the opposite side). The
spectral transmittance τ(λ) of a medium is the ratio of the transmitted spectral
flux Φλt to the incident spectral flux Φλi:

Φ λt
τ(λ) = . (3.1)
Φ λi

Total transmittance τ is the ratio of the total transmitted flux Φt to the total
incident flux Φi:

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Radiometric Properties of Materials 63

∞
Φ  τ ( λ ) Φ λ i dλ
τ= t = ≠  τ(λ) dλ .
0
∞
(3.2)
Φi

λ
Φ λ i dλ
0

Note particularly that the total transmittance is not the integral over
wavelength of the spectral transmittance; it must be weighted by the incident
source function Φλi.
The transmittance may also be described in terms of radiance as follows:

τ=
  0 2 π sr
Ltλ dΩt dλ
, (3.3)
∞
  0 2 π sr
Liλ dΩi dλ

where Lλi is the spectral radiance Lλi(λ;θi,φi) incident from direction (θi,φi), Lλt is
the spectral radiance Lλt(λ;θt,φt) transmitted in direction (θt,φt), and dΩ is the
elemental projected solid angle sinθ cosθdθdφ.
Geometrically, transmittance can be classified as specular, diffuse, or total,
depending upon whether the specular (regular) direction, all directions other than
the specular, or all directions are considered. The bidirectional transmittance
distribution function (BTDF, symbol ft and units sr–1) relates the transmitted
radiance to the radiant incidence (irradiance) as

dLt ( θt , φt ) dLt ( θt , φt )
f t ( θi , φi ; θt , φ t ) ≡ = . (3.4)
dEi ( θi , φi ) Li ( θi , φi ) d Ωi

This descriptor facilitates specification of the angular dependence of all (specular

and diffuse) transmitted radiation.

3.3 Reflection
In reflection, a fraction of the radiant flux incident on a surface is returned into
the hemisphere whose base is the surface containing the incident radiation. The
reflection can be specular (in the mirror direction), diffuse (scattered into the
entire hemisphere), or a combination of both. Table 3.1 shows a wide range of
materials that have different goniometric (directional) reflectance characteristics.
Spectral reflectance is defined at a specific wavelength λ as

Φ λr
ρ( λ ) = , (3.5)
Φ λi

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 64 Chapter 3

Table 3.1 Goniometric classification of materials.6 (Reprinted with permission of CIE.)

Material γ
Scatter σ Structure Example
classification (deg)
Exclusively none 0 ≅0 none Mirror
reflecting
materials micro Matte aluminum
weak ≤0.4 ≤27
macro Retroreflectors
τ=0 none Lacquer & enamel
coatings
micro Paint films, BaSO4,
strong > 0.4 > 27
PTFE
macro Rough tapestries, road
surfaces
Weakly transmitting,
Sunglasses, color filters,
strongly reflecting none 0 ≅0 none
cold mirrors
materials
micro Matte-surface color
weak ≤ 0.4 ≤ 27 filters
τ ≤ 0.35 macro Glossy textiles
none Highly turbid glass
strong > 0.4 > 27 micro Paper
macro Textiles
Strongly none 0 ≅0 none Window glass
transmitting
materials none Plastic film
micro Ground glass
weak ≤ 0.4 ≤ 27
τ > 0.35 macro Ornamental, prismatic
glass
none Opal glass
micro Ground opal glass
strong > 0.4 > 27 macro Translucent acrylic
plastic with patterned
surface

while the total reflectance ρ is the ratio of the reflected flux Φr to the incident
flux Φi:

∞
Φr  ρ(λ) Φ λi d λ
ρ= = ≠  ρ( λ ) d λ .
0
∞
(3.6)
Φi

λ
Φ λi d λ
0

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Radiometric Properties of Materials 65

As in the case of transmittance, above, the integrated reflectance is not the

integral over wavelength of the spectral reflectance; it must be weighted by the
incident source function Φλi.
Reflectance factor R or R(λ) is defined as the ratio of (spectral) radiant flux
reflected by a sample to the (spectral) radiant flux which would be reflected by a
perfect diffuse (Lambertian) reflector under the same irradiation conditions.
While reflectance (in the absence of luminescence) cannot exceed unity,
reflectance factor can assume values from zero to nearly infinity. Since the
reference is a perfect diffuse reflector, reflectance factor is only useful as a
descriptor for diffuse surfaces. Equations for nine types of reflectance factor
appear in Table 3.2.
Some notes on Table 3.1 are in order:

(1) “Structure” refers to the nature of the surface. In a microscattering

structure, the scatterers cannot be resolved with the unaided eye. The
macrostructure scatterers can be readily seen.
(2) Sigma, σ, is a diffusion factor, the ratio of the mean of radiance
measured at 20 deg and 70 deg to the radiance measured at 5 deg from
the normal, when the incoming radiation is normal.
σ = [L(20) + L(70)] / [2L(5)]. It gives an indication of the spatial
distribution of the radiance, and is unity for a perfect (Lambertian)
diffuser.
(3) Gamma γ is a half-value angle, the angle from the normal where the
radiance has dropped to one half the value at normal. Its value is 60 deg
for a perfect (Lambertian) diffuser.
(4) It is suggested that the diffusion factor is appropriate for strongly
diffusing materials and that the half-value angle is better suited for
weakly diffusing materials.

No single descriptor of reflectance suffices for the wide range of possible

geometries. The fundamental geometric descriptor of reflectance is the
bidirectional reflectance distribution function (BRDF) fr. It is defined as the
differential element of reflected radiance dLr in a specified direction per unit
differential element of irradiance dEi, also in a specified direction.7 It carries the
unit of sr–1.

dLr (θr , φr ) dLr (θr , φr ) −1

f r ( θi , φi ; θ r , φ r ) ≡ = sr . (3.7)
dEi (θi , φi ) Li (θi , φi )d Ωi

As shown in Fig. 3.3, the polar angle θ is measured from the surface normal,
z. The azimuth angle φ is measured from an arbitrary reference in the surface
plane, most often the plane containing the incident beam. The subscripts i and r
refer to the incident and reflected beams, respectively.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 66 Chapter 3

dΩr
dΩi
θr
θi

dA y

φi
φr
x

Figure 3.3 Geometrical definitions for BRDF (Adapted from Ref. 7).

Nicodemus et al.7 integrated over various solid angles and applied the earlier
work of Judd8 to obtain nine goniometric reflectances and nine goniometric
reflectance factors. These are listed in Tables 3.2 and 3.3. In these tables, the
term “directional” refers to a differential solid angle dω in the direction specified
by (θ,φ). “Conical” refers to a cone of finite extent centered in direction (θ,φ); the
solid angle Ω of the cone must also be specified. The reflectances are illustrated
in Fig. 3.4.

Table 3.2 Nomenclature for nine types of reflectance factor.7

Bidirectional
R (θi , φi ; θr , φr ) = πf r ( θi , φi ; θr , φr )
reflectance factor

Directional-conical π
R(θi , φi ; ωr ) =  f r ( θi , φi , θr , φr )d Ω r
reflectance factor Ωr ωr

Directional-
hemispherical R(θi , φi ;2π) =  f r ( θi , φi , θr , φr )d Ω r
2π
reflectance factor

Conical-directional π
R (ωi ; θr , φr ) =  f r ( θi , φi , θr , φr )d Ωi
reflectance factor Ωi ωi

Biconical π
R(ωi ; ωr ) =   f r ( θi , φi , θr , φr )d Ω r d Ωi
reflectance factor* Ωi ⋅ Ω r ωi ωr

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Radiometric Properties of Materials 67

Table 3.2 (Continued.)

Conical- 1
hemispherical R (ωi ;2π) =
Ωi  
ωi 2π
f r ( θi , φi ; θ r , φ r ) d Ω r d Ω i
reflectance factor*
Hemispherical-
directional R (2π; θr , φr ) =  f r ( θi φi ; θ r φ r ) d Ω i
2π
reflectance factor
Hemispherical- 1
conical reflectance R (2π; ωr ) =
Ωr  
2 π ωr
f r ( θi , φi ; θ r , φ r ) d Ω r d Ω i
factor*
Bihemispherical 1
π 2 π 2 π
R (2π;2π) = f r (θi , φi ; θr , φr )d Ω r d Ωi
reflectance factor
*
Configurations that are measurable in practice.

Table 3.3 Nomenclature for nine types of reflectance.7

Bidirectional
dρ(θi , φi ; θr , φr ) = f r ( θi , φi ; θ r , φ r ) d Ω r
reflectance
Directional-conical =  f r ( θi , φi ; θ r , φ r ) d Ω r
ρ(θi , φi ; ωr )
reflectance ωr

Directional-
hemispherical ρ(θi , φi ;2π) =  f r (θi , φi ; θr , φr )d Ω r
2π
reflectance
Conical-directional d Ωr
Ωi ωi
dρ(ωi ; θr , ϕr ) = f r ( θi , φi ; θ r , φ r ) d Ω i
reflectance
Biconical 1
Ωi ωi ωr
ρ(ωi ; ωr ) = f r ( θi , φi ; θ r , φ r ) d Ω r d Ω i
reflectance*
Conical- 1
hemispherical ρ(ωi ;2π) =
Ωi  
ωi 2π
f r ( θi , φi ; θ r , φ r ) d Ω r d Ω i
reflectance *
Hemispherical- d Ωr
directional dρ(2π; θr , φr ) =
π  2π
f r (θi φi ; θr φr )d Ωi
reflectance
Hemispherical- 1
π 2 π ωr
conical ρ(2π; ωr ) = f r (θi , φi ; θr , φr )d Ω r d Ωi
reflectance*
Bihemispherical 1
π 2 π 2 π
ρ(2π;2π) = f r (θi , φi ; θr , φr )d Ω r d Ωi
reflectance
*
Configurations that are measurable in practice.

In both Tables 3.2 and 3.3, configurations containing a directional term are
considered theoretical, as dΩ→0.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 68 Chapter 3

Figure 3.4 Nine reflectance configurations.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Radiometric Properties of Materials 69

3.4 Absorption
Absorption is the process in which a fraction of the incident radiant flux is
converted to another form of energy, usually heat. Absorptance is the fraction of
incident flux that is absorbed. Spectral absorptance is defined at a specific
wavelength λ as

Φλa
α (λ ) = , (3.8)
Φλi

with the subscripts denoting absorbed and incident power, respectively. Total
absorptance is defined as

∞
Φ  α ( λ ) Φ λi d λ
α= a = ≠  α (λ ) d λ .
0
∞
(3.9)
Φi

λ
Φ λi d λ
0

Note the analogy with Eqs. (3.2) and (3.6).

Absorption removes power from a beam; directional characteristics such as
direct absorption and bulk scattering are not often taken into consideration.

3.5 Relationship Between Reflectance, Transmittance, and

Absorptance
Because energy is conserved, the sum of the transmission, reflection, and
absorption of flux incident on a surface is unity, or

τ + ρ + α =1. (3.10)

The above statement assumes integration over all wavelengths and directions. In
the absence of wavelength-shifting effects (such as luminescence or Raman
scattering), this relationship is also valid for any specific wavelength:

τ(λ) + ρ(λ) + α(λ) = 1. (3.11)

3.6 Directional Characteristics

3.6.1 Specular transmittance and reflectance

The specular transmittance and reflectance for a single surface can be calculated
via the Fresnel equations using the complex index of refraction n + iκ. The
simplest case is that of no absorption, i.e., κ = 0. The general equations are

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 70 Chapter 3

2
 n′ cos θ − n cos φ 
ρp =  
 n′ cos θ + n cos φ 
2
, (3.12)
 n cos θ − n′ cos φ 
ρs =  
 n cos θ + n′ cos φ 
and
2
 2n cos θ   n′ cos φ 
τp =    
 n′ cos θ + n cos φ   n cos θ 
2
, (3.13)
 2n cos θ   n′ cos φ 
τs =    
 n cos θ + n 'cos φ   n cos θ 

where the subscripts p and s represent the two polarization states, n and θ are on
the incident side of the interface, and n′ and φ are on the transmitted (or reflected)
side. The total transmittance and reflectance for unpolarized light is the average
of the two polarized components
ρ p + ρs
ρT =
2
and (3.14)
τ p + τs
τT = .
2

Figure 3.5 shows reflection and transmission curves for a single surface of a
nonabsorbing optical material with an index of refraction of 2. The three curves
represent s- and p-polarization states as well as total polarization. To compute the
curves for absorbing media, substitute n ± iκ for n in Eqs. (3.12) and (3.13),
where κ = αλ 4π . Since refractive index n is wavelength dependent, the
calculated reflectance and transmittance are also.
In Fig. 3.6, we see a partially transparent plane slab of an optical material.
Reflection, transmission, and absorption are all present.
The Fresnel equations are greatly simplified at normal incidence, in which
θ = φ = 0. For a nonabsorbing material, the reflectance and transmittance at a
single surface are

2
 n′ − n  4nn′
ρss =   τss = . (3.15)
( n′ + n )
2
 n′ + n 

The internal transmittance τi of a piece of optical material describes only the

absorption component and neglects the reflectance losses. The exponential
absorption law, often referred to as the Lambert-Bouguer-Beer law, is

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Radiometric Properties of Materials 71

1
SINGLE SURFACE REFLECTANCE 0.9

0.8

0.7

0.6

(a) 0.5

0.4
S
0.3 TOTAL

0.2
P

0.1

0
0 10 20 30 40 50 60 70 80 90
ANGLE FROM NORMAL (deg)

1
SINGLE SURFACE TRANSMITTANCE

0.9
TOTAL P
0.8

0.7

0.6
(b) 0.5 S

0.4

0.3

0.2

0.1

0
0 10 20 30 40 50 60 70 80 90
ANGLE FROM NORMAL (deg)

Figure 3.5 (a) Reflection and (b) transmission of a single surface, n = 2.

τi (λ) = e −α′ ( λ ) x , (3.16)

where α′(λ) is the spectral absorption coefficient (cm–1) at the specified

wavelength, and x is the thickness (cm). Different units can be used for thickness
and absorption coefficient (i.e., m, km, mm, μm) as long as they are the same for
both the absorption coefficient and the thickness. The product of the absorption
coefficient and the thickness x is often called the optical thickness τo. These units
and symbols are used differently in different fields; be careful!
External transmittance is the quantity that is ordinarily measured, and
includes the Fresnel reflection losses and the absorption. Equation (3.17)
describes the transmittance of a parallel slab at normal incidence with single-
surface Fresnel reflection ρss, absorption coefficient α′, and thickness x:

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 72 Chapter 3

Φi

Φr Φt
Φa

Figure 3.6 Transmitted, reflected, and absorbed rays.

Φ t (1 − ρss ) e (1 − ρss ) τi .
2
−α′ x 2

τ= = 2 −2 α′ x
= (3.17)
Φi 1 − ρss e 1 − ρss 2 τi2

If the slab is nonabsorbing (α = 0), then τi = 1, and Eq. (3.17) reduces to

2n
τ= . (3.18)
n2 + 1

Similar equations can be derived for reflectance and absorptance. For

reflectance:
ρ (1 − ρ ss ) e −2 α′ x ρss (1 − ρss ) τi2
2 2
Φr
ρ= = ρss + ss = ρ ss + . (3.19)
Φi 1 − ρss 2 e −2 α′ x 1 − ρss 2 τi2

For absorptance:

Φ a (1 − ρss ) (1 − e ) (1 − ρss )(1 − τi )

−α′ x

α= = = . (3.20)
Φi 1 − ρss e −α′ x 1 − ρ ss τi

When the optical thickness is large enough, the material becomes opaque and
the transmittance goes to zero. In this case, the reflectance ρ approximates the
single surface reflectance ρss, and the absorptance α approaches 1 – ρ. For the
opposite case, in which the optical thickness approaches zero, the material
becomes transparent and the following relationships hold:

ρss (1 − ρss )
2

ρ = ρss + , (3.21)
1 − ρss 2

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Radiometric Properties of Materials 73

(1 − ρss )
2

τ= . (3.22)
1 − ρss 2

As an example, an ordinary transparent glass (n = 1.5) has a single-surface

reflectance ρss of 0.04. The total reflectance ρ is 0.077 (rather than 0.08), and the
total transmittance τ is 0.923.

3.6.2 Diffuse transmittance and reflectance

A large part of the optical radiation in our environment is the result of reflections,
with sources including sunlight and artificial tungsten and fluorescent lamps. For
nonspecular reflection or transmission, the BRDF or its analog, the BTDF, may
be characterized for any combination of incident and reflected beams. A full
BRDF or BTDF specification is very complex, particularly when the reflectance
and transmittance lie out of the plane that includes the incident and specular
beams. Figure 3.7 illustrates this complexity.9
The set of polar diagrams in this figure for a partially diffuse surface is
specific to just one incidence angle. A complete characterization would require
data at other incidence angles as well, as the pattern is variable. Surfaces become
more specular as the angle of the incoming beam increases towards grazing
incidence. In addition, the diffuse scatter from rough surfaces diminishes with
increasing wavelength, i.e., the surface becomes more specular.
The BRDF characteristics of a surface are normally plotted as BRDF (sr–1)
versus angle as measured from the specular beam. This method places the
specular beam on the left ordinate, regardless of incidence angle. Positive and
negative angles as measured from the specular beam are typically shown on the
same graph, and angles as large as 175 deg are seen (for an 85-deg angle of
incidence). The angle axis can be linear for diffuse surfaces, but a log plot is
better suited for specular surfaces as shown in Fig. 3.8
For some materials, it has been found that the diffusely reflected radiation is
symmetrical about the specular beam. In this case, a single plot suffices for each
wavelength. The materials with this characteristic are spatially uniform (called
isotropic in the literature, but different from isotropic as applied to a point
source) and are either nearly specular or nearly perfectly diffusing. Insight into
their BRDF characteristics can be gained through a plot of the natural log of the
BRDF versus a special parameter, (β – βo), where β is the sine of the scattered
beam angle and βo is the sine of the specular angle. This places the BRDF into
direction cosine space. Figure 3.9 shows such a BRDF plot for a perfectly diffuse
reflector (fr = 1/π) and for a very good specular reflector. The plot for the perfect
specular reflector would be a delta function of infinite height at (β – βo) = 0, and
would not be seen on this plot. What is actually seen is the instrument function
that is primarily the convolution of the incident beam profile with the detected
beam profile.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 74 Chapter 3

Figure 3.7(a) BRDF of rough aluminum at incidence angle of 33 deg. (Reproduced from
Ref. 9 with permission.)

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Radiometric Properties of Materials 75

Figure 3.7(b) BRDF of rough aluminum at incidence angle of 33 deg. (Reproduced from
Ref. 9 with permission.)

Figure 3.8 Conventional BRDF plot of perfect mirror (“instrument”) compared with a
perfect diffuse reflector and two mirrors with scatter.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 76 Chapter 3

Figure 3.9 Special parameter versus BRDF plot of perfect mirror (“instrument”) compared
with a perfect diffuse reflector and two mirrors with scatter.

A perfect specular reflector has a BRDF equal to ρ/Ωi where ρ is the

reflectance and Ωi is the solid angle subtended by the source. For example, a
piece of opaque glass with a reflectance ρ of 0.05 in sunlight (Ωi = 6.8 × 10–5 sr)
has a BRDF of 14706 × ρ = 735.3. To determine the radiance associated with this
specular reflection, multiply the BRDF by the irradiance. The BRDF for a
hemispherical source is ρ/π, the same as for a perfectly diffuse (Lambertian)
reflector. In the confines of an integrating sphere, diffuse and specular samples
having the same reflectance are indistinguishable.

3.7 Emission
So far, we have considered the radiometric properties of materials with respect to
incoming radiation. In fact, all materials above 0 K radiate, so the emission of
radiation by a material is an important property, as well. The Infrared Handbook
defines emissivity as “the ratio of the radiant exitance or radiance of a given body
to that of a blackbody.”10 Its symbol is ε. Emissivity may be considered a
“quality” factor, indicating the capability for thermal radiation by a material. It
has both spectral and directional properties, it is dimensionless, and its values are
between 0 and 1. As with the material properties already discussed, emissivity
refers to the characteristics of a pure substance, while emittance refers to the
properties of a specific sample.
Spectral emittance ε(λ) is defined as emittance at a given wavelength, and it
is not a derivative quantity. In the case that a radiator is neutral with respect to
wavelength, with a constant spectral emittance less than 1, it is called a graybody.
In that case, the spectral emittance is the ratio of the radiance of that source at
that wavelength to the radiance of a blackbody at that wavelength:

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Radiometric Properties of Materials 77

Lλ
ε(λ ) = , (3.23)
LλBB

where LλΒΒ is the value of the Planck function at that wavelength for a
blackbody. Further discussion of this function and the radiometric characteristics
of sources follows in Chapter 4.
As reflectance, transmittance, and absorptance are related, as indicated by
Eq. (3.10), so too are reflectance, transmittance, and emittance related. At
equilibrium, the power emitted by a body to its surroundings must equal the
power absorbed by the body from its surroundings. More succinctly, the body’s
absorptance must equal its emittance:

α = ε. (3.24)

This is Kirchhoff’s law, to be discussed in more detail in Chapter 4. As a

consequence, emittance may be substituted for absorptance in Eqs. (3.10) and
(3.11), so that

ε =1− τ − ρ , (3.25)

and spectrally,

ε(λ) = 1 − τ(λ) − ρ(λ) . (3.26)

If the body is opaque (τ = 0), then

ε =1− ρ , (3.27)

and
ε(λ) = 1 − ρ(λ ) . (3.28)

3.8 Spectral Characteristics

The radiometric properties of materials of interest all share one spectral
characteristic: the property of interest is not independent of wavelength. Many
material characteristics may be “flat” over a portion of the spectrum, but at other
wavelengths may differ significantly. Since these properties are weighting
functions, the values are irrelevant in those spectral regions where the source
function is insignificant. For example, ordinary window glass has a transmittance
of about 0.92 (clean, normal incidence) in the visible part of the spectrum but
drops to zero in the infrared where the eye is nonresponsive.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 78 Chapter 3

Figure 3.10 Spectral emittance for several generic surfaces.11

These materials are often used for temperature control of spacecraft. Heating
is by absorption of sunlight for wavelengths shorter than 3 μm, and cooling
results from thermal radiation for wavelengths longer than 3 μm. If a material has
high reflectance at shorter wavelengths, it will not absorb much of the incident
radiation. If the reflectance is low at longer wavelengths, the absorption and
consequently the thermal emission will be high. The surface will be cold. If the
spectral regions are reversed, the surface will become hot. These surfaces are
known as selective surfaces, and a wide range of surface temperatures have been
achieved.
Designers of these materials often utilize the ratio α/ε to describe the value of
absorptance in one spectral region (usually solar) relative to the emittance value
in another region (usually infrared.) Figure 3.10 shows the spectral emittance
(1 – spectral reflectance) for several generic surfaces; some are selective.
The Infrared Handbook provides detailed examples of the radiometric
properties of both natural and artificial sources. An example of the spectral
reflectances of several metals is shown in Fig. 3.11.

Figure 3.11 Spectral reflectance of films of silver, gold, aluminum, copper, rhodium, and
12
titanium.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Radiometric Properties of Materials 79

3.9 Optical Materials Checklist

Finally, Table 3.4 provides an “optical materials checklist” that includes several
properties useful to the designer selecting materials for an optical design effort.

Table 3.4 Optical materials checklist.

Optical properties
Transmission (function of wavelength, temperature,
direction)
Index of refraction (function of wavelength,
temperature, direction)
Dispersion, partial dispersion
Surface reflectance
Scatter (surface & bulk)
Absorption (bulk)
Homogeneity
Birefringence, stress coefficient
Fluorescence
Anisotropy
Electro-optic and/or acousto-optic coefficients
Mechanical properties
Young’s modulus
Yield point
Hardness
Optical workability
Coating compatibility
Density, specific gravity
Thermal properties
Thermal conductivity
Specific heat, heat capacity
Coefficient of linear thermal expansion
Softening point, melting point
Environmental properties
Solubility in H2O, other solvents
Surface deterioration, devitrification
Radiation susceptibility (UV, hard particle)
Other factors
Availability
Safety factors, toxicity
Cost
Compiled by James M. Palmer 02/21/89

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 80 Chapter 3

For Further Reading

“Standard practice for angle resolved optical scatter measurements on specular
and diffuse surfaces,” ASTM Standard E1392-90, ASTM International,
Philadelphia (1990).
“Standard practice for goniometric optical scatter measurements,” ASTM
Standard E2387-05, ASTM International, Philadelphia (2005).
J. C. Stover, Optical Scattering: Measurement and Analysis, SPIE Press,
Bellingham, Washington (1995).

References

1. A. G. Worthing, “Temperature, radiation, emissivities and emittances,” in

Temperature: Its Measurement and Control in Science and Industry,
Reinhold, New York (1941).
2. J. C. Richmond, “Rationale for emittance and reflectivity,” Applied Optics
21(1), pp. 1–2 (1982).
3. J. C. Richmond, W. N. Harrison, and F. J. Shorten, “An approach to thermal
emittance standards,” in Measurement of Thermal Radiation Properties of
Solids, NASA SP-31, J. C. Richmond, Ed., NASA, Washington, D.C. (1963).
4. W. L. Wolfe, “Proclivity for emissivity,” Applied Optics 21(1), p. 1 (1982).
5. J. C. Richmond, J. J. Hsia, V. R. Weidner, and D. B. Wilmering, Second
Surface Mirror Standards of Spectral Specular Reflectance, NBS Special
Publication SP 260-279, U.S. National Bureau of Standards, Washington,
D.C. (1982).
6. Radiometric and Photometric Characteristics of Materials and Their
Measurement, CIE Publication 38 (1977).
7. F. E. Nicodemus, J. C. Richmond, J. J. Hsia, I. W. Ginsberg, and T.
Limperis, Geometrical Considerations and Nomenclature for Reflectance,
NBS Monograph 160, U.S. National Bureau of Standards, Washington, D.C.
(1977).
8. D. B. Judd, “Terms, definitions, and symbols in reflectometry,” J. Opt. Soc.
Am. 57(4), pp. 445–450 (1967).
9. F. E. Nicodemus, “Directional reflectance and emissivity of an opaque
surface,” Applied Optics 4(7), cover picture (1965.)
10. W. L. Wolfe, “Radiation theory,” Chapter 1 in The Infrared Handbook, W.
L. Wolfe and G. J. Zissis, Eds., pp. 1–28, U.S. Government, Washington,
D.C. (1978).
11. J. C. Richmond, “Coatings for space vehicles,” in Surface Effects on
Spacecraft Materials, F. J. Clauss, Ed., Wiley & Sons, New York (1960).

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Radiometric Properties of Materials 81

12. W. L. Wolfe, “Optical materials,” Chapter 7 in The Infrared Handbook, W.

L. Wolfe and G. J. Zissis, Eds., pp. 7–81, U.S. Government, Washington,
D.C. (1978).

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Chapter 4
Generation of Optical
Radiation

4.1 Introduction
From a discussion in the previous chapter on the interaction of radiation with
materials, we now turn to the subject of how radiation is generated, and the roles
that emission, reflection, and other processes play.
Sources of optical radiation can be classified in a variety of ways. Active
sources emit optical radiation due to their temperature (thermal sources) or as a
result of atomic transitions (luminescent sources). Passive sources reflect optical
radiation from active sources or from other passive sources. Passive sources can
also be classified as thermal or luminescent, depending upon the process that
generated the radiation initially. Examples of thermal sources include blackbody
radiation simulators, tungsten-filament lamps, gases, the sun, the moon, and you
and I. Examples of luminescent sources include lasers, fluorescent lamps,
mercury arcs, sodium lamps, electroluminescent panels, LEDs, and gases. Some
sources combine both thermal and luminescent mechanisms, and some may be
both active and passive, reflecting in one spectral region and emitting in another.
Other means of classification have also been used. Some authors distinguish
between artificial (man-made) and natural sources. Lamps are artificial sources,
whereas the earth, the sun, and stars are natural sources. Still another practice is
to divide sources according to their output spectral characteristics. Continuous
sources have a spectral radiance that is slowly varying with wavelength, typical
of thermal radiation, while line sources emit in narrow, well-defined spectral
regions. Yet another attempt to distinguish sources is by their degree of spatial
and/or temporal coherence.
Thermal radiation has been extensively studied since the late nineteenth
century. Stefan was the first to experimentally examine the relationship between
radiation and temperature in 1879. He analyzed data from Tyndall and found that
the total radiation is proportional to the fourth power of temperature T4.
Boltzmann derived this T 4 relationship from the Carnot cycle in 1884. In 1891
Wien derived his displacement law, which relates the peak radiation to
83

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 84 Chapter 4

temperature, and the T 5 relationship for the magnitude of the peak. In 1896, Wien
derived an equation for the spectral distribution of thermal radiation based on
thermodynamic arguments. In 1900 Rayleigh derived another equation for
spectral distribution based on equipartition, and Jeans in 1905 independently
repeated this derivation. Planck in 1901 published an empirical equation
involving the notion that energy exists as discreet “packets” that fit experimental
data better than either the Wien or the Rayleigh-Jeans equation. This proved
monumental, as it was later accepted as the birth of quantum mechanics.
Verifications of the Planck equation and the physical constants continued up to
1982 with precise measurements of the spectral distribution of blackbody
radiation, determination of the Stefan-Boltzmann constant, and confirmation of
the thermodynamic temperature scale.

4.2 Radiation Laws

4.2.1 Planck’s law

The Planck expression for blackbody radiation is at the heart of all thermal
radiation equations. Any object at a temperature above absolute zero (0 K) will
radiate as a modified Planckian radiator. A full derivation of the Planck equation
will not be presented here, as it can be found in many physics texts. A brief
outline of the derivation will show its important features. It requires the
determination of the number of discrete frequency modes in a cavity. The total
mode density in a cavity (in vacuo) as a function of frequency ν is:

8πν 2
N ν dν = dν . (4.1)
c3

The average energy per mode looks like

hν
q= hν / kT
. (4.2)
e −1

The energy density (energy per unit volume per unit frequency interval) is
the product of the number of available modes and the average energy of each
mode. It is expressed as:

8πhν3 1
uν = . (4.3)
c 3 (e hν / kT − 1)

The energy exits a cavity at velocity c into 4π sr. The behavior is described by
Planck’s law, expressing radiance per unit frequency interval (spectral radiance):

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Generation of Optical Radiation 85

2hν 3 1
Lν = 2 hv / kT
. (4.4)
c e −1

Frequency is fundamental (independent of the medium), but wavelength is

preferred to frequency in radiometric applications due to ease of measurement.
Making use of the identity

Lλ d λ = Lν d ν

and the relationship between wavelength and frequency λ = c/ν, the spectral
radiance of blackbody radiation can be expressed in wavelength terms as

2hc 2 1
Lλ = 5 hc / λkT
. (4.5)
λ e −1

This is one form of Planck’s equation, whose units are W/m2sr·m.

The above equation implies that the process is taking place in vacuum where
the refractive index is unity. A more general expression utilizes n. The most
frequently encountered and useful form of the Planck equation is

c1 1
Lλ = 2 5 c 2 / nλT
. (4.6a)
πn λ e −1

Because blackbody radiation is Lambertian, the spectral radiant exitance may be

expressed as

c1 1
Mλ = 2 5 c 2 / nλT
. (4.6b)
nλ e −1

In this pair of important equations,

2
c1 = 2πhc = 3.74177107(29) × 10–16 Wm2 (7.8 × 10–8) (first radiation
constant)
c2 = hc/k = 1.4387752(25) 10–2 m·K (1.7 × 10–6) (second radiation constant)
h = 6.62606876(52) ×10–34 J·s (7.8 × 10–8) (Planck constant)
c = 299792458 m/s (exact) (velocity of light)
k = 1.3806503(24) × 10–23 J/K (1.7 × 10–6) (Boltzmann constant)
n = index of refraction (1 for vacuum, ≈1.00028 for air)

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 86 Chapter 4

The above notation is a conventional way to present uncertainties in the

physical constants. The number in parenthesis that is attached to the value is the
absolute uncertainty (one sigma, or 1σ) in the last digit(s) of the constant. For
example, c2 = 0.014387752 ± 0.000000025 m·K. The last number in parenthesis
is the relative uncertainty in the value of the constant. It is the absolute
uncertainty divided by the value of the constant.
The complete Planck equation as shown above is valid in any media having
index of refraction n. In a vacuum (n = 1), the n can be dropped, λ = λo, and the
Planck equation is most often seen this way. If in air (nair ≈ 1.00028), this
correction is usually ignored except for extreme low-uncertainty applications.
It is more convenient for optical measurement work to use micrometers (μm)
or nanometers (nm) for wavelength to get W/m2sr·μm or W/m2sr·nm. The
radiation constants then become

c1 = 3.74177107 × 108 W·μm4/m2 (wavelength in μm),

c1 = 3.741771 07 × 1020 W·nm4/m2 (wavelength in nm),
c2 = 14387.752 μm·K = 1.438 7752 × 107 nm·K.

The two curves in Fig. 4.1 show the form of Planck’s equation as a function
of wavelength, with temperature as a parameter. They are strongly peaked, with
the form governed by the λ–5 term for wavelengths longer than the peak and by
the exponential term for shorter wavelengths. Note that only a limited range of
temperatures can be shown on a single linear plot such as these, as the ordinates
are highly nonlinear, varying over many orders of magnitude. The plots also
show a dashed line, the locus of the wavelength of peak spectral radiance, having
a characteristic hyperbolic shape. As will be seen, these dashed lines represent
the Wien displacement law.
If the spectral radiance curves are plotted on log-log axes as in Fig. 4.2,
several interesting things are seen. First, this form allows for a wide range of
temperatures and wavelengths on a single plot. Second, the locus of the
wavelength of peak radiance is a straight line on a logarithmic plot, indicating
hyperbolic behavior. Finally, note that all of the curves have an identical shape
when the logs are plotted. Since the shape of the curve is independent of
temperature, one could construct a nomogram by tracing a single curve and the
straight line locus of maxima onto a transparent sheet and use it as an overlay.
Slide the overlay along the straight line to display the blackbody radiation curve
for any temperature.

4.2.2 Wien displacement law

The Wien displacement law describes this line, the locus of the peak wavelength
of the blackbody curve as a function of temperature. It is determined by taking
the derivative of Lλ with respect to wavelength (in the medium) and setting it to
zero to find the maximum. The result is

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Generation of Optical Radiation 87

nλmax T = 2.897 7686(51) × 10–3 m-K (1.7 × 10–6). (4.7)

A rough but useful approximation for the peak wavelength (in μm, setting n = 1)
is

λ maxT ≈ 3000 μm-K. (4.8)

Table 4.1 lists the peak wavelengths for several common sources.

(a)

(b)
Figure 4.1 Spectral radiance of blackbody radiation for (a) high temperatures and (b)
lower temperatures.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 88 Chapter 4

Figure 4.2 Log-log plot of blackbody spectral radiance as a function of wavelength and
temperature.

Table 4.1 Peak wavelengths for several common sources.

Temperature Peak wavelength

Source
(K) (μm)
Sun ~6000 ~0.5
Tungsten lamp ~3000 ~1
Typical hot IR source ~1000 ~3
Typical IR test source ~500 ~6
Room temperature ~300 ~10
Liquid nitrogen 77 ~40

To find the value of Lλ at the peak wavelength, solve the blackbody equation
using the peak wavelength (in the medium):

n 2 σ′ 5
Lλ max = T , (4.9)
π

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Generation of Optical Radiation 89

where σ' = 1.286 × 10–11 W/m2K5μm. The radiance at the peak wavelength varies
as the fifth power of the temperature.

4.2.3 Stefan-Boltzmann law

To determine the total radiance of blackbody radiation, integrate the Planck
equation over all wavelengths. The result is the Stefan-Boltzmann law,

∞
n2σ 4
L =  Lλ d λ = T , (4.10)
0
π

where σ is the Stefan-Boltzmann constant, 2π5 k 4 15c 2 h3 = 5.670 × 10 –8 W/m 2 K 4 .

The Stefan-Boltzmann law is most often seen in terms of radiant exitance with
n = 1:

M = σT 4 . (4.11)

Insert room temperature (300 K) into the Stefan-Boltzmann equation to find

that one square meter of a black surface emits ≈460 W. Therefore the walls,
ceiling, and floor of a typical 6 × 6 × 2.5-m classroom have a total surface area of
132 m2 and emit over 60 kW into the room. For the moment, we ponder why we
don’t bake under this onslaught.

4.2.4 Laws in photons

It is frequently advantageous to work with photons rather than watts, particularly
in cases where light levels are low. Thus, a photon-based analysis may be helpful
for applications in which photomultiplier tubes provide photon counting or CCDs
are employed to image a faint target. The energy associated with a photon, in
joules, is given by:

c
Q = hν = h , (4.12)
λo

where h is again Planck’s constant. For energy in units of electron-volts (eV),

divide by the electronic charge q (1.602 × 10–19 C).* At a (vacuum) wavelength
λo of 1 μm, it takes 5.034 × 1018 photons per second to equal one watt, and more
at longer wavelengths. Conversely, a single photon at a (vacuum) wavelength λo
of 1 μm has an energy of 1.986 × 10–19 joules (W·s), or 1.239 eV, and has
proportionally more energy at shorter wavelengths.

*
More accurate values for the constants that appear in this section are given in Appendix B; higher
accuracy is sometimes required.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 90 Chapter 4

The Planck equation rewritten for photons (n = 1) is

c1q 1
Lqλ = 4 c2 / λT
, (4.13)
πλ e −1

where c1q = 2πc = 1.883 × 109 m/s. For wavelength in μm, c1q becomes 1.883 ×
1027 m2/μm·s. Thus, Lqλ is expressed in photons per second per area per unit
wavelength.
Plots of spectral photon radiance are shown in Figs. 4.3 and 4.4. The curves
appear similar to the previous radiance curves, but the range of ordinate values is
not so extreme. The equation is subjectively the same; the exponential term is
identical but the wavelength in the denominator is only raised to the fourth
power.

(a)

(b)

Figure 4.3 Photon spectral radiance versus wavelength for (a) high temperatures and (b)
lower temperatures.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Generation of Optical Radiation 91

Figure 4.4 Log-log plot of spectral photon radiance as a function of wavelength and
temperature.

The Wien displacement law for photons is derived as described earlier; the
result is

λ q ,maxT = 3.6696986(62) × 10−3 m ⋅ K . (4.14)

If the wavelength is expressed in micrometers, a useful approximation is

λ q ,maxT ≈ 3700 μm ⋅ K . (4.15)

The photon spectral radiance at the peak wavelength is

σ′q
Lqλ max = T4 , (4.16)
π

where σ' q = 2.101 × 1011 s–1m–2K–4μm–1.

The total photon radiance is the integral of the Planck equation for photons:

∞
σq
Lq =  Lqλ d λ = T3, (4.17)
0
π

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 92 Chapter 4

where σq = 1.520 × 1015 s–1m–2K–3. At a temperature of 300 K, one square meter

of a black surface emits about 4 × 1022 photons per second. Therefore there are
more than 5 × 1024 photons per second being emitted by the walls, ceiling, and
floor in our typical classroom.

4.2.5 Rayleigh-Jeans law

There are two common approximations to the Planck equation. The Rayleigh-
Jeans equation for blackbody radiation was derived independently in 1900 by
Rayleigh and in 1905 by Jeans from the thermodynamic principle of
equipartition. It successfully predicted the shape of the spectral curve at long
wavelengths, but was clearly inappropriate at short wavelengths. It predicted
infinite energy at λ = 0, and the integral did not converge, giving rise to the
labeling of this equation as the “ultraviolet catastrophe.” Assume that n = 1 in Eq.
(4.6) and begin the derivation using the Planck equation for spectral radiance:

c1 1
Lλ = 5 c 2 / λT
.
πλ e −1

The exponential term can be expanded:

c 2 / λT c2 (c2 / λT ) 2 (c2 / λT )3
e =1+ + + + ... .
λT 2! 3!

If c2/λT << 1 (corresponding to a large value of λT), drop all of the higher-order
terms. Then,

c 2 / λT c2
e −1 ≈ .
λT

Rearranging terms and substituting into the Planck equation, the result becomes:

2ckT
Lλ = . (4.18)
λ4

This is the Rayleigh-Jeans law.

This expression is valid with less than 1% error if λT > 0.778 m·K (i.e., long
wavelengths and/or high temperatures). This approximation is not particularly
useful, even for far-infrared work, as less than 0.1% of the output of a blackbody
is located at λT values larger than 0.8 m·K, and the wavelength where it becomes
valid is some 250× the peak wavelength. This means for sunlight (6000 K) and
room temperature (300 K), the minimum wavelengths for which the
approximation is valid are 130 and 2600 μm, respectively.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Generation of Optical Radiation 93

4.2.6 Wien approximation

To derive the Wien approximation, begin with the Planck equation (n = 1).
Assume that e( c2 / λT ) >> 1 . Then, the 1 can be dropped and the equation
becomes:

−c
c1 λT2
Lλ = e (4.19)
πλ 5

The Wien approximation is valid with less than 1% error if λT < 3000 μm·K
(short wavelengths and/or low temperatures). It is quite useful for a great deal of
radiometric work as it is valid for blackbody radiation at all wavelengths shorter
than the peak. Figures 4.5 and 4.6 depict the curves and the ranges of validity for
the two approximations.

4.2.7 More on the Planck equation

A career can be spent fiddling with the Planck equation, presenting it for various
frequencies, wavenumbers, etc. We will keep it simple and just do some
normalization. Divide both sides of the Planck equation by T5:

Lλ c1 1
= . (4.20)
T 5 π(λT )5 ec2 / λT − 1

(μm·K)

Figure 4.5 Wien and Rayleigh-Jeans blackbody approximation curves.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 94 Chapter 4

(μm-K)
Figure 4.6 Errors associated with Wien and Rayleigh-Jeans approximations.

The result is an expression which is now a function of a new variable,

(λΤ ). The result of this normalization is a single curve as shown in Fig. 4.7 for
Planckian radiation for any wavelength and temperature.
The Planck function can also be normalized by dividing by Lλmax such that
the resulting curve peaks at unity. This curve is denoted as f(λT). An additional
very useful curve shows the cumulative (integrated) radiance from 0 up to λT
divided by the total radiance at temperature T. This cumulative curve is labeled
F(λT). The defining equations for these two functions are:

Lλ (λT ) L (λT )
f (λ T ) = = λ (4.21)
Lλ (λ maxT ) σ' 5
T
π
and
λ λ

 Lλ (λT )d λ  L (λT ) d λ
λ

F (λT ) = 0
= 0
. (4.22)
∞
σ 4
 L (λT ) d λ
0
λ
π
T

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Generation of Optical Radiation 95

5E-12

4E-12
RADIANCE/T^5

3E-12

2E-12

1E-12

0
0 5000 10000 15000
WAVELENGTH x TEMPERATURE (um-K)

Figure 4.7 Normalization of the Planck curve.

Since these definitions are ratios, exitance M can readily be substituted for
radiance L. These two functions are graphed in Fig. 4.8.
To use these curves to determine radiance in a narrow wavelength interval
(Δλ < 0.05λc), first select T for the blackbody radiation and the desired center
wavelength λc. Determine the radiance at the peak using the equation for Lλ(max).
Finally, use the function f(λT) from the graph and the wavelength interval Δλ to
arrive at the result

λ c +Δλ / 2


λ c −Δλ / 2
Lλ (λT )d λ = Lλ (λ maxT ) f (λ cT )Δλ . (4.23)

If the wavelength interval is large, typically greater than 0.05× the center
wavelength, use the other function F(λT) to determine the radiance in a finite
wavelength interval. Again select T and the two desired wavelengths, λ1 and
λ2. From the graph, read F(λ1T) and F(λ2T) and compute the total radiance using
the Stefan-Boltzmann law. The result is

λ2
σ
 L (λT ) d λ = π T [ F (λ T ) − F (λ T ) ]
4
λ 2 1 λ2 > λ1. (4.24)
λ1

Figure 4.8 also shows the corresponding curves for photons, fq(λT) and
Fq(λT). The defining equations are

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 96 Chapter 4

Figure 4.8 Curves of f(λ) and F(λ) for watts and photons.

Lqλ ( λT )
f q ( λT ) = (4.25)
Lqλ ( λ maxT )
and
λ λ

 Lqλ (λT )d λ L qλ (λ T ) d λ
Fq (λT ) = 0
= 0
. (4.26)
∞
σq
L
3
qλ (λT ) d λ T
0
π

The application of the fq and Fq photon curves is identical to the curves for
energy.
In some applications, it is desirable to maximize the radiation contrast
between a target and background of similar temperature. What wavelength might
one choose for this task? The problem occurs regularly in the infrared where both
the target and the background radiate near 300 K. Take the second derivative
d2Lλ/(dTdλ), and set it to zero. The result is:

λ contrast T = 2411 μm ⋅ K . (4.27)

This equation implies that the best “visibility,” or contrast with the
background, occurs at a wavelength somewhat shorter than the peak wavelength,
at λcontrast = 0.832λmax. Since the result is on the short wavelength side of the
peak, the Wien approximation is valid and it simplifies the calculus considerably.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Generation of Optical Radiation 97

Thus, if your target and background temperature were 305 K, the peak
wavelength is 9.50 μm and the wavelength of maximum radiation contrast λcontrast
is 7.9 μm. The wavelength for maximum photon contrast is 2898 μm·K, the same
as the peak wavelength for energy.
We often need to know how the spectral radiance Lλ changes with
temperature; this can be determined by differentiating the Planck function with
respect to temperature. The result, shown in differential form, is

ΔLλ xe x ΔT
= x , (4.28)
Lλ e −1 T

hc c
where x = = 2 .
λkT λT
xe x
We define Z ≡ x . For a small change in temperature ΔT such that the
e −1
change in x is also small, the change in Lλ with temperature at any wavelength is

ΔLλ ΔT
=Z . (4.29)
Lλ T

If ex is significantly greater than unity (λT <3000 μm·K, the Wien

approximation), then Z = x and ΔLλ/L = xΔT/T. Inspection of the blackbody
curves and this equation shows that Z approaches infinity as the wavelength
approaches zero, and Z approaches unity at longer wavelengths. Figure 4.9 shows
Z plotted against λT; the curve is identical for watts and photons. The region
where the Wien approximation is valid is λT < 3000 K (Z = 5).

4.2.8 Kirchhoff’s law

Blackbody radiation exists in any closed cavity at thermal equilibrium. As such,
it is idealized, because whenever a hole is made into the cavity to allow radiation
to exit, the conditions for blackbody radiation are altered. As noted in Chapter 3,
emissivity is defined as the ratio of the radiance of an object to the radiance of a
blackbody at the same temperature. The directional spectral emittance of a
specific sample, ε(λ;θ,φ), is

Lλ (θ, φ)
ε(λ; θ, φ) = . (4.30)
LλBB

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 98 Chapter 4

Figure 4.9 Parameter Z as a function of λT.

This equation differs slightly from Eq. (3.23), as it includes the directional
component.
Total emissivity is the integral of spectral directional emissivity over all
angles and wavelengths, also weighted by the Planck function, which introduces
a temperature dependence if the spectral emissivity is not uniform (that is, gray):

ε=
 (θ, φ; λ) L λBB sin θ cos θd θd φd λ
. (4.31)
(σ / π)T 4

Since ε(λ;θ,φ) is not a derivative (per unit wavelength interval) quantity, it

makes no sense to integrate it alone:

ε ≠  (θ, φ; λ )d λ . (4.32)
λ

Kirchhoff’s law was stated in Eq. (3.24):

α = ε. (4.33)

To illustrate, Fig. 4.10 shows a small body within a large isothermal

enclosure. At equilibrium, in the absence of other sources or sinks, the source and
enclosure must be the same temperature. The power absorbed by the small body
(by the definition of absorptance) is Φα = αΦi. The power emitted by the body
(by the definition of emittance) is Φe = εΦBB. At equilibrium, the power emitted

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Generation of Optical Radiation 99

by the body to the enclosure must equal the power absorbed by the body from the
enclosure:

αΦ i = εΦ BB .

As the incident and emitted power are the same (no other sources or sinks), the
conclusion is α = ε.
This seems too simple; is it really true? No! Look at the following equation
for the equilibrium temperature of a flat plate in space, facing the sun and
insulated on the back side. It is a simple statement of conservation of power,
equating radiant incidence (absorbed) to radiant exitance (emitted):

αEo = εσT 4 . (4.34)

Here Eo is the solar constant (1368 W/m2). The units on both sides of the
equation are W/m2. Solving for T, we find that it is a function of the ratio (α/ε).
But Kirchhoff’s law says that α = ε, and these terms therefore cancel. The
equilibrium temperature is therefore a function only of Eo. The implication is that
a white (allegedly reflective) car and a black (supposedly absorptive) car have the
same equilibrium temperature after sitting out in the sun all day. People from
Arizona and Florida know better than that, and often buy white cars to keep
cooler in the summer! So what’s wrong?
The answer is that Kirchhoff's law does not apply in all situations. Table 4.2
indicates the applicability in terms of spectral and directional conditions.
Specifically, α(λ;θ,φ) = ε(λ;θ,φ); the absorptance equals the emittance at a single
wavelength in a single direction. After integrating over wavelength and
geometry, test for the stated restrictions before applying Kirchhoff’s law. The

1
Figure 4.10 Illustration of Kirchhoff’s law. [Reprinted with permission of author from
Optical Radiation Measurement series, Vol. 1, F. Grum and R. J. Becherer, Radiometry, p.
98 (1979).]

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 100 Chapter 4

simple equation above attempts to relate the absorbed energy from the sun, which
is concentrated in the spectral range between 0.3 and 3 μm, to the thermal
radiation located in the wavelength range between 3 and 30 μm. White and black
paints have completely different reflectances in the solar region, but their
emittances at 300 K are very similar.
The thermal radiative properties of any material can be given in terms of its
temperature T and its spectral directional emittance ε(λ;θ,φ). The spectral
dependency is of primary importance, and the directional properties and
temperature coefficients are usually of lesser concern.
The blackbody equation is highly nonlinear with both wavelength and
temperature, and is thus not particularly tractable. Fortunately, there are many
computational and visual aids to help. The Infrared Handbook gives several
calculator programs for now-obsolete HP and TI calculators, and some early
calculators were available with plug-in cards to do blackbody calculations. The
latest incarnation, The Infrared and Electro-Optical Systems Handbook,
substitutes an extensive set of programs in BASIC to do the computations. There
have been several slide rules (remember them?) that do calculations on
blackbody radiation. The GE Radiation Calculator is a plastic rule that
occasionally surfaces at a reasonable price. Cardboard knock-offs are often given
away by vendors. For a little more money, Electro-Optical Industries at one time
sold a high-quality metal rule. One side of my venerable GE calculator is shown
in Fig. 4.11.
Back when computers were scarce, tables generated by mainframes were
commonly used to do precise blackbody calculations. Several tables may still be
found in musty libraries.2,3,4
Today, simple computer programs can easily be written. Spreadsheets are
sufficiently powerful to do these calculations with relative ease and provide
superior graphics as well. Tools like Mathcad, MATLAB, and Mathematica also
work well.

Figure 4.11 The venerable GE radiation calculator—J. M. Palmer’s own!

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Generation of Optical Radiation 101

1
Table 4.2 Summary of absorptance-emittance relations. [Adapted with permission of
author from Optical Radiation Measurement series, Vol. 1, F. Grum and R. J. Becherer,
Radiometry, p. 115 (1979).]

Directional spectral α(λ;θ,φ,Τ ) = ε(λ;θ,φ,Τ )

Required conditions:

None, other than thermal equilibrium

Directional total α(θ,φ,Τ ) = ε(θ,φ,Τ )

Required conditions:

(1) Spectral distribution of incident energy proportional to blackbody at T, or

(2) α(λ;θ,φ,Τ ) = ε(λ;θ,φ,Τ ) independent of wavelength

Hemispherical spectral α(λ,Τ ) = ε(λ,Τ )

Required conditions:

(1) Incident radiation independent of angle, or

(2) α(λ;θ,φ,Τ ) = ε(λ;θ,φ,Τ ) independent of angle

Hemispherical total α(Τ ) = ε(Τ )

Required conditions:
(1) Incident energy independent of angle and spectral distribution proportional to
blackbody at T, or
(2) Incident energy independent of angle and α(λ;θ,φ,Τ )
= ε(λ;θ,φ,Τ ) independent of wavelength, or
(3) Incident energy at each angle has spectral distribution proportional to
blackbody at T and α(λ;θ,φ,Τ )
= ε(λ;θ,φ,Τ ) independent of angle, or
(4) α(λ;θ,φ,Τ ) = ε(λ;θ,φ,T ) independent of angle and wavelength

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 102 Chapter 4

Spectral emittance was defined as the ratio between the spectral radiance of
an object and the spectral radiance of a blackbody at the same wavelength and
temperature. In general, it is also a function of angles θ and φ. Using this
definition, any isothermal object has a spectral radiance given by

 ε(λ; θ, φ)c1   1 
Lλ (θ, φ) =  5   c2 / λT , (4.35)
 πλ  e −1 

and if ε(λ;θ,φ) and T are known, so is the spectral radiance. As ε(λ) increases
towards unity, blackbody radiation is approached. Blackbody radiation
simulators attempt to do just this, and many are commercially available. Since
the spectral radiant exitance can be described by this simple equation, blackbody
radiation simulators are most often used as standard sources of optical radiation
in the calibration laboratory.

4.3 Emitter Types and Properties

4.3.1 Metals
An important factor in the behavior of a material is its electrical conductivity.
Metals have a high conductivity, meaning large quantities of free electrons are
available to interact directly with the radiation field.
Since metals are generally good reflectors (a direct consequence of their high
conductivity), they are poor emitters (conservation of energy). The emittance is a
slowly varying function of wavelength, as shown by a simple empirical equation
given by Hagen and Rubens:5

resistivity
ε ≈ constant (λ > 2 μm) (4.36)
λ

This equation, which is generally valid at wavelengths longer than 2 μm, shows
that increasing resistivity (decreasing conductivity) increases emittance, and that
emittance decreases at longer wavelengths. Figure 4.12 gives examples.
The directional properties can be derived from Maxwell’s equations. The
radiation is highly polarized at angles off normal, as shown in Fig. 4.13. Note,
however, that the total radiation, the sum of both polarizations, is quite constant
with angle (i.e., Lambertian) to within a few percent out to nearly 60 deg from
specular.

4.3.2 Dielectrics
Dielectrics and gases have much lower conductivity than metals. Their electrons
are more tightly bound to their parent nuclei and require specific atomic
interactions with the radiation field. This implies that dielectrics tend to radiate in
specific, fairly well-defined spectral regions, and not elsewhere.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Generation of Optical Radiation 103

Figure 4.12 Spectral emissivity for some metals.6

The emitting properties of dielectrics are closely associated with the complex
index of refraction n + iκ where κ, the extinction coefficient, is in turn related to
the absorption coefficient α′, as discussed in Chapter 3. As noted in that chapter,
the product of the absorption coefficient and the thickness of a material is the
optical thickness τo. A material is considered optically thin when τo < 0.1
(transmission high) and optically thick when τo > 2 (transmission low). Optically
thin materials approach transparency and have low emittance; for optically thick
materials, the normal emittance is (1 – reflectance) and depends on the index of
refraction as determined by the Fresnel equation at normal incidence as
illustrated in Fig. 4.14. Emittance at other angles also comes from the Fresnel
equations and is polarized.

4.3.3 Gases
Gases are optically thin over wide wavelength ranges and may be transparent
over long paths. Therefore, their emittance is essentially zero at these
wavelengths. However, there are specific spectral regions where absorption, and
therefore emission, occur. Each species has its own absorption characteristics,
correlated with its atomic and molecular structure and energy levels. These
characteristics take the form of a series of spectral lines at regular locations in the
spectrum. They are occasionally seen as discrete lines, but more often as a series
of overlapping lines called bands.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 104 Chapter 4

Figure 4.13 Directional emittance as a function of angle from specular.

The important species identified with our own atmosphere include “fixed”
gases O2, N2, and CO2. Those species classified as variable include water, ozone,
and methane (H2O, O3, and CH4.) It must be stressed that the energy absorbed by
a gas is dependent upon the concentration and absorbing characteristics of the
gas, and the energy emitted by the gas depends upon the temperature and spectral
emittance of the gas. The processes of absorption and emission are not
independent, but occur simultaneously.

4.4 Practical Sources of Radiant Energy

4.4.1 Two major categories

Table 4.3 presents a division of sources into the two categories of thermal and
luminescent. Several of these sources will be discussed in this section. Natural
sources of radiation, which may be thermal or luminescent, will also be
discussed.
1
NORMAL EMITTANCE

0.9

0.8

0.7

0.6
1 2 3 4
INDEX of REFRACTION

Figure 4.14 Normal emittance versus index of refraction.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Generation of Optical Radiation 105

Table 4.3 Practical radiation sources.

Thermal sources Luminescent sources

Natural radiators Natural sources
Direct radiators Glow worms, fireflies
Earth (longwave) Aurora, airglow
Sun
Stars Artificial sources
Atmospheric gases Semiconductor devices
Reflective natural objects LEDs, IREDs
Earth (shortwave) Diode lasers
Moon, planets Phosphors
Skylight, clouds zodiacal light CRT tube (TV, computer
Atmospheric scatter display, oscilloscope)
Electroluminescent panels
Artificial radiators Lasers
Incandescent substances Gas (He-Ne, CO2, Ion, N2)
Tungsten lamps Solid-state (Ruby, YAG)
Hot metals Metal vapor (He-Cd)
Silicon carbide rods Liquid (dye, cyanide, chelate)
Ceramic tubes Chemical (HF, DF)
Nernst glower Metal-vapor lamps
Electrical transmission Sodium vapor lamp
equipment Mercury vapor lamp
Machinery Fluorescent lamp
Personnel Gas discharge lamps
Carbon arc Plasma display
Optical elements (lenses, Neon lamp
mirrors, etc.) Glow discharge
Open coal, wood, oil fires Xenon arc and flashtube
Hot gases
Welsbach mantle
Bunsen burner
Exhaust gases

4.4.2 Thermal sources

4.4.2.1 Tungsten and tungsten-halogen lamps

The most common household source of optical radiation (other than fire) has
been the tungsten-filament lamp. Equation (4.35) may be used to describe the
spectral radiance of a tungsten lamp; it is the equation for blackbody radiation
with the emittance term included. Tungsten’s spectral emittance has been
extensively studied. The most frequently used data are from DeVos, shown in

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 106 Chapter 4

Fig. 4.15. He measured the emittance of a flat ribbon of tungsten, but the results
are also applicable to a round wire. Most lamps are made by drawing the
tungsten into a round wire and then tightly coiling it, enhancing the emittance by
making a partial cavity. If this tight coil is then further loosely coiled, emissivity
is enhanced further.
Tungsten lamps are designed to operate at a nominal voltage and current. The
design compromise is between light and lifetime. If more light is needed, voltage
(or current) may be increased, but not for very long. Figure 4.16 shows light
(lumens), efficiency (lumens/watts), and lifetime as a function of operating
voltage.
Tungsten-filament lamps decay due to evaporation of the filament, leaving
brown deposits on the inside of the lamp envelope. An uneven rate of
evaporation creates “hot spots” which cause the lamp’s overall rate of decay to
increase. Additionally, filaments may crystallize and become brittle particularly
when the lamp is operated on dc, mechanically weakening the filament and
making it susceptible to breakage from mechanical or thermal shock.
Tungsten-halogen lamps are better suited than tungsten for most applications,
as they either eliminate or delay the onset of both decay mechanisms. The
addition of a halogen such as bromine or iodine creates a regenerative cycle, in
which the evaporated tungsten combines with the halogen rather than plating on
the envelope. A hot (minimum 250° C) envelope, usually of fused silica, is
required. The resulting halide compound decomposes at a rate proportional to
temperature. This decomposition occurs preferentially at the “hot spot,” causing
tungsten to plate back onto the filament.

Figure 4.15 Spectral emittance of tungsten. (Reprinted from Ref. 7 with permission from
Elsevier.)

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Generation of Optical Radiation 107

Figure 4.16 Tungsten lamp characteristics as a function of percent normal operating

voltage.

The advantages of the tungsten-halogen lamp are many. They have a

significantly longer lifetime—at least a factor of ten—than the tungsten filament
lamp alone for a given filament temperature, and they can be operated at much
higher temperatures. They exhibit superior long-term stability because the
envelopes don’t darken. Because they require a high envelope temperature, they
are physically of much smaller size than the simple tungsten lamp for a given
wattage.
There are disadvantages to this approach as well. The hot quartz envelope
requires greater care, both in fixture design and in handling. Fingerprints are
verboten; in calibration applications, fingerprints and grease on the lamp can
severely impair measurement accuracy. If the temperature of the envelope is not
hot enough, the regenerative cycle fails, and the lamp will fail sooner. The lamps
have high internal pressure, and the increased ultraviolet content requires that
users take safety measures. Finally, their purchase price is considerably higher
than tungsten lamps alone.
It should be noted that at the time of publication (2009), incandescent
tungsten lamps have fallen out of favor, largely due to energy efficiency
considerations in comparison to alternatives such as compact fluorescent bulbs
and light-emitting diode (LED) lighting. In fact, sales of incandescent light bulbs

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 108 Chapter 4

are being phased out in the European Union, Australia, the United States, and
other nations.

4.4.2.2 Other metallic sources

Metallic sources have some general properties worth noting. Since they are fair-
to-good conductors of electricity, they can be heated to the operating temperature
by simply applying an electric current. The power dissipated is given by I2R. The
resistance of metals increases with increasing temperature, meaning that a current
surge can be expected when a device is turned on. This is particularly noticeable
with tungsten, where the starting current required is an order of magnitude
greater than the operating current. Other materials with lower operating
temperatures show less of this effect.
Refractory (high-temperature) metals that may find occasional application
include tantalum and molybdenum. They have lower melting points than
tungsten and higher vapor pressures. They are more reflective than tungsten,
hence at normal operating temperatures their emittance is lower. They also are
more susceptible to oxidation. Other than the above, they work well.
Noble metals, those that remain in pure form, include platinum, palladium,
and iridium. They also have lower melting points and lower emittances than
tungsten, but are resistant to oxidation and will therefore operate in air.
Certain alloys have been developed for use in heating applications. The most
often-used alloy is 80Ni-20Cr (one trademark is Nichrome), which is the element
in toasters, space heaters, and other devices. The resistance of Ni-Cr is essentially
constant with temperature, simplifying the power supply design. To season it, run
it at an elevated temperature in air to maximize the oxidation, which enhances the
emittance.

4.4.2.3 Dielectric thermal sources

Nominal characteristics of dielectric sources include a high emittance in those
spectral regions where they are optically thick, and a negative resistance versus
temperature characteristic. The resistance at room temperature is usually
sufficiently high that extreme voltages would be required to generate sufficient
heat to get them started. Large amounts of ballast are needed to ensure an overall
positive resistance versus temperature characteristic. Alternatively, they can be
indirectly heated by a conventional resistive heater.
The globar is a rod fabricated from silicon carbide (SiC). It is typically 5 mm
in diameter and 50 mm long and requires about 200 W to reach the operating
temperature of 1000 to 1500 K. Its emittance is approximately 0.75 out to a
wavelength of 15 μm. This source can be directly heated, but requires an
auxiliary ballast to overcome the slight negative resistance characteristics of the
SiC. Establishing electrical contact is difficult and is usually accomplished with
water-cooled silver electrodes.
The Nernst glower is a small (1-mm diameter by 10 mm long) ceramic rod
comprising mixed oxides of Zr, Y, Ce, Th, Be and the like. It dissipates about

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Generation of Optical Radiation 109

200 W and operates at 1500 to 2000 K. The emittance is about 0.75 out to a
useful wavelength of 30 μm. It is rather fragile and has a lifetime of 200 to 1000
hours. Its resistance at room temperature is extremely high, requiring an auxiliary
platinum heater coil to get it started. Nernst glowers are most often found in
infrared spectrometers where their shape is ideally suited to image onto a narrow
slit.
A popular low-cost IR source consists of an aluminum oxide (Al2O3) ceramic
tube heated with a coaxial nichrome wire. This source is often used in
inexpensive infrared spectrophotometers and operates at about 1200 K with an
emittance of about 0.8. The power source is a simple line-operated transformer.
The Welsbach mantle is a woven fabric mesh impregnated with refractory
oxides such as thorium oxide. It is heated with propane or white gas to a
temperature up to 2400 K. The emittance is high for λ > 10 μm. This source is
commonly known as the “Coleman” lantern.
The carbon arc is a valuable source for announcing grand openings, detecting
hostile aircraft, simulating solar radiation, etc. The radiation comes from the
plasma-heated carbon at its sublimation temperature of 3800 K. By placing
refractory oxides in the carbon rod, higher temperatures commensurate with the
evaporating points of these oxides can be achieved, in the range 5000 to 8000 K.
The high brightness comes from the small (about 10-mm diameter) size of the
carbons. The disadvantages are lack of stability, mechanical issues involving
continuous feeding of carbons, the power required (usually a noisy, smelly,
unregulated motor generator) and the need for ventilation (hydrogen cyanide is
generated).

4.4.2.4 Optical elements

The components in our optical systems also radiate in spectral regions where they
are not perfectly transparent or reflective. Mirrors are coated with thin metallic
films for high (ρ > 0.9) reflectance. Their emittance is consequently low in the
infrared (ε = 0.02 is a reasonable approximation). Even a surface with this low
emittance still generates copious quantities of photons per second unless cooled.
Refractive elements are transparent at wavelengths where they are normally
used; the emittance is low. At wavelengths where the material is absorptive,
however, the emittance is high. If our detectors are responsive to wavelengths
where the windows and lenses emit, they will respond to such radiation,
contributing nothing but noise.
Choppers are used to impart modulation to a beam. One must analyze what is
being seen when a chopper is used. If the chopper is reflective, you must
determine what the system sees in reflection. If the chopper is nonreflective
(black), then it has self emission governed by its own temperature and emittance.

4.4.2.5 Miscellaneous thermal sources

Flames emit large quantities of radiation, principally in the infrared. There are
two components to radiation from flames. The first is from particulate matter,

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 110 Chapter 4

like the hot soot from burning hydrocarbons with insufficient oxygen. Examples
include the oxyacetylene torch before the oxygen is turned on, and flames from
solid rocket motors. These particles are effectively “black,” emitting as small
blackbody radiators. The other component is radiation from the gases, including
impurities. For example, the yellow color from a Bunsen burner is due to atomic
emission from sodium in the supply gas.
Exhaust gases emit radiation in spectral regions where they are optically
thick. The important gases (from a radiometric standpoint) are the hot
combustion products CO, C2O, and H2O. Particulates, “black” if their size is
> 30 μm, are often present, particularly from diesel exhaust and muzzle flash.
If a gas both emits and absorbs at the same wavelength, how can we see it?
Our initial reaction is that the absorption lines from these gases block the
emission from being seen. In fact this is not the case. A phenomenon known as
“line reversal” takes place wherein hot gases have slightly different spectral
profiles than their unheated counterparts, broader because of increased
temperature and pressure. Absorption will occur at the center wavelength of the
profile, but the hot gas may be observed in two wavelength bands bordering the
center.
You and I make good infrared sources. Our temperature is about 300 K and
our emittance is nearly unity for all wavelengths longer than about 5 μm. We are
all black in the thermal infrared, regardless of skin pigmentation in the visible.
Appliances and conveyances also make interesting sources. Cars, trucks,
trains, tanks, and aircraft all have different temperatures than their surroundings
when at work. Even such little things as insulators on a power transmission line
get hot when leakage occurs, rendering them observable targets in the infrared.
Table 4.4 lists the total (integrated over wavelength) directional (normal
incidence) emittance for several materials at the indicated temperatures (in
degrees kelvin).

4.4.3 Luminescent sources

4.4.3.1 General principles

The term “luminescence” refers to emission of light from materials that receive
energy from various sources. Table 4.5 details some of the types of luminescence
and the sources of energy for each.
As we have seen, temperatures must be high to get significant radiation from
thermal sources in the visible and ultraviolet regions of the spectrum.
Luminescence can occur at much lower temperatures, with light generated as
atoms decay to a lower energy level from a higher level. In this process, photon
emission occurs at a specific wavelength, but emission lines may be spectrally
broadened by Doppler (Gaussian) and pressure (Lorentzian) effects. In practice,
this means that the emission occurs in a narrow band, rather than being confined
to a single wavelength.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Generation of Optical Radiation 111

Table 4.4 Emittances of some common materials.

Material Temperature (K) Emittance
Metals — —
Aluminum, polished 300–900 0.04–0.03
Aluminum, heavily anodized 373 0.84
Brass, polished 373 0.1
Chromium 310–1370 0.08–0.4
Copper, polished 300 0.02
Copper, black oxidized 300 0.8
Gold 300 0.02
Nickel, polished 300 0.05
Nickel, oxidized 500–1500 0.4–0.8
Silver 300–800 0.01–0.03
Stainless steel, polished 300 0.16
Stainless steel, oxidized 1000 0.85
Steel 300 0.08
Tungsten 300–3000 0.03–0.39
Dielectric Materials — —
Alumina 300–1000 0.96–0.6
Brick, red 300 0.93
Carbon, lampblack 300 0.95
Concrete, rough 300 0.93
Glass, plate 300 0.94
Ice 270 0.97
Magnesium oxide 400–750 0.69–0.55
Oil, 0.001 in. thick on nickel 300 0.27
Oil, thick on nickel 300 0.82
Paint, oil 350 0.94
Paint, laqcuer 350 0.97
Skin, human 300 0.98
Snow 270 0.82
Soil, dry 300 0.92
Soil, wet 300 0.95
Water 273–373 0.96

At low pressures (below 100 mbar), the Doppler effect predominates. The
shape of the curve is Gaussian, and a typical equation takes the form:

 ( v − vo )2 
0.47 S − ln 2  α2 
k (ν ) = e , (4.37)
α
where k(ν) is an extinction coefficient at frequency ν, S is a line strength, α is an
absorption coefficient related to the halfwidth of the line, and νo is the line

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 112 Chapter 4

Table 4.5 Luminescence processes and associated energy sources.

Process Energy source
Bioluminescence Organic chemical reactions
Cathodoluminescence Electron beam
Chemoluminescence Inorganic chemical reactions
Electroluminescence Electric field
Photoluminescence Optical radiation
Radioluminescence Charged particles
Sonoluminescence Acoustic energy
Thermoluminescence Mild heating below that required for
incandescence
Triboluminescence Energy from friction or pressure

center frequency. This shape is characteristic of low-pressure discharge sources

and atmospheric absorption and emission lines at elevations greater than 80 km
in the stratosphere. The key element in determining this profile is a long mean-
free path.
At higher pressures, where the mean-free path is short, the lines are further
broadened by collisions among the molecules. This pressure broadening gives a
Lorentzian shape of the form

Sα
k (ν ) = . (4.38)
π  (ν − ν o ) 2 + α 2 

The Lorentzian shape characterizes the emission of high-pressure discharge

sources and atmospheric absorption and emission lines at elevations less than
50 km in the troposphere.
At intermediate pressures, the line shape takes on a hybrid shape, a
convolution of the Gaussian and the Lorentzian curves. Figure 4.17 compares the
shapes of the two curves.
The comparatively narrow, Doppler-broadened linewidth of low-pressure
discharge is exploited in a number of sources. Because of the narrow width, there
is little power in each of the lines; sources are limited to low-power applications
involving sharp lines, such as wavelength calibration of spectrometers, excitation
of fluorescence spectra, and interferometry. The means of excitation include
alternating current and radio frequencies. In the latter case, unknown gases
placed in a sealed tube are excited by an external radio frequency (RF) field and
produce emission spectra which allow their identification. Common gases
identified in this manner include mercury, helium, neon, sodium, potassium, zinc,
cadmium, and cesium. These gases have further application in colorful signs and
illumination. An example of the emission lines of mercury and argon appears in
Fig. 4.18.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Generation of Optical Radiation 113

Doppler

Lorentz

Figure 4.17 Doppler and Lorentzian line shapes.

The high-pressure regime is entered when the pressure is increased to the

point where a low-resistance arc is observed. The line shape becomes Lorentzian,
with broad wings extending to either side of the line center. These wings
combine to provide a continuum. Arc lamps are efficient and extremely bright.
Common types include mercury, xenon, mercury-xenon, and sodium. They can
be modulated and flashed, as in the photographic strobe lamp. They are useful for
illumination, ultraviolet exposure for photoresist applications, projectors,
searchlights, and solar radiation simulators. Figure 4.19 depicts the spectral
irradiance of xenon arc sources.

8
Figure 4.18 Emission lines for Hg (< 600 nm) and Ar (> 600 nm). (Copyright held by
Ocean Optics, Inc. Reproduced with permission.)

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 114

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

Figure 4.19 Spectral irradiance of xenon arc lamps. (Permission to use granted by Newport Corporation. All rights reserved.)
Chapter 4
 Generation of Optical Radiation 115

Phosphors exhibit luminescent mechanisms and are highly temperature

dependent. They represent a variety of colors and persistences (decay time
constants greater than 20 ms). They are used in fluorescent lamps, cathode ray
tubes (CRTs), x-ray and gamma-ray screens, UV detectors, charged particle
detectors, and flat-panel displays, to name a few applications. Figure 4.20 depicts
the spectral characteristics of several phosphors.

4.4.3.2 Fluorescent lamps

The modern fluorescent lamp is a combination of a moderate-pressure (about 1
atm) mercury discharge lamp and a phosphor. The phosphor is excited by
radiation from the mercury discharge, predominantly from the 254-nm Hg line.
The phosphor, coated on the inside of the cylindrical glass envelope, reradiates at
longer wavelengths in the visible spectrum. Many phosphors are available with
varying color characteristics. Several types of daylight simulation are available,
from the rather harsh but inexpensive “daylight” to more subdued “warm white.”
Several pale colors are available for special effects, and a plant growth phosphor
is common for hothouse use. This latter phosphor emits in the red and blue
portions of the visible spectrum, and lacks green, since plants reject green light
by reflection. The mercury lines may be prominent, particularly when the more
transparent phosphors are used.
General characteristics of fluorescent lamps include a long lifetime and
relatively high illumination efficiency, as their output is concentrated in the
visible. They are a low-radiance extended source, principally suited for
illumination. They operate with dc, ac, and RF excitation. The discharge needs
help in getting started, as the mercury must be vaporized. Be careful in disposal
of spent lamps, as both the mercury and some of the phosphors are toxic.

Figure 4.20 Phosphor spectral characteristics.9 (Reprinted by permission of Pearson

Education, Inc., Upper Saddle River, NJ.)

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 116 Chapter 4

The compact fluorescent lamp (CFL) finds increasing use in today’s

environment, particularly as a replacement for tungsten incandescent bulbs
common in household use. The efficiency of these lamps is high, primarily due to
triphosphors and high-frequency solid-state electronic ballast. They are available
in several colors, and their liftetime extends to 12,000 hours. Compact
fluorescents are more expensive to buy than incandescent lightbulbs, but their
longer lifetime has rendered them a cheaper choice than the latter for many uses.
Figure 4.21 shows the components of a compact fluorescent lamp.

Figure 4.21 Components of the compact fluorescent lamp. (Reprinted with permission
th
from the IESNA Lighting Handbook, 9 Edition, by the Illuminating Engineering Society of
10
North America.)

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Generation of Optical Radiation 117

4.4.3.3 Electroluminescent sources

Electroluminescent sources are activated by an ac field, typically at 60 Hz. These
lamps are constructed by layering the phosphor on a metallic electrode, covered
by a semitransparent electrode and glass. They draw very little power and are
quite efficient. Their radiance is low, and their geometrical characteristics are
nearly Lambertian. The color selection is limited to available phosphors. Large
areas and complex shapes are possible. Their output depends upon voltage,
temperature, and excitation frequency. They are stable and exhibit long lifetimes.
They find uses for nightlights, emergency signs, backlighting for portable LCD
computer screens, and calibration sources for x-ray film. Other forms of
excitation are available, including biological (fireflies and glowworms), chemical
(plastic-encapsulated light sticks), and atomic (tritium-excited watch dials), for
example.

4.4.3.4 LED sources

A forward-biased p-n junction is an excellent emitter of optical radiation. The
emission comes from recombination radiation at the band edge excited by the
high current density. The wavelength region of emission is narrow,
approximately 5% of the center wavelength. These light-emitting diodes (LEDs)
are small, consume little power, and have lifetimes in excess of 10,000 hours.
The radiance from an LED is high due to the small emitting area, but the total
radiated power is low. They may be modulated to frequencies up to 1 GHz. They
are useful for displays and back-illuminating liquid crystal displays. They are
also used extensively in airline illumination for exit signs, floor and overhead
lighting, and cockpit lighting.
Traffic and roadway lighting applications are increasingly benefiting from
LED technology. Many older traffic lights are being replaced by LED sources in
a matrix configuration, which allows the light to continue to function although
several individual LEDs burn out. LEDs are also seen increasingly in
architectural displays. Table 4.6 lists several materials used for LED lighting
with their corresponding wavelengths or wavelength bands.
OLEDs (organic light-emitting diodes) are a special class of LED. They are
created from thin organic emitters sandwiched between a transparent anode and a
metallic cathode. Enhancement of efficiency and color control are provided by
doping emissive layers with highly fluorescent molecules. The structure of
organic layers and the choice of anode and cathode materials are designed to
maximize the recombination process in the emissive layer, thus maximizing light
output.
Due to the emitters’ thin width (a few micrometers), OLED display designs
can replace LCDs in many applications, including television monitors and
automotive displays. Figure 4.22 shows the OLED physical structure.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 118 Chapter 4

Table 4.6 LED materials and corresponding wavelengths.

Material Wavelength (nm)

GaN 350
SiC 465
GaP 565
ZnTe 620
CdTe 855
InP 1000
ZnCdTe 530 – 830
GaAsP 550 – 900
InGaAsP 550 – 3500
GaAlAsSb 580 – 1800
AlGaAs 620 – 900
InGaAs 850 – 3150
InPAs 910 – 3150

4.4.3.5 Lasers
The LASER (light amplification by stimulated emission of radiation), was first
demonstrated in 1960, and represented a radical departure from conventional
sources known at the time. The combination of stimulated (rather than
spontaneous) emission to provide gain and a frequency-selective feedback
mechanism gives the laser its unique properties. The stimulated emission results
from a population inversion, where there are more electrons at a higher energy
level than at a lower energy level. This population inversion can be generated
using one of several different energy sources, including electrical, optical,
magnetic, chemical, or nuclear. The frequency-selective feedback mechanism

Figure 4.22 OLED structure.11 (Reproduced by permission of Silicon Chip.)

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Generation of Optical Radiation 119

most often takes the form of a resonant cavity with mirrors. For frequencies
having overall gain greater than unity, oscillation is possible.
Lasers differ from incoherent sources in many ways. Since the amplification
process depends upon strict phase relationships, the radiation produced is
coherent, i.e., the waves are in phase with each other. Since we are dealing with
atomic transitions and a resonant cavity, the output is very nearly
monochromatic, with a very narrow bandwidth. Because of the requirements
placed on the resonant cavity, the beam diameter and the beam divergence angle
are both small.
Lasers come in all forms, shapes, and sizes. Some generate unmodulated
light [continuous wave (CW)] with power levels from milliwatt to megawatt and
above, while others generate pulses. Pulse widths can vary from several
femtoseconds (10–15 s) to several milliseconds with repetition rates varying from
gigahertz to millihertz. Size can be as small as a TO-18 transistor to as large as a
full-size laboratory.
A wide variety of materials have been found useful as gain media. Gas lasers
include He-Ne and CO2. Examples of ion lasers are argon and krypton. Solid-
state lasers are exemplified by ruby and Nd:YAG. A semiconductor laser is
similar to an LED but incorporates the resonant cavity needed for laser operation
within the semiconductor structure. Tunable lasers are available using organic
dyes and special solid-state crystals such as alexandrite. Erbium can be doped
into a fiber to provide in situ gain.

4.4.4 Natural sources

4.4.4.1 Sunlight
As Table 4.3 shows, natural sources appear in both thermal and luminescent
categories.
The most prominent natural source of radiation, indeed the most important, is
our sun. The best estimate for the solar constant is 1368 W/m2 with an
uncertainty and drift of about 0.2%. Since the orbit of the earth around the sun is
elliptical, there is an additional diurnal variation of ±3.5%, with the maximum
experienced in January.
Things are somewhat different at the surface of the earth. Our atmosphere
selectively attenuates by two primary means: scattering by molecules (Rayleigh
scatter) and aerosols (Mie scatter), and absorption by molecules (H2O, CO2, and
O3). The term “air mass” is frequently used to indicate how much atmosphere the
solar radiation is traversing; it is approximately equal to the secant of the solar
zenith angle.† Figure 4.23 shows extraterrestrial solar spectral irradiance in the
visible and near infrared, along with transmitted solar spectral irradiance for
several air masses.

†
At or near sea level. As the altitude increases significantly, and air pressure goes down, the air
mass decreases significantly.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 120 Chapter 4

12
Figure 4.23 Spectral characteristics of direct sunlight.

4.4.4.2 Skylight, planetary, and astronomical sources

Skylight is of some importance, as it contributes to both heating and illumination.
It arises from both Rayleigh and Mie scattering. The sky appears blue during
daytime because of the strong λ–4 wavelength dependence of Rayleigh scattering
at shorter wavelengths. In the morning and evening, with airmass high, longer-
wavelength oranges and reds appear. Figure 4.24 shows some typical spectral
irradiance curves. Note the dependence on solar zenith angle (z) and albedo
(surface reflectance.)

Figure 4.24 Irradiance from skylight.13 [Reprinted from Daylight and its Spectrum, S. T.
Henderson, p. 113 (1970).].

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Generation of Optical Radiation 121

Astronomers and planetary scientists are greatly interested in stellar and

planetary radiation. Optical radiation from the planets in our solar system has two
components: reflected solar radiation appearing like a 5700 K blackbody, and
emitted thermal radiation corresponding to the temperature of the planet.
Radiation from stellar bodies is even weaker, requiring ever larger telescopes to
capture enough radiation to be measured. Astronomers use a number of systems
to classify stars. Visual magnitude is a logarithmic system, wherein a difference
of 1 magnitude equals an irradiance ratio of 2.512. It has been determined that a
zero-magnitude star produces an illuminance of 2.65 × 10–6 lm/m2. The sun has a
magnitude of about –27 (the scale is upside down.)
Progressing to even weaker diffuse sources, we encounter the aurora, the
airglow, and zodiacal light. Both the aurora and airglow arise from solar
electrons and protons exciting atomic species in the upper regions of our
atmosphere. The earth’s magnetic field dictates that most of this activity occurs
in polar regions. The aurora is transient and related to intense solar activity, while
the airglow is a constant background. Both are characterized by spectral line
structure. Zodiacal light arises from solar radiation scattered by dust in the
ecliptic plane.

4.4.4.3 Application: energy balance of the earth

The temperature of the earth is determined by the combination of the energy
absorbed from the sun and the energy radiated to space. The globally averaged
earth albedo weighted by the solar spectrum is 35%. This means we absorb 65%
of the incident irradiance, or about 900 W/m2. To maintain thermal equilibrium,
we must emit about 225 W/m2 to space (recall that the ratio of the total area to
the projected area of a sphere is four). This irradiance corresponds to a blackbody
equivalent temperature of about 250 K. The radiation from the earth does not
appear as blackbody radiation because of the spectral characteristics of the
atmosphere. In the atmospheric windows (regions of relative transparency), the
radiation comes from the surface. In those regions where the atmosphere is
opaque, the radiation comes from the cold upper atmosphere. Figure 4.25 shows
the spectral radiance of a 272 K blackbody for reference (dashed curve) and the
outward radiance of the earth as seen from space (solid curve). The dotted curve
is the radiance seen when looking upward from the surface of the earth, which
includes upwelling earth radiance and whose primary atmospheric component is
emission at these wavelengths.

4.5 Radiation Source Selection Criteria

Several factors must be considered when selecting a source of optical radiation
for experimental or design use. Table 4.7 details selection criteria in the form of
questions, the first eight of which were posed by A. G. Worthing in 1937.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 122 Chapter 4

Figure 4.25 Earth radiance in the thermal infrared region. Dashed curve: blackbody at 272
K; solid curve: radiance of earth from space; dotted curve: radiance seen from earth’s
14
surface, looking up.

Table 4.7 Source selection criteria.

Does it supply energy at such a rate or in such an amount as to make

measurements possible?
Does it yield an irradiation that is generally constant or that may be varied with
time as desired?
Is it reproducible?
Does it yield irradiations of the desired magnitudes over the areas of the desired
extent?
Has it the desired spectral distribution?
Has it the necessary operating life?
Has it sufficient ruggedness for the proposed problem?
Is it sufficiently easy to obtain and replace, or is its purchase price/construction
cost reasonable?
What is its physical size? Will it fit into the allocated space?
What are its operational limitations (cooling, shock & vibration, etc.)?
Where do you plug it in?

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Generation of Optical Radiation 123

4.6 Source Safety Considerations

Table 4.8 provides tips on the safe handling of light sources.

Table 4.8 Necessary considerations for safe handling of light sources.

Ultraviolet eye damage is possible with UV sources; shield the lamp and wear
protective goggles.
There is an explosion hazard from high-pressure lamps. Use an approved
housing, clean quartz envelopes thoroughly, handle lamps with gloves, and
wear eye protection.
A fire hazard exists with powerful lamps; keep flammable materials away from
them.
Ozone is generated by ultraviolet lamps; use ozone-free lamps (envelopes that do
not transmit UV) or provide proper ventilation.
There are electrical hazards from lamp power supplies; take normal precautions,
particularly with high-voltage starters for arc sources and capacitor banks for
pulsed lasers.
Careful handling and disposal of CFLs is a must! See http://www.energystar.gov
for more information.

4.7 Summary of Some Key Concepts

All objects above 0 K radiate according to Planck’s law modified by the
directional spectral emittance term, Eq. (4.35):

 ε(λ; θ, φ)c1   1 
Lλ (θ, φ) =  5   c2 / λT .
 πλ  e − 1 

For blackbody radiation (as opposed to blackbody radiation simulator and

graybodies), the emittance term can be neglected.

Metals are poor emitters.

Opaque dielectrics are good emitters.
Transparent dielectrics are poor emitters.
Gases and glasses radiate only in absorption bands and are transparent
elsewhere.

For Further Reading

F. E. Carlson and C. N. Clark, “Light sources for optical devices,” Chapter 2 in
Applied Optics and Optical Engineering, Vol. 1, R. Kingslake, Ed.,
Academic Press, New York (1975).

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 124 Chapter 4

J. E. Eby and R. E. Levin, “Incoherent light sources,” Chapter 1 in Applied

Optics and Optical Engineering, Vol. 7, R. R. Shannon and J. C. Wyant,
Eds., Academic Press, New York (1975).
D. Kryskowski and G. H. Suits, “Natural sources,” Chapter 3 in Sources of
Radiation, G. J. Zissis, Ed., Vol. 1 of The Infrared & Electro-Optical
Systems Handbook, SPIE Press, Bellingham, Washington (1993).
A. J. La Rocca, “Artificial sources,” Chapter 2 in Sources of Radiation, G. J.
Zissis, Ed., Vol. 1 of The Infrared & Electro-Optical Systems Handbook,
SPIE Press, Bellingham, Washington (1993).
A. J. La Rocca, “Artificial sources,” Chapter 10 in Handbook of Optics, Vol. 1,
Part 4, Optical Sources, McGraw Hill, New York (1995).
H. Z. Malacara and A. Morales, “Light sources,” Chapter 5 in Geometrical and
Instrumental Optics, Academic Press, New York (1988).
J. B. Murdoch, Illumination Engineering—from Edison’s Lamp to the Laser,
Macmillan, New York (1985).
M. S. Rea, Ed., Lighting Handbook: Reference and Application, 9th Ed.,
Illuminating Engineering Society of North America, New York (2000).
J. C. Richmond and F. E. Nicodemus, “Blackbodies, blackbody radiation, and
temperature scales,” Chapter 12 in NBS Self-Study Manual on Optical
Radiation Measurements, Part 1, U.S. Government (1985).
R. Siegel and J. R. Howell, Thermal Radiation Heat Transfer, Hemisphere
Publishing Corp., New York (1981).
W. T. Silfvast, “Lasers,” Chapter 11 in Handbook of Optics, Vol. 1, Part 4,
Optical Sources, McGraw Hill, New York (1995).
R. H. Weissman, “Light emitting diodes,” Chapter 12 in Handbook of Optics,
Vol. 1, Part 4, Optical Sources, McGraw Hill, New York (1995).

References

1. F. Grum and R. J. Becherer, Radiometry, Vol. 1 in Optical Radiation

Measurements, F. Grum, Ed., Academic Press, New York (1979).
2. M. A. Bramson, Infrared: A Handbook for Applications, Plenum, New York
(1966).
3. M. Pivivonsky and M. Nagel, Tables of Blackbody Radiation Functions,
Macmillan, New York (1961).
4. M. Czerny and A. Walther, Tables of the Fractional Function of the Planck
Radiation Law, Springer-Verlag, Berlin (1961).
5. E. Hagen and H. Rubens, “Uber die beziehung des reflexions und
emissionsvermogens der metalle zu irhrem elektrischen leitvermogen,” Ann.
d. Physik, 4(11), pp. 873–901 (1903).

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Generation of Optical Radiation 125

6. J. A. Jamieson et al, Infrared Physics and Engineering, McGraw-Hill, New

York (1963).
7. J. C. DeVos, “A new determination of the emissivity of Tungsten ribbon,”
Physica, 20, p. 690 (1954).
8. “HG-2 mercury argon calibration source,” Ocean Optics, Inc.
http://www.oceanoptics.com/products/hg1.asp.
9. I. Csorba, Image Tubes, Howard W. Sams & Co, Inc., Indianapolis, (1985).
10. IES Lighting Handbook, 9th Ed., Illuminating Engineering Society of North
America, New York (2000).
11. P. Smith, “OLED displays: better than plasma or LCD,” Silicon Chip Mag.
179 (1 Aug. 2003). http://www.siliconchip.com.au/cms/A_30650/article.html
12. S. L. Valley, Handbook of Geophysics and Space Environments, Office of
Aerospace Research, USAF (1965).
13. S. T. Henderson, Daylight and its Spectrum, Elsevier, New York (1970).
14. F. Kneizys, et al., “Optical properties of the atmosphere,” Air Force
Cambridge Research Laboratory (1972).

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Chapter 5
Detectors of Optical Radiation
Section 5.7.2 contributed by Robert C. Schowengerdt

5.1 Introduction
Optical radiation detectors are transducers that transform optical radiant energy
into a different form of energy that is more readily measured. Electrical energy is
typically used for this purpose, as electrical measurement technologies are well
established. Both thermal and photon detectors convert incident optical energy
into electrical signals; in the thermal detector, the initial output takes the form of
heat before conversion. Either detector type may be a “point” or an “area”
detector. The former are single-element detectors, designed to respond to incident
energy. The latter are one- or two-dimensional arrays used particularly for
imaging, and include mechanisms to read out the signal on the array. Table 5.1
gives examples of photon and thermal detectors, while Table 5.2 lists differences
between them.

Table 5.1 Detector types.

Thermal Photon (quantum)

Bolometer Photodiode
Point
Thermocouple Photoconductor
Pyroelectric vidicon CCD array
Area Linear thermoelectric array CMOS array
Microbolometer array UV photodiode array

Table 5.2 Some important differences.

Thermal Photon
Low detectivity High detectivity
Slow response time Fast response time
Do not require cooling Typically require cooling for
IR operation

127

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 128 Chapter 5

Figure 5.1 Characteristic response curves of (a) photon and (b) thermal detectors.1
[Reprinted from Optical Radiation Measurement series, Vol. 1, F. Grum and R. J.
Becherer, Radiometry, p. 177 (1979).].

The parameters in Table 5.2 will be discussed in more detail later in this
chapter. Figure 5.1 illustrates the generic differences in spectral response
between photon and thermal detectors. Note that the thermal detector response is
wavelength independent, and that the photon detector response is a function of
several parameters including wavelength and quantum efficiency. A specific
example of the latter is the commonly used silicon detector, whose wavelength of
peak response is 950 nm and whose cutoff wavelength is near 1100 nm.

5.2 Definitions
Several terms commonly used to describe detector parameters are not often used
outside the field of optics. Recall that the term wavelength refers to the optical
regime below 100 μm. The term frequency is used to describe the direct current
(dc) to 1012 Hz audio/radio regime. The following defintions will aid our study of
detectors.
Spectral responsivity ℜ(λ) is the ratio of the output of a detector or
radiometer to that of a monochromatic source of optical radiant power. The
detector ouput is typically a current (amperes, A) or a voltage (volts, V), while
the incoming optical quantity is power, measured in watts. It is also a function of
wavelength, in that it is measured at specific wavelengths, and is correctly
reported in A/W or V/W at a specific wavelength. It can also vary as a function
of detector temperature and input power level.
Responsivity ℜ is the ratio of the output of a detector or radiometer to the
incoming optical radiant power, integrated over the spectral range which is
common to the source and detector. The input and output signals are those
described above. Since this parameter is the result of the quotient of two integrals
as shown below, it is source dependent for nonmonochromatic or nonflat sources.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 129

ℜ=
 R (λ )Φ (λ ) d λ . (5.1)
 Φ (λ ) d λ
On occasion, one will also find responsivities related to other radiometric
quantities, such as irradiance responsivity ℜE in amps (or volts) per W/m2, or
radiance responsivity ℜL in amps (or volts) per W/m2sr. These responsivities are
more likely to characterize radiometric instruments than detectors alone.
Photon spectral responsivity ℜq(λ) is the ratio of the output of a detector or
radiometer to the monochromatic optical radiant photon flux incident upon the
detector or radiometer. The output is typically a current (A) or a voltage (V),
while the incoming optical quantity is photon flux (photons/s). Therefore
ℜq(λ) carries the units (A·s) or (V·s). This parameter is wavelength dependent for
most detectors and also can vary with temperature and input power level. Like
spectral responsivity, photon spectral responsivity is a function of wavelength.
Photon responsivity ℜq is the ratio of the output of a detector or radiometer
to the incoming photon flux, integrated over the spectral range which is common
to the source and detector. The input and output signals are those described
above. Since this parameter is the result of the quotient of two integrals, it is also
source dependent for nonmonochromatic or nonflat radiation sources.
The electrical signal that is output by a detector is governed by the electrical
characteristics of the detector (resistance, capacitance, etc.) as well as its
associated circuitry. A Bode plot is a convenient way to depict a system’s
electrical frequency response, with gain or phase plotted on the ordinate and
frequency on the abscissa. For our purposes, voltage or current gain are often
depicted. Both axes are logarithmic, as seen from the generalized Bode plot in
Fig. 5.2. With the proper choice of axes, slopes associated with single time
constants (6 dB/octave, 20 dB/decade) plot as 45-deg lines, and the cut-on and/or
cutoff frequencies are at the intercepts of the straight-line asymptotes.

Figure 5.2 Bode plot.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 130 Chapter 5

Time constant τ is the time required for a signal to achieve 63% of its final
output, given a step input. For a simple single-section resistance-capacitance
(RC) circuit, τ = RC. The rise time is the time the signal takes to rise from 10%
of its final value to 90%. For a single time-constant circuit, its value is
2.2τ. Figure 5.3 illustrates this concept.
Cutoff frequency fc is the frequency at which the voltage or current response
of a detector or circuit falls to 1 2 , or 0.707× its dc or midband maximum value.
It is related to the time constant by fc = 1/(2πτ). In terms of power, it is the
frequency where the power drops to half of the dc or midband maximum value.
This is also called the 3-dB frequency.
The root-mean-square (rms) value of a quantity (voltage in this example) is
defined as

T
1 2
T 0
vrms = v (t )dt , (5.2)

where T is a single or integer multiple period for periodic waveforms. The rms
values discussed in this chapter may be values of voltages or currents and are
often referenced in descriptions of noise.
Signal is the component of the output voltage or current from a detector that
arises from a specific radiometric input. Signal is what you want to work with. It
is an rms current or voltage and is denoted vs or is. The signal is the integral of
the product of the spectral responsivity and the radiometric input such that:

∞
SIGNAL =  R(λ)Φ (λ)d λ . (5.3)
0

RISE TIME
1

0.8

0.6
OUTPUT

0.4

0.2

0
0 0.5 1 1.5 2 2.5 3
TIME (time constants)

Figure 5.3 Signal output as a function of time constant.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 131

Noise is the component of the output voltage or current from a detector that
arises from random fluctuations in the detector circuit, in the incoming radiation,
or from a number of other sources discussed in this chapter. It is characterized by
an rms current or voltage and denoted vn or in.
Signal-to-noise ratio (SNR) is the ratio of the rms signal current is to the rms
noise current in. Voltage may be substituted for current when calculating SNR,
which is dimensionless.
Impedance Z is the slope of the current versus the voltage curve of a device
at the designated operating point:

dV
Z= (οhms) . (5.4)
dI

Linearity is the proportionality of output to input for a detector or an

instrument. Linearity implies that the responsivity of a detector or instrument is
constant over a defined range of input power, irradiance, etc. Expressed
differently, it is the region in which the slope of a plot of output versus input is
equal to a constant.

5.3 Figures of Merit

Several figures of merit have been defined for optical radiation detectors. Among
the more common are the following:
Responsive quantum efficiency (RQE, η) is the number independent output
events per incident photon. This definition eliminates gain terms that could mask
the fundamental detection mechanism. It is frequently expressed as electrons per
photon, which, in the case of a detector with internal gain (i.e., a photomultiplier
or an avalanche photodiode) gives misleading information. RQE is between 0
and 1 and is often simply called “quantum efficiency.”
Detective quantum efficiency (DQE) is the ratio of the square of the output
SNR to the square of the input SNR. It is a measure of the SNR degradation
caused by the detector and has value between 0 and 1. DQE may be expressed
mathematically as

SNR 2 out
DQE = . (5.5)
SNR 2 in

Noise-equivalent power (NEP) is the incoming signal, in watts, that produces

a signal-to-noise ratio of 1. It is therefore the ratio of the rms noise current (or
voltage) to the responsivity. Like responsivity ℜ, this term results from
integration over wavelength.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 132 Chapter 5

in i
NEP = = Φ n ; or
ℜ is
(5.6)
vn vn
NEP = = Φ .
ℜ vs
Note that NEP is a useful figure of merit only when the limiting noise is
inherent in the detector, and not due to the input signal.
Spectral NEP(λ) is the NEP for a monochromatic single-wavelength input
signal. Again, this is not a derivative quantity, per-unit wavelength, but merely a
value at a specific wavelength.
Noise-equivalent photon flux NEΦq is the incoming signal in photons/s that
produces a signal-to-noise ratio of unity. It is therefore the ratio of the rms noise
current (or voltage) to the photon responsivity.
Detectivity D is the reciprocal of NEP, originally defined as a term because
“bigger is better.” The unit is W–1. It is conveniently thought of as the SNR for a
1-W input. Multiply D by the input power to get the SNR. Like responsivity ℜ
and NEP above, this term is a result of integration over wavelength:

1 SNR
D= = . (5.7)
NEP Φinput

Specific (or normalized) detectivity D* is pronounced “dee-star.” Most

detectors display noise that is proportional to (AdB)1/2, where Ad is the detector
area and B is the noise bandwidth (to be defined later). D* permits a useful
comparison of detectors of different materials, unequal areas, and different noise
bandwidths. Its units are cm Hz1/2W–1. Derived from responsivity ℜ and NEP,
this term is again a result of integration over wavelength. Its defining equations
are

Ad B Ad B vs
D* = D Ad B = = cm·Hz1/2/W. (5.8)
NEP Φ vn

In the expression above, signal and noise currents may replace signal and
noise voltages. To obtain the SNR from D*, multiply D* by the input power and
divide by (AdB)1/2.
This term is called blackbody D* if the incident power comes from a
blackbody radiation simulator. The notation used in this case is D*(T,fc,B), where
T is the blackbody radiation simulator temperature in degrees kelvin (500 K is
common), fc is the chopping frequency in Hz, and B is the effective noise
bandwidth, by convention, 1 Hz.
Spectral D*(λ) is D* at a specific wavelength. It is defined as

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 133

Ad B
D * (λ ) = cm·Hz1/2/W. (5.9)
NEP (λ)

A notation often seen is D*(λp,fc,B), where λp is the wavelength at the peak

spectral responsivity (μm), fc is the chopping frequency (Hz), and B is the
effective noise bandwidth.
D**, pronounced “dee double star,” is a further normalization of D* to a
π steradian, hemispheric, projected-solid-angle field of view. The term applies
only to what we will later define as background-limited detectors. There is a
spectral D** as well. D** is defined as

Ω
D** = D* = D* sin Θ1 2 , (5.10)
π

since Ω = πsin2Θ1/2, where Θ1/2 is the system or detector’s half-angle field of

view.
Photon D* is the specific or normalized detectivity in terms of photons:

hc AdB is
D *q = D * = cm·Hz1/2/(phot/s), (5.11)
λ n in

where n-bar in the denominator is the photon flux in photons per second.
BLIP is an acronym for background-limited infrared photodetector, the
condition in which the limiting noise in a detector output arises from background
photons. A BLIP detector’s internal noise has been reduced to the point where it
is not significant. This ideal condition allows SNR to be easily calculated.
RA product is the product of the detector resistance and area and is a
constant for many materials. D* is proportional to (RAd)1/2 for many photovoltaic
detectors; thus the RA product can be used as a figure of merit for material
comparisons.

5.4 #N$O%&I*S@E~^

5.4.1 Introduction to noise concepts

Noise places a fundamental limit on the detection process. Even if all noises
intrinsic to the detector could be reduced to insignificance, the noise associated
with the random arrival of incident photons will still limit the amount of power
that can be detected. Noise is of three types:
(1) intrinsic (i.e., noises from physical processes in detectors, associated
preamplifiers, and signal processing circuits)
(2) anthropomorphic (man-made noises, i.e. from motors, radio, TV, etc.)
(3) environmental (i.e., lightning or ionospheric effects).

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 134 Chapter 5

Assuming that (2) and (3) can be minimized using good engineering
practices (shielding, grounding, proper component and circuit layout), we
concentrate on (1). Figure 5.4 is a plot of noise spectral density versus frequency
that includes several of the above noise mechanisms.
Noise comes in many forms and colors. The most prevalent model of noise is
white noise, having a power spectrum independent of frequency. Of course real
noise is never white, as that would imply infinite total power. Noise density
actually goes to zero above 1012 Hz because of the finite mass of electrons. In
Fig. 5.4, the region at frequencies greater than about 50 Hz is white
(superimposed on spikes from fixed frequency sources). Pink noise has a power
spectrum that increases with decreasing frequency at a nominal rate of 3 dB per
octave, proportional to the inverse of the frequency (1/f ). It occurs at low
frequencies, below about 5 Hz. Noise increasing at a rate of 6 dB per octave
(proportional to 1/f 2) at low frequencies is referred to as red (or brown) noise.
There are special names for relatively rare noises that increase with increasing
frequency. Blue noise has a power spectrum that increases 3 dB per octave with
increasing frequency (proportional to f ), and purple noise has a power spectrum
that increases 6 dB per octave with increasing frequency (proportional to f 2).

Figure 5.4 Noise spectral density as a function of frequency. (Reproduced from Ref. 2
with permission of Wiley-Blackwell.)

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 135

Table 5.3 Gaussian noise distribution characteristics.

Fraction of time
Peak-to-peak value
p-p exceeded
2 × rms (±1σ) 0.32
4 × rms (±2σ) 0.046
6 × rms (±3σ) 0.0027
6.6 × rms 0.001 (1000 ppm)
8 × rms 60 ppm
10 × rms 0.6 ppm
12 × rms 2 × 10–9 ppm

Most noise can be accurately modeled as white. Also, many forms of noise,
such as thermal noise within detectors, can be effectively modeled as Gaussian.
There is no necessary correlation between the white noise model and the
Gaussian noise model. The oscilloscope trace in Fig. 5.5 shows the characteristic
appearance of Gaussian noise.
White noise without a dc component has a zero average over time, and its
peak is infinite (though with zero probability). The preferred way to characterize
this noise is by its rms value denoted by σ, which is also called the standard
deviation. Table 5.3 shows peak-to-peak values of Gaussian noise, along with the
fraction of the time that the peak-to-peak (p-p) value is exceeded.
A good estimate of the rms amplitude of the noise can be obtained by
estimating the difference between the maximum and minimum amplitudes in an
oscilloscope trace and dividing this difference by six: rms amplitude = (peakmax –
peakmin)/6. This can also be done by looking at amplitudes in the data set.
Noise can be expressed as a voltage, a current, or a power, and is best
expressed as a spectral quantity if its magnitude depends upon frequency. The
expression for a mean square noise voltage (equivalent to a noise power) looks
like
T
1
v = (vi − vavg ) =  (vi − vavg ) 2 dt .
2 2
(5.12)
T0

Figure 5.5 Gaussian noise oscilloscope trace and accompanying probability distribution.
(Reprinted from Ref. 3 with permission from John Wiley & Sons, Inc.)

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 136 Chapter 5

Its units are volts squared. In the expression, vi is the instantaneous voltage,
vavg is the average voltage, and T is the observation time. The rms noise
is vn = vn2 . Current may also be used according to Ohm’s law, v 2 = i 2 × R 2 .
If the various noises in a system have different origins, they may be
considered independent or uncorrelated. The noise powers then add algebraically,
while the noise voltages (or currents) add in quadrature:

vn2 = v12 + v22 + ... + vi2 , or

(5.13)
vn = v12 + v22 + ...vi2 .

If the noises are partially correlated, emanating from the same source, a
correlation coefficient must be included in the equation, and the resulting noise
will be greater. In any case, the total noise will not exceed the algebraic sum of
the noise voltages or currents.

5.4.2 Effective noise bandwidth

The effective noise bandwidth (ENB, but more often seen as B or Δf), first
mentioned in the definition of D*, is defined by the equation

∞
1
2 
B= G ( f )v( f ) 2 df , (5.14)
G ( f o )vo 0

where G is the power gain and fo is the frequency where G is a maximum.

Spectrally flat white noise is often assumed, simplifying the equation to

∞
1
G ( f o ) 0
B= G ( f )df . (5.15)

The ENB is an equivalent square-band bandwidth and differs from the

conventional 3-dB bandwidth, as illustrated in Fig. 5.6. It is realized by an
electrical filter operating in the audio- to low-radio-frequency range, 20 Hz to
several MHz, or possibly by the detector itself. For a typical single-section (one-
time constant) filter, the relationship between these two bandwidths is B = (π/2) ×
f3dB. A combination of white plus 1/f noise increases B, whereas pure 1/f noise or
a combination of white plus generation-recombination (G-R) noise decreases B.
These noises will be discussed in more detail later in this chapter.
For multipole electrical filters, the ENB and the 3-dB bandwidth converge as
the number of poles increases, and the filter response becomes more rectangular.
For a common two-pole low-pass filter with equal time constants (Bessel filter),
B = 1.22 × f3dB. For other filter types, e.g. Butterworth or Chebyshev, the

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 137

G(fo) 3-dB
bandwidth

Actual response

Noise
bandwidth

Frequency

Figure 5.6 Effective noise bandwidth. (Adapted from Ref. 4.)

relationship differs and must be derived via integration. If the noise is nonwhite,
integration must be performed. ENBs can also be specified in terms of time
constants. The common two-pole low-pass filter with equal time constants
(Bessel filter) has B = 1/(8τ), and the simple single-pole filter has B = 1/(4τ).
Note that the ENB is always stated in frequency space at frequencies below
1012 Hz in the electronic realm. Do not confuse this bandwidth with the passband
of an optical filter operating at frequencies greater than 1012 Hz, whose center
and passband are specified in units of wavelength, nm or μm.

5.4.3 Catalog of most unpleasant noises

Most texts give the noise equations including the ENB term. The power spectral
densities for each noise will be given first here, and then the total noise
expression that results from integration over frequency will be presented.

5.4.3.1 Johnson noise

Johnson (Nyquist, thermal) noise arises from the random motion of carriers in
any electrical conductor. Both the amplitude and frequency distribution are
Gaussian. The power spectral density of Johnson noise is given by the equation

S J ( f ) = 4kTR V2/Hz, (5.16)

where

k = Boltzmann's constant, 1.380658 × 10–23 J/K,

R = the resistance of the conductor (Ω), and
T = the absolute temperature (K).

Note that since there is no explicit frequency term, Johnson noise is “white”
at least to 1012 Hz. Integrating over frequency to obtain the mean-square noise
voltage:

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 138 Chapter 5

v 2j = 4kTRB , (5.17)

where B is the effective noise bandwidth, defined above. The rms Johnson noise
voltage is the square root of the mean square noise voltage, or

v j = 4kTRB . (5.18)

As an example, a 106 Ω resistor generates 4 μV rms for B = 103 Hz, T = 290

K. Since v2/R is power, the noise power per unit bandwidth is simply 4kT,
independent of resistance. Invoking Ohm’s law, the mean-square Johnson noise
current is given by

4kTB
i 2j = . (5.19)
R

If the bandwidth is increased, it appears as if the noise voltage approaches

infinity. Not so; at extremely high frequencies, a quantum correction must be
applied. The energy per degree of freedom kT in the noise expressions must be
replaced:

hf
kT → ( hf / kT )
. (5.20)
e −1
Then,

x
v 2j = 4kTRB ,x
(5.21)
e −1
where x = hf/kT.
It is easy to show that Eq. (5.21) reduces to Eq. (5.17) by applying the
approximation e s ≈ 1 + s when s is small, that is, much less than 1. This
substitution is only necessary for frequencies greater than 1012 Hz, so it is not
significant for most applications. This expression also indicates the linkage
between thermal noise and blackbody radiation.
In most practical applications, the noise bandwidth is established by an RC
circuit time constant. Under these circumstances, the mean-square Johnson noise
is

∞
S ( f ) df
v 2j =  , (5.22)
0
1 + (2πfRC ) 2

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 139

which integrates to

kT
v 2j = , or
C (5.23)
v j = kT / C .

Note that the expressions in Eq. (5.23) are independent of resistance R, but
dependent upon circuit capacitance C. As an example, if T = 300 K and
C = 1 picofarad (pf), then vj= 64 μV rms. Using the identity q = CV, the noise can
also be expressed as 6.43 × 10–17 coulombs (C), or about 400 electrons.

5.4.3.2 Shot noise

Shot noise appears due to the discrete nature of electronic charge and occurs
whenever current flows across a potential barrier. Its power spectral density is

S s ( f ) = 2qI dc , (5.24)

where q equals the charge on an electron, 1.60217733 × 10–19 C, and Idc equals
the direct current flowing across the potential barrier.
The mean square shot noise current is

2
ishot = 2qI dc B . (5.25)

Like Johnson noise, shot noise is also spectrally flat (“white”). As an

example, if 2 × 10–5 amps flow across a barrier, then the noise current in a 103-Hz
bandwidth is 8 ×10–11 amps, or 80 pA. If this noise current were to flow through a
105 Ω resistance, then the noise voltage is 80 μV rms. Contrary to some claims, a
potential barrier must exist in order for shot noise to be present, as it is in
photovoltaic detectors that will be discussed later in the chapter. In a purely
resistive conductor, the flow of electrons is highly correlated, not independent.
Shot noise follows Poisson statistics: the events all have the same amplitude
at low photon-arrival rates. At high rates, the amplitude distribution becomes
Gaussian.

5.4.3.3 1/f noise

1/f (contact, modulation, excess, flicker) noise is an empirical noise having
various sources. A typical power spectral density equation is

KI α
S( f ) = dc
, (5.26)
fβ

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 140 Chapter 5

where K = a “constant” for the particular technology/noise process, 1.25 < α < 4
(usually 2), and 0.8 < β < 3 (usually 1).
Note the spectral frequency dependence of this noise, which has been
demonstrated to occur at frequencies as low as 10–5 Hz and below. This is a
particularly nasty noise, ultimately limiting dc and low-frequency detection and
amplification. This form of fluctuation is ubiquitous, found throughout nature. A
few of the manifestations of 1/f noise are seen in Table 5.4.
The low-frequency form of 1/f noise is often called “drift.” Other names are
“pink noise” (β = 1), “red” or “brown noise” (β = 2), “excess,” “flicker,” and
“contact.” Causes in semiconductor detectors, for example, include nonohmic
contacts and surface impurities.
This noise is particularly insidious, as the noise power is constant in each
frequency decade. The total noise is the integral of the power spectral density
over frequency, which is proportional to ln(fhigh/flow) for pure 1/f noise. Evaluation
is easy for ac-coupled systems, but in direct-coupled systems, we must choose
some number other than zero for flow. Unless a system has a particularly long
integration time, flow of 0.1 Hz is probably sufficient. While integration of
“white” noise over time reduces white-noise effects, integration fails to decrease
1/f noise due to the decrease in flow with increased integration time.

Table 5.4 Where does 1/f noise occur?

Waves on a beach
Fluctuation in axon membrane
Earthquakes
Economic variables
Ecological time series
Self-organizing systems
Fluctuations in human heart rate
Photon counting
Most music (not Metallica!)
Frequency of rotation of earth (β = 2)
Feedback controls in nuclear reactors
Base arrangement of DNA sequences
Traffic flow (both vehicular & network)
Motion of man standing on one foot

5.4.3.4 Generation-recombination noise

Generation-recombination (G-R) noise arises from fluctuations in the rate at
which charge carriers are generated and recombined in semiconductor devices. It

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 141

includes variations in the carrier lifetime. The expression for G-R noise can be
complicated, and the specific form depends upon the detector configuration. An
example of the power spectral density of G-R noise for an extrinsic
photoconductor (to be described later in this chapter) is

4 R 2i 2 τ
SG - R ( f ) = , (5.27)
N (1 + 4π2 f 2 τl2 )

where

N = the mean number of carriers,

i = the average photocurrent due to N,
R = the resistance,
τl = the carrier lifetime, and
f = frequency.

Note that G-R noise is frequency dependent (nonwhite). Its Bode plot looks
like that of a low-pass filter.

5.4.3.5 Temperature fluctuation noise

Temperature fluctuation noise is found only in thermal detectors. This noise has
to do with microfluctuations in the temperature of thermal detectors and is the
limiting noise for this detector class. The equation is related to statistical noise in
blackbody radiation and is given by an equation of the form

4kT 2 KB
S ΔT 2 ( f ) = , (5.28)
K 2 + 4π 2 f 2 H 2

where

K = thermal conductance (W/deg),

H = heat capacity (J/deg), and
k = Boltzmann’s constant.

Thermal conductance and heat capacity define a thermal time constant τT =

H/K that limits the temperature fluctuations at higher frequencies (nonwhite).
The effect of this noise must be determined via the responsivity of the particular
detector.

5.4.3.6 Photon noise

Photon noise occurs because in a beam of light, photons do not propagate in an
orderly fashion. They obey Poisson statistics, giving the result that the mean

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 142 Chapter 5

square fluctuations in the photon arrival rate are equal to the arrival rate. The rms
photon noise equals the square root of the number of photons:

(Δn) 2 = n , (5.29)

where n is the number of photons.

The assumption of independence breaks down at extremely high photon
arrival rates (“photon clumping”) and requires a quantum correction:

ex
(Δn) 2 = n . (5.30)
ex − 1

There is still some controversy in the literature about the relationship

between photon noise and other noises. The controversy stems from the carrier
generation mechanism and its associated noise. If the generation of carriers in a
photodetector is a random process, independent of the randomness of the photon
arrival rate, then shot noise or G-R noise can exist independently from the noise
in the photon stream. Careful measurements have not yet resolved this dilemma,
but high-level measurements indicate that the noise is essentially that given by
the shot or G-R equations. The implication is that the generation process is
instantaneous, i.e., when a photon is absorbed, the associated carrier is generated
instantaneously.

5.4.3.7 Microphonic noise

Microphonic noise, acoustically generated, has two primary manifestations. The
most important is found in pyroelectric detectors, which are by nature
piezoelectric and act like microphones. A more subtle form occurs when
unanchored wiring is allowed to vibrate within a vacuum dewar, causing small
but noticeable changes in circuit capacitance.

5.4.3.8 Triboelectric noise

Triboelectric noise is a curious electrostatic noise originating from charges that
are built up in dielectrics, typically in coaxial cables used to prevent the entry of
external noises. It is particularly nasty when the cable is permitted to flex, and the
charge may take some time to dissipate.

5.4.3.9 CCD noises

CCD noises are those specific to the charge transport and readout mechanisms in
charge-coupled device (CCD) arrays, including charge-transfer efficiency (CTE)
variations, readout noise, KTC reset noise, and others. CCDs will be discussed in
more detail in Sec. 5.7.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 143

5.4.3.10 Amplifier noise

Amplifier noise is a combination of the noises typically found in the low-level
circuits that directly couple and signal-condition the detector. There are both shot
noise components due to the flow of current through p-n junctions in diodes and
transistors and Johnson noise components because of finite resistances in the
circuit. There may also be “popcorn” noise, found in some integrated circuit
amplifiers due to instabilities in the forward current gain. Amplifier noise is nasty
in that it is a form of 1/f noise with a typical exponent f 2. Other forms of 1/f noise
are also present in amplifier noise and may be seen in data sheets.

5.4.3.11 Quantization noise

Quantization noise occurs in systems requiring digitization of analog signals. It
arises from the discrete nature of the digitization process and is related to the
number of bits n in the digital word by

SIGNALmax
LSB = , (5.31)
2n

where

n = number of bits,
SIGNALmax = the full-scale signal (amps, volts, or electrons), and
LSB = the magnitude of the least-significant bit.

Since the quantization noise is proportional to the LSB, the larger the number
of bits, the lower the quantization noise.
Johnson noise, G-R noise, and 1/f noise are usually uncorrelated and
therefore add in quadrature as shown in Fig. 5.7. This figure is typical for a
photoconductive detector. The three frequencies labeled f1, f2, and f3 are the
“corner” frequencies on the curve. In particular, since a plot of responsivity ℜ
versus frequency looks like the G-R noise curve, the region where the D* is
nominally flat extends from f1 to f3 and is the typical choice for operation.

5.4.4 Noise factor, noise figure, and noise temperature

Noise factor and noise figure are terms often seen in the electrical engineering
literature. Noise factor F is the ratio of the actual noise to the theoretical
(Johnson-limited) noise:

real noise power SNRin 1

F= = = . (5.32)
ideal (Johnson) noise power SNRout DQE

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 144 Chapter 5

Figure 5.7 Combination of several significant noises.

Noise figure NF is the log (base 10) of the noise factor, given by

 real noise power 

NF = 10 × log   dB. (5.33)
 ideal (Johnson) noise power 

Noise figure is often seen in the specification of preamplifiers and is a valid

measure of the additional noise introduced by such a device. One can attempt to
artificially lower the noise figure by the simple maneuver of increasing the circuit
resistance, thereby increasing the Johnson (theoretical) noise. Only one comment
about this: don’t do it!
Another quantity often encountered is noise temperature Tn. It is the
temperature of a thermal source that provides a signal power level equal to the
noise power level. It is defined as:

vn2 + in2 R 2
Tn = . (5.34)
4kRB

5.4.5 Some noise examples

It is beneficial to see what some noise sources look like on an oscilloscope. The
figures below are taken from Motchenbacher.3 Figure 5.8 shows scope traces of
noise that is white over three finite bandwidths, listed in Table 5.5.
As can be seen in Fig. 5.8, reducing bandwidth has an effect on the peak
amplitude as well as the rms noise value. Figure 5.9 presents traces for 1/f noise
within three different bandwidths, listed in Table 5.6.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 145

Figure 5.8 Noise that is white over three different bandwidths. (Reprinted from Ref. 3 with
permission from John Wiley & Sons, Inc.)

Table 5.5 White-noise bandwidths.

Curve Bandwidth
Upper 200 KHz
Center 20 KHz
Lower 2 KHz

Figure 5.9 1/f noise at three different bandwidths. (Reprinted from Ref. 3 with permission
from John Wiley & Sons, Inc.)

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 146 Chapter 5

Table 5.6 1/f noise bandwidths.

Curve Bandwidth
Upper 2 KHz
Center 200 Hz
Lower 20 Hz

Unlike the case for white noise, limiting bandwidth does not proportionately
reduce peak amplitude for 1/f noise.*
Finally, Fig. 5.10 presents a comparison of sinusoidal signals within white
and 1/f noise, respectively. In both cases, the signal-to-noise ratio is
approximately 1. As seen in the figure, white noise has a “furry” or “fuzzy”
quality in the trace, while 1/f noise is “jumpy.”

Figure 5.10 (a) White noise and (b) 1/f noise for SNR = 1. (Reprinted from Ref. 3 with
permission from John Wiley & Sons, Inc.)

*
From the frequency plane perspective, the smoothing shown is the result of filtering out higher-
frequency terms. From the time-plane perspective, rapidly changing signals are smoothed out by
the energy storage elements in the filter (capacitors and inductors).

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 147

5.4.6 Computer simulation of Gaussian noise

Those considering computer simulations of noise should note that the random
number generator found in most programming languages and spreadsheets
generates a uniform distribution rather than a Gaussian distribution.
Consideration of the central limit theorem leads to a simple algorithm

k
xi − 0.5
y= , (5.35)
i =1 k / 12

where x = RND(1). Use of this series generates random numbers y with mean = 0
and standard deviation σ = 1. The starting point is k uniformly distributed
numbers xi, where x is between 0 and 1. A value of k = 12 is suggested, which
will give maximum values for y of 6 at the 3σ point. If we wish to generate a new
random variable y′ with mean m and standard deviation σ, we form the
expression

y′ = m + σy . (5.36)

5.5 Thermal Detectors

5.5.1 Thermal circuit

Thermal detection of optical radiation is a two-step process. Incident optical
radiation is absorbed by the receiving surface of the detector, giving rise to an
increase in the temperature of the surface. The rise in temperature can then be
detected by one of several means, which will be discussed in this section.
All thermal detectors follow the same equation relating a change in
temperature ΔT = Td – To to incident radiation and the thermal properties of the
detector and its surroundings:

αΦRT
ΔT = Td − To = , (5.37)
1 + ω2 RT2 H 2

where

ΔT = the rise in temperature of the receiving element over a local reference

temperature,
Td = the detector temperature,
To = the temperature of the heat sink,
α = the absorptance of the receiver,
RT = thermal resistance between the receiving element and the heat sink,
H = the heat capacity of the receiving element (J/deg),

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 148 Chapter 5

ω = radian frequency of the incoming signal, and

Φ = the incident radiant power, defined as Φoejωt to allow for either steady-
state or modulated-input radiation.

Because the thermal time constant τT may be written

τT = RT H , (5.38)

Eq. (5.37) may also be written as

αΦRT
ΔT = Td − To = . (5.39)
1 + ω2 τT 2

The thermal circuit is illustrated conceptually in Fig. 5.11.

A receiver having α as close as possible to 1 (a perfect absorber) provides the
best detector performance, and several black coatings provide good absorption
characteristics. For example, Lampblack (carbon) is quite black in the visible and
near IR, but becomes somewhat transparent at longer wavelengths. Metallic
blacks (gold, platinum, etc.) formed by evaporation in a poor vacuum are of very
low density (low RT) and very porous, and are black due to multiple reflections.
They are also extremely fragile, sintering at modest temperatures (~65° C for
gold black). Black paints can be effective absorbers, yet their heaviness yields
rugged but slow (high RT) detectors. Newer etched and anodized coatings and
conversion processes that yield extremely black coatings with less mass are also
available.
As seen from Eq. (5.37), the thermal resistance RT must be as high as
possible for maximum sensitivity (high ΔT.) However, high values of RT come at
the expense of speed and ruggedness. If speed and sensitivity are equally
important, H must be small. This requirement necessitates a very small structure,
a lightweight substrate, and a black coating. The ultimate limit of RT is achieved

H
RT
HEAT SINK Ad

To

Td
Figure 5.11 Thermal circuit.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 149

when conduction and convection to the surroundings are minimized. One way to
achieve this goal is to place the detector in a vacuum, support it with low-
conductivity materials, and utilize small thin connecting wires. It can be shown5
that under these circumstances

1
RT = , (5.40)
4ασAd Td 3

where Ad is the area of the detector.

The thermal conductance K expressed in W/deg is the reciprocal of RT:

K = 4ασAd Td 3 . (5.41)

Under the conditions described, the K in Eq. (5.41) is the only conductance
between the detector and its surroundings. The input power producing a
temperature change ΔT may be expressed as

Φ = K ΔT . (5.42)

Setting the signal-to-noise ratio equal to one so that Φ=NEP, and assuming
the limiting case in which the noise is due only to fluctuations in the incident
power, we obtain:

ΔT = ΔT 2 . (5.43)

The mean square value of the temperature fluctuations in the incident beam
may also be expressed as
4kTd 2
ΔT 2 = B, (5.44)
K

where B is the noise bandwidth.6 Equation (5.42) may be rewritten as

Φ = 4kTd 2 KB . (5.45)

Substituting for conductance from Eq. (5.41) and rearranging terms, we

obtain

kTd 5 σAd B
NEP = ΔΦ 2 = 4 , (5.46)
α

bearing in mind that α= ε by Kirchhoff’s law for a system at thermal equilibrium.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 150 Chapter 5

The maximum D* that can be achieved for a 300 K detector (Ad = 1 cm2,
B = 1) is thus 1.8 × 1010 cm·Hz1/2/W. This is called the Havens limit for a thermal
detector. If the detector element is cooled, this limit increases dramatically,
reaching a D* of nearly 1015 cm·Hz1/2/W at 4 K and close to 1019 cm·Hz1/2/W at
0.1 K. At these low temperatures, other noises predominate; we cannot reach the
theoretical maximum.

5.5.2 Thermoelectric detectors

5.5.2.1 Basic principles

Thermoelectric detectors operate via the thermoelectric effect, in which a
temperature difference produces a voltage difference and vice versa. These two
conditions are described below. The thermoelectric effect was first used for
optical radiation measurements (solar radiometry) by Nobili and Melloni in 1835,
with subsequent applications including early investigation of infrared spectra.
Two dissimilar metals connected in series form a thermocouple, which may
be used in a circuit as shown in Fig. 5.12. If the junctions between the wires are
at different temperatures such that T2 > T1, a current will flow around the loop in
the direction indicated. The current’s magnitude will be proportional to ΔT =
(T2 – T1); its exact value depends upon the resistance of the circuit and the
difference between the thermoelectric powers of the two metals. This
phenomenon is called the Seebeck effect, named after its discoverer, T. J.
Seebeck (1821).
Opening the circuit, as shown in Fig. 5.13, and measuring the voltage results
in the following relationship:

ΔV
S= (V/deg), (5.47)
ΔT

where S = the Seebeck coefficient, or, alternatively, thermoelectric power. The

open-circuit voltage is SΔT.

METAL 1

J1@T1 I J2@T2

METAL 2

6
Figure 5.12 Thermoelectric circuit. [Reprinted from Optical Radiation Measurement
series, Vol. 4, W. Budde, Physical Detectors of Optical Radiation, p. 101 (1983).]

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 151

METAL 1

J1@T1 J2@T2

METAL 2 J3@T3 J4@T3 METAL 2

METAL 3 METAL 3

V
6
Figure 5.13 Open-circuit thermoelectric pair. [Reprinted from Optical Radiation
Measurement series, Vol. 4, W. Budde, Physical Detectors of Optical Radiation, p. 101
(1983).]

In this case, an additional pair of metal wires represents a voltage-measuring

instrument inserted into the circuit. These wires are typically made of copper.
Two additional junctions, J3 and J4, are formed, whose thermoelectric
contributions cancel out if they are at the same temperature, according to the law
of intermediate materials commonly applied in mechanical engineering. In
addition, if the wires are homogeneous, the voltage depends only on the
temperatures of the junctions, and not on temperature distributions along the
wires.
In 1834, Jean C. A. Peltier experimented with thermoelectric circuits by
passing currents through them. He noted that one of the junctions became warm,
while the other cooled. The coefficient describing the magnitude of this effect is
called the Peltier coefficient. It is expressed as:

1  dQ 
Π=  , (5.48)
I  dt 

where dQ/dt is heat flow and I is current. This effect is exploited in

thermoelectric coolers, which are used to cool detectors, laser diodes, small
refrigerators, dew-point sensors, and many other things. The Peltier coefficient is
related to the Seebeck coefficient by the second kelvin relationship, Π = T × S,
where T is absolute temperature in degrees kelvin. This second-order effect
occurs in opposition to the Seebeck effect; that is, a current flow due to a change
in temperature causes a reduction in that temperature change. It is thus
undesirable in radiation detectors, in which it is desirable to maximize ΔT for
optimum sensitivity. Hence, such detectors are usually operated with little or no
current flow.
Signals from thermal detectors can be increased by placing several junction
pairs in series, connecting alternate junctions to blackened receivers exposed to
incoming radiation, and connecting the other junctions to heat sinks. Devices

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 152 Chapter 5

using multiple thermocouple junctions are called thermopiles and are used in
most thermoelectric transducers.
Some of the theory behind thermoelectric detectors can help to understand
their operation. Going back to the thermal equations common to all detectors, we
have

ΔT = αΦRT (dc case) (5.49)

and

α ΦRT
ΔT = (ac case). (5.50)
1 + ω2 τT 2

Neglecting Peltier cooling, the responsivities are given by

ℜ = αSRT (dc case) (5.51)

and

α SRT
ℜ(ω) = (ac case). (5.52)
1 + ω2 τT 2

If current is permitted to flow, the Peltier effect causes a reduction in

sensitivity. In that case, ΔΤ is given by

 R S 2Td 
ΔT = αΦRT 1 − T , (5.53)
 R 

where unsubscripted R is the electrical resistance. Since thermocouples are

purely resistive, the limiting noise is Johnson noise in the resistance R.
Performance can thus be enhanced somewhat by cooling. It can also be enhanced
by reducing the value of R; however, this results in decreased RT as well. The
overall effect is to decrease thermal sensitivity ΔΤ.
A better solution is to use materials having low electrical resistance and high
thermal resistance, but for most metals this is not much of an option. Instead, we
choose materials for their thermoelectric properties and then optimize either
speed or sensitivity, depending upon application. Table 5.7 lists common
thermoelectric materials, along with their thermoelectric power values.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 153

Table 5.7 Thermoelectric materials and thermoelectric power values.

Material S (μV/°C) Material S (μV/°C)

Al -0.5 Bi -60
Cu +2.7 Sb +40
Ag +2.9 Si -400*
Fe +16 Ge +300*
Constantan 38
*
variable, depends on doping

5.5.2.2 Combinations and configurations

There are many material combinations that furnish useful thermoelectric pairs,
the majority of which are designed for industrial thermometry over wide
temperature ranges. Constantan, for example, is an alloy of copper and nickel
developed specifically for thermoelectric measurement. It is paired with copper
or iron to form the popular type J and T thermocouples, respectively. The type T
thermocouple has an output voltage of approximately 40 μV for a 1° C difference
between the hot and cold junction. Seebeck worked with bismuth-antimony (Bi-
Sb) pairs. These have the highest thermoelectric power for any of the commonly
used metals, 100 μV/°C for the pair. Specially doped silicon and germanium
yield the highest output but are difficult to fabricate.
Four distinct configurations for thermocouple/thermopile radiation detectors
are shown in Fig. 5.14. All early thermopiles were fabricated by soldering or
welding fine wires (usually Bi-Sb), culminating with the Coblentz designs [Fig.
5.14(a)]. The linear models are still used for large spectroscopic detectors and the
circular for laser power meters and radiometry. The Schwarz design [Fig.
5.14(b)] features two pins of doped Si and Ge, connected by a 0.3-μm-thick
blackened gold foil. They are extensively used for small spectroscopic detectors.
A wirewound thermopile [Fig. 5.14(c)] can be fabricated by winding many turns
of Constantan wire onto a thin insulator and electroplating silver on half of it.
The junctions are not very efficient, but they are easy to fabricate.
More recent thermopile designs have been realized by vacuum evaporation
of alternating layers of bismuth and antimony onto a substrate [Fig. 5.14(d)]. One
example utilizes a thin Mylar™ substrate placed over an insulating channel. The
thickness of the Mylar determines the speed/responsivity tradeoff for a given
application. Another example utilizes micromachined silicon with a thin oxide
layer (vanadium oxide for example) or aluminum to create monolithic structures
used for thermal imaging. Typical characteristics for both wirewound and
evaporated thermopile detectors are shown in Table 5.8.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 154 Chapter 5

Figure 5.14 Thermopile configurations: (a) Coblentz, (b) Schwartz (c) wirewound, and (d)
evaporated.7 [Reprinted from Semiconductors and Semimetals series, Vol. 5, N. B.
Stevens, “Radiation Thermopiles,” pp. 300–304 (1970).]

Table 5.8 Characteristics of wirewound and evaporated thermopile detectors.

Parameter Wirewound Evaporated

Active area 1 × 3 mm to 1 × 10 mm 0.5 × 0.5 mm to 4 × 4 mm

NEP 0.1 to 1 n·W/Hz1/2 0.3 to 1 n·W/Hz1/2

D* 108 to 109 cm·Hz1/2/W 1 to 3 × 108 cm·Hz1/2/W

Time constant 4 to 400 ms 25 to 100 ms

Responsivity 0.1 to 10 V/W 10 to 50 V/W

Resistance 10 Ω to 2 kΩ 2 to 20 kΩ

Spectral range 0.3 to 100 μm 0.3 to 30 μm

Window materials SiO2, KRS-5 BaF2, CaF, KBr

Filter types Long-wave pass, bandpass

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 155

Table 5.9 Some applications of thermopile detectors.

Passive intrusion alarms

Spectral gas analyzers
Flame detection
Radiometry—laser, spectral, and
broadband
IR thermometry

Thermopile detectors are extremely versatile due to their small size, low cost,
and wide wavelength range of operation at dc and room temperature. They can be
ruggedized to survive space applications such as horizon sensing and earth
radiation budget measurements. Some terrestrial uses of thermopile detectors are
shown in Table 5.9.

5.5.3 Thermoresistive detector: bolometer

When a thermoresistive material absorbs incident radiation, it becomes warmer
and its electrical resistance changes. The resistance change can be sensed using a
device called a bolometer, first invented by S. P. Langley in 1880. The bolometer
is a resistor that possesses a high temperature coefficient of resistance, often
abbreviated as TCR and symbolized here as β, with units of K–1 . Bolometers are
fabricated from metals (the classical approach) and semiconductors (modern).
General characteristics of resistance as a function of temperature for both
material types are shown in Fig. 5.15.
The equation for resistance as a function of temperature is

R(T ) = Ro (1 + βΔT ) , (5.54)

where β = (1/R)(dR/dT) and Ro is the resistance at some nominal temperature,

often 25° C.
R
7
5
6 METAL

5
5
5
4 SEMI
5
3 T
260 280 300 320 340 36
Figure 5.15 General characteristics of resistance as a function of temperature for metal
and semiconductor materials used in bolometers.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 156 Chapter 5

As can be seen from Eq. (5.54), the change in resistance ΔR is proportional to

RoβΔT, where ΔT may be obtained using Eq. (5.37). Note that the resistances
referred to above are electrical, not to be confused with RT, a material’s thermal
resistance described earlier in the chapter.
For metals, β is quite small, on the order of 0.5%/°C. These bolometers are
rather insensitive. For the semiconducting materials, mixed oxides called
thermistors, β is proportional to T–2, making them potentially quite sensitive.
In order for a resistance to be measured, we must force a current through the
circuit and measure a voltage drop. Although many circuit types may be used, the
half-bridge circuit depicted in Fig. 5.16 is the most popular. A voltage source
drives two resistors in series, with the upper resister the load resistor, and the
lower resistor the bolometer itself. A coupling capacitor (C) is frequently used to
block the dc signal across the sensitive component, RB. This is necessary due to
the sensitivity of the component to dc voltage. As a result, modulated beams are
required for successful operation of the circuit.
The expression for signal voltage measured across the terminals is

VB RL ΔRB
Vs = ΔV = . (5.55)
( RB + RL ) 2

Taking a page from electrical engineering, we invoke the maximum power

transfer theorem to set RL equal to RB. Then,

VB RΔR  VB   ΔR   VB   RoβΔT 
Vs = ΔV = =    =   . (5.56)
( R + R) 2  4   R   4  R 

Figure 5.16 Bolometer half-bridge circuit.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 157

Substituting for ΔT from Eq. (5.50) and recognizing that for small ΔT, R~Ro,
the bolometer voltage responsivity ℜv is

Vs  VB   αβRT 
.
ℜv = =  (5.57)
Φ  4   1 + ω2 τT 2 


Inspection of this equation shows that to maximize the responsivity, we

should make RT as large as possible. However, by Eq. (5.38), this also increases
τT, making the device slower. Thus, we have a tradeoff between responsivity and
speed. Because responsivity varies linearly with bias voltage VB, we may
consider increasing the bias voltage. This also increases the current through the
bolometer, which heats it. Increased heating is typically not a problem in metal
bolometers, but can lead to thermal runaway and burnout in thermistor
bolometers. A constant current bias can prevent burnout, but the additional
electronics add noise to the circuit, making this approach unattractive unless wide
variations in ambient temperature are encountered. A better solution is to make
the load resistor a matched bolometer element and shield it from the incident
radiation.
Because the bolometer is a resistive device, Johnson noise predominates. If
the electrical contacts are less than perfect, we can also get 1/f noise. The
ultimate performance limit for the device, in which temperature fluctuation noise
predominates, is rarely seen until the device temperature approaches 4 K.
Typical room temperature bolometers exhibit D* values on the order of 109
cm·Hz1/2/W, with responsivities varying between 10 and 104 V/W depending on
the material and temperature, with time constants varying between 1 and 100
milliseconds, depending upon the size of the bolometer element. (Uncooled
microbolometer arrays, part of a relatively recent development in thermal
imaging, feature thermal time constants of a few milliseconds.) Cooling the
bolometer to cryogenic temperatures increases the D* to 1012 cm·Hz1/2/W at 2 K
and 1016 cm·Hz1/2/W at 0.1 K. Bolometers used at these temperatures are typically
fabricated from germanium (the Low bolometer) or composite materials. Another
interesting variant is the superconducting bolometer, which operates at the
superconductor transition temperature. It is extremely sensitive but has limited
dynamic range, even with active bias control.

5.5.4 Pyroelectric detectors

5.5.4.1 Basic principles

The pyroelectric detector is unusual in that it is capable of high-speed operation
and responds only to changing signals. The pyroelectric effect was first suggested
for radiation detection by Yeou Ta in 1938. Certain ferroelectric materials with
asymmetric crystal orientations display the pyroelectric effect, which is a change
in surface charge (spontaneous polarization) with temperature. The effect is not

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 158 Chapter 5

observed at constant temperature as mobile charges within the material align to

maintain neutrality. The pyroelectric coefficient p is the change in electric
polarization per change in temperature:

dPs
p= C/cm2K, (5.58)
dT

where dPs is the change in polarization.

The pyroelectric coefficient increases with temperature as shown in Fig. 5.17
until the Curie temperature is reached, when it abruptly drops to zero with
attendant loss of response. Fortunately, this is not usually a permanent condition,
and the device can often be reactivated.
The equivalent circuit is shown in Fig. 5.18(b) as a current generator in
parallel with a capacitor and a shunt (or load) resistance RL. A change in
temperature ΔT produces a charge Q such that

Q = pAd ΔT , (5.59)

where Ad = the sensitive area of the detector and p = the pyroelectric coefficient.
The pyroelectric current ip is the product of radian frequency ω and charge:

i p = ωQ = ωpAd ΔT . (5.60)

6
Figure 5.17 Pyroelectric coefficient versus temperature. [Reprinted from Optical
Radiation Measurement series, Vol. 4, W. Budde, Physical Detectors of Optical Radiation,
p. 129 (1983).]

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 159

Figure 5.18 (a) Initial circuit and (b) equivalent circuit of a pyroelectric detector, with the
8
current generator in parallel with a capacitor and load resistor. [Reprinted from
Semiconductors and Semimetals series, Vol. 5, E. Putley, “The Pyroelectric Detector”
(1970).]

The current responsivity for the pyroelectric detector may be found by

inserting the expression for ΔT in Eq. (5.37) and substituting the expression for τT
in Eq. (5.38):

ip ωpAd ΔT ωpAd αRT ωpAd αRT

ℜi = = = = , (5.61)
Φ Φ 1 + ω2 RT 2 H 2 1 + ω2 τT 2

where τT is the thermal time constant.

To determine the output voltage across a load resistor, RL in Fig. 5.18, recall
that the output voltage signal is the product of current and impedance.9 In this
case, the output voltage is given by

i p RL
v= , (5.62)
1 + ω2 RL 2C 2

where RLC = the circuit’s electrical time constant τ and ip is given by Eq. (5.60).
Therefore, the expression for voltage may be rewritten as

ωpAd ΔTRL
v= . (5.63)
1 + ω2 τ2

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 160 Chapter 5

Figure 5.19 Voltage responsivity as a function of radian frequency.8 [Reprinted from

Semiconductors and Semimetals series, Vol. 5, E. Putley, "The Pyroelectric Detector"
(1970).].

Substituting for ΔT from Eq. (5.37) and applying the definition of thermal
time constant, the voltage responsivity may be expressed as

αωpAd RL RT
ℜv = . (5.64)
1 + ω2 τ2 1 + ω2 τT 2

In this case, the responsivity increases from zero to a flat region, then
decreases at even higher frequencies. The width of the flat region depends upon
the separation of the thermal and electrical time constants. The generic voltage
responsivity behavior is shown in Fig. 5.19, while Fig. 5.20 depicts voltage
responsivity as a function of electrical frequency with load resistance as the
parameter.

5.5.4.2 Pyroelectric materials

Several pyroelectric materials and their properties are listed in Table 5.10. Tc is
the Curie temperature and FM is a figure of merit defined as

p
FM = , (5.65)
εC ′
where

p = the material’s pyroelectric coefficient,

ε = the material’s dielectric constant, and
C′ = the product of the material’s specific heat and density.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 161

Figure 5.20 Typical voltage responsivity curve for pyroelectric detectors.6 [Reprinted from
Optical Radiation Measurement series, Vol. 4, W. Budde, Physical Detectors of Optical
Radiation, p. 131 (1983).]

Table 5.10 Pyroelectric materials and their properties.

Material Tc (οC) p (C/cm2·K) FM D* (cm·Hz1/2/W) Remarks

–8 9
TGS 49 4 × 10 5200 10 High D*
Doped TGS 60 5 × 109 Best D*
LaTaO3 618 2 × 10–8 1500 6 × 108 Bulletproof
SrBaNbO3 60 to 200 6 × 10–8 750 5 × 108 Fastest
PVF2 80 2 × 10–8 1300 2 × 108 Cheap,
flexible

TGS is the acronym for the organic compound triglycine sulfate. It depoles
(loses its internal charge) readily, requiring the periodic or even continuous
application of an electric field to maintain operation. It is also sensitive to
moisture and needs protection. The D* can be enhanced by doping the material
with L-alinine.
SBN is strontium barium niobate, a mixture of the general form
SrxBa1–xNbO3. Polyvinylidene fluoride, PVF2, is a plastic film (tradename Kynar)

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 162 Chapter 5

which can be cut and formed into custom configurations. Other pyroelectric
materials include lead zirconate titanate (PZT), ceramic, barium titanate, and
barium strontium titanate (BST).

5.5.4.3 Operational characteristics of pyroelectric detectors

The outstanding feature of the pyroelectric detector is its uniform wavelength
response (depending on crystal absorption or external blackening) coupled with
high sensitivity attainable without cooling. These detectors are small and require
no external bias source. Several are available with built-in FET preamplifiers and
matched load resistors. The impedance of the device is extremely high such that
restoration of charge equilibrium is a relatively slow process.
The crystal fitted with electrodes behaves like a capacitor with a variable
dielectric. It can be used in either a current or voltage mode. Although the
response of the pyroelectric is slow compared to photon detectors, the
pyroelectric can be used in current mode for higher-speed applications when
sufficient input power is available.
The noises inherent in pyroelectric detectors include Johnson noise, thermal
fluctuation noise, and microphonics. Pyroelectric detectors are piezoelectric and
act as microphones; they must be isolated from the surrounding acoustic
environment for successful operation. Preamplifiers provide an additional noise
source.
In terms of performance, D* values greater than 109 cm·Hz1/2/W have been
achieved, as shown in Table 5.5 above, with voltage responsivities on the order
of 104 V/W. If the Curie temperature is exceeded, the following procedure may
be used to restore operation of the device:
(1) Heat the pyroelectric element to a temperature slightly above the
Curie temperature.
(2) Apply a bias voltage across the electrodes.
(3) Slowly lower the temperature back to ambient with the bias voltage
applied.
Some pyroelectric detectors will spontaneously depole, albeit slowly, at room
temperature. These detectors require periodic application of a poling bias, or
even a constant bias, to maintain proper poling.

5.5.4.4 Applications of pyroelectric detectors

Perhaps the most pervasive use of pyroelectric detectors is in infrared motion-
detection systems, in which two detectors are placed behind a Fresnel lens array
to monitor a designated area. At the equilibrium, or no motion condition, there is
no output signal from either detector. Motion of any thermally radiating object
across the detectors’ fields of view causes a change in received power in one or
both detectors, setting off an alarm or triggering the lighting of an area. A
separate photocell within the system inhibits operation during daylight hours.
Pyroelectric detectors are also employed in:

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 163

(1) tympanic thermometers

(2) laser power meters and beam profilers
(3) thermal imaging systems
(4) fire detection
(5) pollution monitoring/gas analysis systems.
Pyroelectric detectors are available as single elements and in one- and two-
dimensional arrays. Some two-element arrays are differential, with the elements
wired in opposition. This scheme enhances sensitivity for motion detection
systems. Other linear arrays are available with 128 and 256 elements. Two-
dimensional arrays have been made to 320 × 240 elements.

5.5.5 Other thermal detectors

The Golay cell was invented by M. J. E. Golay in 1947 for use in what was at
that time the relatively new field of infrared spectrophotometry. It remains one of
the most sensitive thermal detectors ever built. Figure 5.21 shows its
construction.
In operation, incoming radiation is absorbed by the blackened membrane
which heats the gas within the cell and distends the membrane, which is silvered
on the opposite side. Light from a lamp illuminates a Ronchi screen which is
imaged on the membrane. The reflected beam passes through another Ronchi
screen, and the photocell detects the modulation when the membrane moves.
The Crooke radiometer is an early example of an optomechanical detector. It
is pictured in Fig. 1.1 and consists of a rotating element with four flat vanes.
Each vane is absorptive (black) on one side and reflective (polished) on the other.
Theoretically, the rotation would be such that the reflective side would recede
when radiant energy is incident. The momentum of the incident photons would
be absorbed on the black side, and would recoil from the shiny side. This mode is
never observed, as a very hard vacuum and virtually perfect bearing would be
required. In practice, a small amount of residual gas is present inside the glass
envelope, and the gas is heated by the black side of a vane. The local gas pressure
is slightly higher at the black face than at the shiny face, resulting in a slight
pressure differential. This causes the black face to recede from the incoming
radiation.

Figure 5.21 Golay-cell-detector schematic.6 [Reprinted from Optical Radiation

Measurement series, Vol. 4, W. Budde, Physical Detectors of Optical Radiation, p. 133
(1983).]

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 164 Chapter 5

Other optomechanical detectors include the liquid-in-glass thermometer, a

bimetallic strip, and a piezoelectric bimorph.

5.6 Photon Detectors

This section describes the most common photon detection schemes: the
photoconductive, photoemissive, and photovoltaic. First, we need a brief review
of some simple semiconductor physics.

5.6.1 Detector materials

Intrinsic semiconductors are nearly pure materials with a minimum concentration
of impurities. Most have a well-defined energy gap. Photons with energies less
than the energy gap value are not absorbed, but are either reflected or
transmitted. For a photon to be absorbed, it must have a minimum energy and
will then elevate an electron from the valence band (bound state) up into the
conduction band (free state) such that it is available for conduction under an
applied electric field. Figure 5.22 demonstrates this concept schematically, while
Eq. (5.66) expresses the high wavelength cutoff beyond which electrons will not
have the energy to reach the conduction band:

hc 1.2398
λc = = , (5.66)
Eg Eg

where λc = the high wavelength limit (cutoff wavelength) in micrometers, and

Eg = the gap energy, or the minimum energy required to elevate an electron into
the conduction band. It is expressed in electron volts (eV).
Thermalization loss
Conduction band

Eg = Gap energy
Ep = Photon
energy

Ef = Fermi
level

Valence band

Figure 5.22 Illustration of valence and conduction bands, with Eg the energy necessary to
promote an electron from the former to the latter.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 165

Another way to think about the concept is simply to recall that when Ep =
hc/λ is greater than or equal to Eg, for a given material, electrons generated by
photons at wavelength λ will possess the energy necessary to elevate into the
conduction band.
In Fig. 5.22, the excess energy Ep – Eg appears as heat. In the case of a
photovoltaic cell (to be discussed in greater detail later) this thermalization loss
causes the cell’s voltage and power to decrease. It is one of the two primary loss
mechanisms responsible for the fairly low peak theoretical efficiency of about
28% for simple photovoltaic cells.
A number of interesting intrinsic semiconductor materials are candidates for
optical radiation detection, as shown in Table 5.11. For example, silicon is seen
to require a maximum wavelength of 1.1 μm, 1100 nm, and this leads to the
dramatic falloff on the right side of the photon detector curve seen in Fig. 5.1.
Most of these materials may be categorized in a straightforward manner, but the
last combination deserves special mention. Mercury telluride (HgTe) is classified
as a semimetal, with a small negative energy gap. This means that HgTe is a
conductor at room temperature, albeit a rather poor one. Cadmium telluride
(CdTe) is a semiconductor which has been exploited for visible radiation
detection. When the two tellurides are combined as Hg1–xCdxTe, then the energy
gap depends on x, the fraction of CdTe in the mix. A mixture where x = 2 is
common and yields a detector with response out to about 12 μm. The longest
wavelengths that are practical with this trimetal detector are about 25 μm, and
cold temperatures are required for effective operation.

Table 5.11 Semiconductor materials useful for optical radiation detection.

Material Energy gap (eV) Cutoff λ (μm)

SiC 3.0 0.41
CdS 2.4 0.52
CdSe 1.74 0.71
GaP 2.25 0.55
GaAs 1.4 0.89
InP 1.25 0.99
Si 1.12 1.1
Ge 0.68 1.8
PbS 0.37 3.35
PbSe 0.26 4.8
InAs 0.33 3.8
InSb 0.23 5.4
CdTe 1.6 0.78
HgTe –0.3 —
Hg1–xCdxTe variable 1 to 24
HgCdTe (x = 2) 0.1 12.4

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 166 Chapter 5

Table 5.12 Intrinsic carrier concentrations for Si and InSb.

Material T (K) T3 Eg (eV) e–Eg/kT ni

Si 300 2.7E7 1 3.3–19 1.3410
InSb 300 2.7E7 23 1.4–4 2.717
InSb 77 4.6E5 23 8.8–16 9.010

The relationship between intrinsic carrier concentration and temperature is

due to thermal activity, and is expressed as:

− Eg / kT
ni 2 = constant × T 3 × e , (5.67)

where

ni = the concentration of carriers,

constant ~ 2 × 1031 cm–6K–3 for most materials, and
kT = 0.02585 at 300 K with units of eV.

Figure 5.23 shows the intrinsic carrier concentration plotted versus

temperature, with energy gap as the parameter. As seen from the figure, an
increase in temperature produces an increase in carrier concentration, and
materials whose energy gap is large have lower carrier concentrations. Table 5.12
shows these effects for silicon and indium antimonide (InSb).

Figure 5.23 Intrinsic carrier concentration as a function of temperature for several Eg

values.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 167

Looking at the two temperatures for InSb in Table 5.12, it is apparent that the
T in the exponent in Eq. (5.67) has more influence on the calculated value of ni
than does the T3 term. Since we want sufficient carriers available for photon
excitation even at low values of ni, we must cool low Eg detectors that are
designed for long-wavelength operation. This fact explains the need, for
example, to cool HgCdTe detectors to temperatures of 77 K or lower. Detectors
should be cooled so that

Eg 600
kT < or T < , (5.68)
25 λc

with λc given by Eq. (5.66).

Extrinsic semiconductors are those in which a small amount of a selected
impurity (called a dopant) is introduced into an intrinsic material. Dopants create
additional energy levels within the intrinsic energy gap. The most common host
materials have been silicon (Si) or germanium (Ge), both from group IV of the
periodic table of the elements. Germanium has historical precedence, and was for
many years the only photon detector material available for wavelengths longer
than 5 μm. Silicon-based detectors are used extensively in visible and near-
infrared applications; their ease of integration with other monolithic circuit
components is a strong element in their favor.
Donors are atoms from group V of the periodic table which have five
electrons in their outer shell. Only four are needed to match up with Si or Ge,
however. The fifth electron is rather loosely bound to the host atom. It resides at
the donor level at 0 K, but may be thermally elevated to the conduction band at
an elevated temperature. Figure 5.24 shows this schematically.
Acceptors are atoms from group III which are short an electron when
compared with the host material. These holes (absence of an electron) are at the
acceptor level at 0 K. When heated, electrons are thermally elevated from the

Figure 5.24 Energy band structure for donor semiconductors.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 168 Chapter 5

Figure 5.25 Energy band structure for acceptor semiconductors.

valence band to acceptor levels, leaving behind a hole for conduction, as shown
schematically in Fig. 5.25. Table 5.13 lists some of the impurity dopants and
their levels and cutoff wavelengths.
The maximum practical doping for extrinsic materials is about 1 ppm, or
about 6 × 1017 atoms/cm3. At room temperature, they are nearly all used up. As
the temperature increases, the material reverts to an intrinsic conductor.
There are a number of other interesting semiconductor materials, among
them PbxSn1–xTe, that have made good photovoltaic detectors. Many alloys can
be formed from combinations of materials in groups III to V or II to VI in the
periodic chart. The materials in Table 5.14 below have been successfully used in
the fabrication of heterostructure alloys for solid-state sources and detectors.
Some are useful in the ultraviolet, others in the visible, and many others in the
infrared.

Table 5.13 Donor and acceptor levels for germanium and silicon.

in Germanium in Silicon
Dopant eV λc (μm) eV λc (μm)
Au 0.15 A 8.3 0.54 A 2.3
Cu 0.041 A 30 0.24 A —
Zn 0.035 A 35 0.26 A —
Hg 0.087 A 14 none —
Cd 0.055 A 22 0.3 A —
Ga 0.011 A 112 0.0723 A 17.8
B — — 0.045 A 27.6
Al — — 0.0685A 18.4
In 0.011 A — 0.155 A 7.4
S — — 0.187 D 6.8
As 0.013 D — 0.054 D 23
Sb 0.0096 D 129 0.039 D 32
Note: “A” denotes an acceptor and “D” denotes a donor.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 169

Table 5.14 Semiconductor materials used for solid-state sources and detectors.

AlP
AlAs
AlSb
GaN
GaP
GaAs
GaSb
InP
InAs
InSb

5.6.2 Photoconductive detectors

5.6.2.1 Basic principles

Intrinsic and extrinsic photoconductive detectors have been discussed in the
previous section. To recap, intrinsic photoconductors are typically found at
shorter wavelengths than extrinsic photoconductors, whose doping with an
impurity provides a longer-wavelength response. Figure 5.26 shows the general
layout of a photoconductive detector, with L the distance between electrodes.
The conductivity of a slab of semiconductor material is

σe = q(nμ n + pμ p ) Ω–1cm–1, (5.69)

where

q = unit of electric charge on one electron or one hole,

n = electron concentration (electrons/cm3),
p = hole concentration (holes/cm3),
μn = electron mobility [cm2/(Volt·s)], and
μp = hole mobility [cm2/(Volt·s)].
INCIDENT PHOTONS

CONTACT

Figure 5.26 Photoconductive detector structure.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 170 Chapter 5

In an extrinsic semiconductor, one carrier is present. An n-type extrinsic

semiconductor will be assumed to simplify the analysis, so that

σe = qμ n n . (5.70)

In the presence of incoming light, the conductivity is

σe = qμ n ( n + Δn) , (5.71)

where Δn = change in carrier concentration due to incident radiation.

The relative change in conductivity due to incident radiation is

Δσe qμ n Δn
= . (5.72)
σe σe

To obtain an expression for Δn, recall the definition of responsive quantum

efficiency, RQE (symbolized as η), from Sec. 5.3 as the ratio of independent
output events per incident photon. In this case, the RQE is the number of
electrons elevated to the conduction band per incident photon. The photon
irradiance on the detector is Eq, expressed in photons/s·cm2, and the carrier
lifetime is τl. The physical processes occur in a detector of thickness z. The
change in carrier concentration due to incident radiation is

η Eq τ l
Δn = . (5.73)
z

Equation (5.72) can now be rewritten as

Δσe qμ n ηEq τl
= . (5.74)
σe σe z

The relative change in conductivity can also be written in terms of detector

resistance R:10

Δσe −ΔR
= , (5.75)
σe R

with the negative sign indicating that the relative change in resistance has
opposite slope to the relative change in conductivity.
We wish to derive an expression for the voltage responsivity ℜv at
wavelength λ. To do so, we consider the placement of a photoconductive detector
in a circuit, as shown in Fig. 5.27.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 171

Figure 5.27 Photoconductive detector bias circuit.

The signal voltage out Vs is

RL
VS = VB , (5.76)
RL + RD
where

VB = bias voltage,
RL = load resistance, and
RD = detector resistance.

In Fig. 5.27, C(opt) is an optional capacitor in the circuit.

The optional capacitor placed in the circuit does not factor into the analysis,
but is included because values of Vs can be very large, and VB can get as high as
200 V. Placing a capacitor in the circuit allows a modulated signal. Note that the
device is symmetrical, and that the polarity of the applied bias in Fig. 5.27 is
unimportant.
We need an expression for ΔVs, the change in output signal voltage due to a
change in resistance. Differentiating Eq. (5.76), we obtain

−VB RL
ΔVs = × ΔRD . (5.77)
( RL + RD ) 2

Note that VB /(RL+RD) is Idc, the dc current flowing through the detector, so that

− I dc RL
ΔVs = × ΔRD . (5.78)
( RL + RD )

Eq in Eq. (5.74) is the photon irradiance on the detector of area Ad. At a

particular wavelength, it is Φλ – (hcAd)–1. Making this substitution in Eq. (5.74)

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 172 Chapter 5

and applying the results of Eqs. (5.75) and (5.78) to the definition of voltage
responsivity, we obtain

ΔVs I R R ληqμ n τl
ℜv = = dc L D . (5.79)
Φ hczAd σe ( RL + RD )

According to the maximum power transfer theorem of electrical engineering,

the output signal Vs is maximized when RL = RD. For small signals, we can call
them both R. In that case, Eq. (5.79) becomes

ΔVs I dc Rληqμτl
ℜv = = , (5.80)
Φ 2hczAd σe

where μ, the carrier mobility, may be μn as above, or more generally, μn + μp.

For good photoconductivity, we want high values for η, τl, and μ, and low
values for σe and z. As in most other practical applications, compromises and
tradeoffs are required. If τl is large, the device will have a slow response. If the
device is thin, small z, the RQE η will be reduced because of incomplete
absorption.
If a carrier in transit comes too close to a nucleus having a vacancy, it may
recombine. In that case, an electron will return to the valence band or top the
relevant donor or acceptor level. The carrier lifetime τl is the statistical time
between generation and recombination. If a photogenerated carrier arrives at one
of the electrodes without recombining, another carrier leaves the opposite
electrode in order to maintain charge neutrality. In this way, more than one
carrier may exist for each absorbed photon, a situation that constitutes a gain.
The photoconductive gain G is defined as the ratio of the carrier lifetime to the
carrier transit time

τl
G= , (5.81)
τtr

where the transit time, in turn, is defined as

l2
τtr = . (5.82)
μVB

The gain is increased by decreasing the spacing between electrodes or applying a

large electric field. The price paid here is an increase in response time.
We can also derive an expression for voltage responsivity in terms of gain.
The signal photocurrent is

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 173

λΦ λΦ τl μVB
is = ηq G = ηq ⋅ 2 . (5.83)
hc hc l

Applying Ohm’s law to the circuit in Fig. 5.27, with RL = RD as above, we

obtain

 λΦ   τl μVB R 
Vs = ηq   2 , (5.84)
 hc  l 2 

and

 λ   τ μV  R
ℜv = ηq   l 2 B  . (5.85)
 hc  l  2

Note that the current responsivity ℜi may be obtained directly from Eq.
(5.83) by dividing the signal current by the power term, such that
ℜI = ηq(λ/hc)G A/W.
We can maximize ℜv by increasing VB, but if Joule heating occurs and the
PC detector heats up, decreasing its resistance, we will burn it out!

5.6.2.2 Noises in photoconductive detectors

The noises commonly found in photoconductive detectors are Johnson, G-R, and
1/f. A typical noise expression is

  λ  kT kT 
in 2 = 4q  ηqΦ   G 2 + qG 2 N ′ + + B. (5.86)
  hc  qRD qRL 

The terms inside the bracket, according to their order, are

(1) G-R noise from incident photons (signal and background),

(2) dark current noise due to N′ thermally generated carriers,
(3) Johnson noise in the detector resistance RD, and
(4) Johnson noise in the load resistor RL.

The limiting noise is G-R noise from radiation (signal plus background).
When G-R noise overpowers all other noises, the rms signal current is

 λ 
is 2 = is = ηqΦ   G , (5.87)
 hc 

while the rms noise current is expressed as

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 174 Chapter 5

 λ 
in 2 = in = 4q 2 ηΦ   G 2 B , (5.88)
 hc 

where B is the effective noise bandwidth.

Recall from Eq. (5.8) that D* may be expressed as

i   Ad B 
D* =  s  ,
 in   Φ 

where Ad is the detector area. Noting that Φ = EAd, and rearranging terms in Eqs.
(5.87) and (5.88), D* for the photodetector becomes

 η  λ 
D *BLIP (λ, f ) =    . (5.89)
 4 E  hc 

This quantity is called D*BLIP because, as discussed in Sec. 5.3, the limiting
noise arises from incident photons. Equation (5.89) expresses a quantity
referenced to a particular wavelength and having a specific modulation frequency
f. Expressing this irradiance in terms of photon incidence Eq we also obtain

λ η
D *BLIP (λ, f ) = . (5.90)
2hc Eq

The background can be reduced by appropriate shielding and filtering using

cold filters when necessary. A detector can be background limited for a 300-K
background, but may become Johnson-noise limited for cold backgrounds.
Extrinsic detectors require more cooling for a given cutoff wavelength than do
intrinsic detectors, and their physical thickness is typically much greater, due to
the low absorption coefficient of the host material. A typical infrared
photoconductor installation consists of a liquid nitrogen-cooled dewar with an
appropriate window, and an optional cold shield and cold filter, as illustrated in
Fig. 5.28.

5.6.2.3 Characteristics of photoconductive detectors

In the infrared, PbS, PbSe, InSb, and HgCdTe intrinsic photoconductors along
with doped silicon and germanium extrinsic photoconductors are common. Each
is optimized for a different wavelength region. HgCdTe is unique in that it is an
alloy of CdTe (Eg ≈ 1 eV) and HgTe (Eg < 0 eV, a semimetal). By selecting the
composition of the alloy, almost any bandgap and therefore almost any peak

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 175

LIQUID
NITROGEN

COLD SHIELD

COLD FILTER
WINDOW
DETECTOR

VACUUM

Figure 5.28 Detector in vacuum dewar.

wavelength can be obtained. The doped silicon detectors are currently in favor
for focal plane applications as on-chip signal processing can be accomplished
with conventional silicon technology. The lead-salt detectors will operate at room
temperature, albeit poorly, but do much better when cooled to –193° C. Most
InSb and HgCdTe detectors prefer 77 K while the extrinsic detectors based on Si
and Ge require even lower temperatures. The lead-salt detectors have somewhat
slower response times than the others.

5.6.2.4 Applications of photoconductive detectors

In the visible portion of the spectrum, CdS, CdSe, and mixtures thereof are the
most common photoconductive detectors. They are reliable and rugged if
protected from the environment, fairly sensitive, though not very linear. They can
handle large amounts of power. They are most often used in industrial and
commercial photoelectric controls. In the infrared, photoconductors find use as
point and array detectors, particularly beyond 5 μm where photovoltaic detectors
are either unavailable or do not perform as well.

5.6.3 Photoemissive detectors

5.6.3.1 Basic principles

The process of photoemission relies on an external photoeffect, in which an
electron receives sufficient energy from an incident photon to physically escape a
photosensitive material called a photocathode. Photoemission can take place
from virtually any surface given sufficient incident photon energy. It was first

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 176 Chapter 5

observed in metals in the UV by Heinrich Hertz (1887). Einstein received the

Nobel Prize in physics in 1921 for his explanation of photoemission, originally
published in 1905.
Photoemission is a three-step process:

(1) a photon is absorbed; the result is a “hot” electron;

(2) the electron moves to the vacuum interface; and
(3) the electron escapes over the surface barrier to the vacuum.

In practical devices, the electron is attracted to and collected by a positively

charged anode.
The energy required for an electron to escape the surface barrier is known as
the “work function,” symbolized as φ. It is material specific. In metals, it can be
used to calculate the cutoff wavelength of an incident photon, beyond which an
electron will not obtain the necessary kinetic energy to escape the barrier:

hc 1239.8
λc = = , (5.91)
φ φ

where λc is the cutoff wavelength in nm, and φ is in eV.

5.6.3.2 Classes of emitters

Figure 5.29 schematically depicts photoemission in metals. Because no two
electrons may occupy the same energy state according to the Pauli exclusion
principle, energy states within the metal are separated into a number of closely
spaced levels.10 The lines to the left in the figure denote energy levels, the highest
being Ef, the Fermi level. Typical work functions are shown in Table 5.15, along
with the corresponding long-wavelength cutoffs.
Metallic photocathodes are used in the ultraviolet as they are stable and
“blind” to photons having lower energies. However, due to metallic reflection
and internal absorption and electron scattering, the quantum efficiency is very
low (η ≈ 0.001).

Figure 5.29 Photoemission in metals.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 177

Table 5.15 Work functions for metals.

Metal φ (eV) λc (nm)

Pt 6.3 200
W, Cu 4.5 275
Mg 3.67 340
Ca 2.71 460
Na 2.29 540
K 2.24 555
Cs 2.14 580

Higher quantum efficiency and longer wavelength operation can be achieved

with the “classical” photocathodes fabricated from alkali metals, chiefly cesium,
and with semiconductor photocathodes. In semiconductors, the photon must
impart enough energy to an electron in the valence band to reach the conduction
band and have enough energy left to escape the material into the surrounding
vacuum. The energy difference between the bottom of the conduction band and
the vacuum level is called the electron affinity Ea. Recall that the energy required
to elevate an electron from the valence band to the conduction band is called the
gap energy Eg. Their sum defines the work function for semiconductors,
φ = Eg + Ea. This is the total energy required to escape the potential barrier. In
practice, Eg should be larger than Ea to minimize scattering losses within the
material.
As was the case for metals, the cutoff wavelength can be calculated
according to Eq. (5.91). Figure 5.30 schematically depicts photoemission in
semiconductors. The electron behavior described above, for both semiconductors
and metals, may be described according to Fermi-Dirac statistics:

1
P ( En ) = ( En − E f )/ kT
, (5.92)
1+ e

Figure 5.30 Photoemission in semiconductors.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 178 Chapter 5

FERMI-DIRAC FUNCTION
1
100K
400K
0.8
300K 0K
PROBABILITY

200K
0.6

0.4

0.2

0
0 0.5 1 1.5 2
ENERGY

Figure 5.31 Fermi-Dirac function, plotted with temperature as a parameter.

where

En = the energy of the nth state,

P(En) = the probability that state n is occupied, and
k = Boltzmann’s constant.

At En = Ef, the probability of a state being occupied is 0.5. This statistical

function is plotted as Fig. 5.31.
With semiconductor materials, quantum efficiencies up to 0.3 can be
achieved at wavelengths as long as 800 nm, and a quantum efficiency of 0.01 is
found at wavelengths out to 1.2 μm. Table 5.16 gives the pertinent parameters.
The higher quantum efficiencies result from lower reflection losses and less
scattering. More recent materials based on photoconductive single-crystal
semiconductors with a thin Cs-based surface demonstrate even higher quantum
efficiency and response out to 1.65 μm. Table 5.17 shows some of those
materials.

Table 5.16 Photocathode energies and cutoff wavelengths.

Material Eg (eV) Ea (eV) φ=Eg+Ea (eV) λc (nm)

LiF 12 1 13 95
CsI 6.3 0.1 6.4 195
GaAs 1.4 4.1 5.5 225
Si 1.1 4 5.1 245
Ge 0.7 4.2 4.9 255
Cs2Te ~3.2 ~0.3 3.5 350
K2CsSb (bi-alkali) 1 1.1 2.1 590
CsSb (S-11) 1.6 0.45 2.05 605
Ag-Bi-O-Cs (S-10) 0.7 0.9 1.6 775
Na2KSb:Cs (S-20) 1 0.55 1.55 800

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 179

10
Table 5.17 Nominal composition and characteristics of various photocathodes.
(Reprinted by permission of Burle Technologies.)

JETEC Conversion Luminous

Nominal PC
response (lumen/W responsivity
composition type
designation at λmax) (μA/lm)
Ag-O-Cs O S-1 92.7 25
Ag-O-Rb O S-3 285 6.5
Cs3Sb O S-19 1603 40
Cs3Sb O S-4 1044 40
Cs3Sb O S-5 1262 40
Cs3Bi O S-8 757 3
Ag-Bi-O-Cs S S-10 509 40
Cs3Sb S S-13 799 60
Cs3Sb S S-9 683 30
Cs3Sb S S-11 808 60
Cs3Sb S S-21 783 30
Cs3Sb O S-17 667 125
Na2KSb S S-24 758 85
K2CsSb S — 1117 85
Rb-Cs-Sb S — 767 120
Na2KSb:Cs S — 429 150
Na2KSb:Cs S S-20 428 150
Na2KSb:Cs S S-25 276 160
Na2KSb:Cs S ERMA II 220 200
Na2KSb:Cs S ERMA III 160 230
GaAs:Cs-0 O — 116 1025
GaAsP:Cs-0 O — 310 200
In.06Ga.94As:Cs-0 O — 200 250
In.12Ga.88As:Cs-0 O — 255 270
In.18Ga.82As:Cs-0 O — 280 150
Cs2Te S — — —
CSI S — — —
Cul S — — —
K-Cs-Rb-Sb S — 672 125
S = semitransparent
O = opaque

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 180 Chapter 5

Table 5.17 (Continued.)

Dark
Wavelength Responsivity Quantum
Nominal emission
of maximum at λmax efficiency
composition at 25° C
response (nm) (mA/W) at λmax (%)
(fA/cm2)
Ag-O-Cs 800 2.3 0.36 900
Ag-O-Rb 420 1.8 0.55 —
Cs3Sb 330 64 24 0.3
Cs3Sb 400 42 13 0.2
Cs3Sb 340 50 18 0.3
Cs3Bi 365 2.3 0.77 0.13
Ag-Bi-O-Cs 450 20 5.6 70
Cs3Sb 440 48 14 4
Cs3Sb 480 20 5.3 —
Cs3Sb 440 48 14 3
Cs3Sb 440 23 6.7 —
Cs3Sb 490 83 21 1.2
Na2KSb 420 64 19 0.0003
K2CsSb 400 95 29 0.02
Rb-Cs-Sb 450 92 25 1
Na2KSb:Cs 420 64 19 0.4
Na2KSb:Cs 420 64 19 0.3
Na2KSb:Cs 420 44 13 —
Na2KSb:Cs 530 44 10.3 2.1
Na2KSb:Cs 575 37 8 0.2
GaAs:Cs-0 850 119 17 92
GaAsP:Cs-0 450 61 17 0.01
In.06Ga.94As:Cs-0 400 50 15.5 220
In.12Ga.88As:Cs-0 400 69 21 40
In.18Ga.82As:Cs-0 400 42 13 75
Cs2Te 250 25 12.4 0.0006
CSI 120 24 20 —
Cul 150 13 10.7 —
K-Cs-Rb-Sb 440 84 24 —

These conversion factors are the ratio of the radiant responsivity at the peak
of the spectral response characteristic in amperes per watt (A/W) to the luminous
responsivity in amperes per lumen (A/lm) for a tungsten lamp operated at a color
temperature of 2856 K.
A newer class of photoemitters known as negative electron affinity (NEA)
materials feature a special surface treatment on a p-type semiconductor substrate
to “bend” the band structure. In extreme cases, the vacuum level is below the
bottom of the conduction band. The advantages of this NEA photocathode
include longer wavelength operation and higher quantum efficiency. Figure 5.32
schematically depicts photoemission from these materials.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 181

Figure 5.32 Photoemission from NEA materials.

5.6.3.3 Dark current

Dark current is a limiting factor in photocathodes, and is indistinguishable from
photocurrent. The principal source is thermionic emission from the photocathode,
characterized by the Richardson equation:

J = CT 2 e( −φ / kT ) , (5.93)

where

J = current density in A/m2,

C = a constant, approximately 1.2 × 106, and
k = Boltzmann’s constant, 8.617385 × 10–5 eV/K.

The temperature dependence of dark current is illustrated in Fig. 5.33.

This graph also shows that cooling is required to achieve best possible
performance, especially with photocathodes designed for longer-wavelength
operation (i.e., low work function). For intrinsic semiconductor photocathodes,
substitute Ea + (Eg / 2) for φ in the Richardson equation.
Other sources of dark current are thermionic emissions from the dynodes,
leakage current between the anode and other structures, photocurrent from
scintillation from the envelope or electrode supports, field-emission current,
ionization from residual gasses in the envelope, and currents caused by cosmic
rays, environmental gamma rays, and radioisotope radiation from the envelope
and support structures.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 182 Chapter 5

RICHARDSON EQUATION
1E-6

1E-8
DARK CURRENT (A)

1E-10

1E-12

1E-14
1E-16

1E-18

1E-20

1E-22
200 220 240 260 280 300 320 340 360
TEMPERATURE (K)

Figure 5.33 Richardson equation expressing dark current as a function of temperature.

5.6.3.4 Noises in photoemissive detectors

Noise sources in photoemissive detectors include the following:

(1) shot noise from signal photocurrent,

(2) shot noise from background photocurrent,
(3) shot noise from dark current, and
(4) Johnson noise from the load resistor.

A noise expression may be developed by considering the quantities that make

up these sources.
The signal current from a photoemissive detector is

λ
is = ηqΦ q = ηqΦ . (5.94)
hc

If signal current flows through load resistor RL, then the signal voltage is

λ
vs = RL ηqΦ . (5.95)
hc

Applying Eq. (5.95) and the results from Eqs. (5.18) and (5.25), the noise
voltage is

1/2
 λ 4kT  
vn = RL  2qid + 2q 2 ηΦ + B . (5.96)
 hc RL  

The three terms in the inner bracket are as follows:

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 183

(1) shot noise due to dark current id.

(2) Shot noise due to signal + background current,
(3) Johnson noise in load resistor RL.

The signal-to-noise ratio is therefore

 λ 
η qΦ  
SNR =  hc  . (5.97)
1/ 2
 2 λ 4kT  
 2qid + 2q ηΦ +  B
 hc RL  

The ultimate limit is achieved when the dark current shot noise and the
Johnson noise from the load resistor can be reduced, leaving only the signal-
dependent shot noise. Under these conditions, the SNR is

ηλΦ
SNR = . (5.98)
2hcB

5.6.3.5 Photoemissive detector types

Photomultiplier tubes. The impact of an electron onto a secondary emitting
material releases several secondary electrons. The gain is defined as the number
of secondary electrons per incident electron; its symbol is δ. Values are ~8 in
MgO, ~9 in Cs3Sb, and variable at ~ 50/keV for GaP:Cs. Special structures called
electron multipliers arrange a series of these secondary emitting materials such
that electrons can be accelerated towards the next electrode (dynode) which has a
more positive potential. The total electron multiplier gain is δn, where n is the
number of dynodes. The gain also depends upon applied voltages.
In a photomultiplier tube (PMT), a photosensitive photocathode is combined
with an electron multiplier. In operation, a photoelectron is ejected from the
photocathode and accelerated towards the first dynode. Several electrons are
released and accelerated towards the second dynode, the third, and so on. There
are many interesting designs for electron multiplier structures, yielding up to 14
stages of gain.
There is some additional noise introduced in the multiplication process. A
noise factor (NF) may be calculated as

δ( n +1)
NF = . (5.99)
δ n (δ − 1)

For large values of δ, Eq. (5.99) becomes

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 184 Chapter 5

δ
NF = . (5.100)
(δ − 1)

This noise factor is quite small, typically less than 1.2. The gain of the
electron multiplier is essentially noise free. Table 5.18 lists some of the positive
and not-so-positive characteristics of photomultiplier tubes.
Photomultiplier tubes have found a number of different uses in areas
including photon counting, spectroradiometry, and imaging. In the latter, many
PMT-based devices have been replaced with solid-state imagers.

Microchannel plates. Microchannel plates (MCPs), useful in many UV, visible,

and x-ray applications, are disks built up from millions of microchannels, small
glass tubes whose diameters may range from 10 to 40 μm. They provide an
electron multiplication function and form the core of many image intensifier
systems, with each channel of the disk (plate) contributing one picture element
(pixel) to the resulting image. Typical MCP disk sizes range from 18 to 75 mm in
diameter with lengths between 0.5 and 1 mm.

Table 5.18 Photomultiplier tube characteristics.

PMT characteristics (good) PMT characteristics (not so good)

Large number of photocathode spectral FRAGILE! Most are made of glass
sensitivities
Detectors with S-numbers are Require stable high-voltage power
“classical” photocathodes supply (~1 kV)

Newer NEA photocathodes described Voltage divider string required

by base semiconductor material

Very fast, limited by transit time Require shielding from electrostatic

and magnetic fields
Crossed-field version confines electron May require light shielding to prevent
paths via a magnetic field photons from getting to dynodes
Quantum efficiencies from 0.01 to 0.5 Residual response to cosmic rays,
radioactive materials in tube
Can be physically large Can be physically large
Phosphorescence in window
Photocathode memory and fatigue
Photocathode spatial nonuniformity
Photocathode stability (particularly S-1)

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 185

Figure 5.34 Photomultiplier dynode arrangements: (a) circular-cage type, (b) box-and-grid
type, (c) linear-focused type, (d) venetian blind type, (e) fine mesh type, and (f)
microchannel plate.11 (Reprinted by permission of Hamamatsu Photonics K. K.)

The principle of operation of the microchannel plate is very similar to that of

the photomultiplier tube, with the difference that the microchannel replaces a
series of dynodes as the vehicle for amplification. The microchannel’s inner
surface is coated with a high-resistivity material having good secondary emission
characteristics. In operation, a primary electron entering from a photocathode
strikes the wall and causes secondary emission; this process continues until a
high number of electrons have been accelerated toward the positive electrode at
the other end of the tube.7
Fig. 5.34 shows the dynode arrangement for several photomultipliers. There
are a number of photocathode spectral sensitivities from which to choose; a
representative sample is shown in Fig. 5.35.

5.6.4 Photovoltaic detectors

5.6.4.1 Basic principles

The photovoltaic detector is a popular detector whose operation relies upon an
internal potential barrier with an electric field applied. A p-n junction in a
semiconductor material is typically used to provide this condition. The potential
barrier is formed by doping adjacent regions such that one is an n-type (donor)
region and the other a p-type (acceptor).

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 186 Chapter 5

11
Figure 5.35 Typical spectral responses of common photocathode materials. (Reprinted
by permission of Hamamatsu Photonics K. K.)

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 187

During the process of junction formation, the following events occur:

(1) Free electrons in the n region are attracted to the positive charge in the p
region and drift over.
(2) Free holes in the p region are attracted to the negative charge in the n
region and they drift over.
(3) Carrier drift leaves the n region with a net positive charge and the p
region with a net negative charge. The crystal stays neutral with no net
carrier gain or loss.

In the n-type material (dopants are As, Sb, and P), the electrons are the
majority carriers and the holes are the minority carriers. In the p-type material
(dopants are Al, B, In, and Ga), the holes are the majority carriers, and the
electrons are the minority carriers. Majority carriers are far more mobile than
minority carriers, and they are the primary contributors to current flow.
The barrier height depends upon the donor and acceptor levels and
concentrations. This is shown schematically in Fig. 5.36. The region between the
n and p regions is called the depletion region, and there is an electric field across
it.
The barrier height is calculated as

kT nn p p
φ≈ ln 2 , (5.101)
q ni
where

nn = electron concentration in the n-region (majority carriers),

pp = hole concentration in the p-region (majority carriers), and
ni2 = the intrinsic carrier concentration.

Figure 5.36 Energy levels in a p-n junction.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 188 Chapter 5

Figure 5.37 Application of forward bias to a p-n junction.

Since ni is an extremely strong function of temperature, so is φ. As T

increases, φ decreases. In the equilibrium junction, the Fermi level is constant.
If we apply an external bias across the junction, we can change the energy-
level structure. Application of a forward bias to the p-type region reduces the
barrier height by the amount of applied voltage (Vf in in Fig. 5.37). The positive
terminal of the bias source attracts carriers from the other side of the junction (n-
type) and vice versa. The consequence is a high current flow due to conduction
by majority carriers. Lowering the barrier height reduces the depletion region.
Application of a reverse bias (Vr in Fig. 5.38) to the n-type region increases
the barrier height by the amount of applied voltage. The positive terminal of the

Figure 5.38 Application of reverse bias to a p-n junction.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 189

bias source repels carriers from the other side of the junction (n-type) and vice
versa. The consequence is a low current flow due to conduction by minority
carriers. Increasing the barrier height widens the depletion region.
The equation expressing the I-V characteristic of a p-n junction diode is
derived from a continuity equation:

 βqVkT 
I d = I o  e − 1 , (5.102)
 
 
where

q = electronic charge,
k = Boltzmann’s constant,
T = absolute temperature in degrees kelvin,
V = applied voltage,
β = a “constant” to make the equation fit the data, sometimes called the
“ideality” factor, and
Io = reverse saturation current.

At 300K, q/kT is equal to 38.7. The “constant” β varies with applied voltage.
It is typically 1, but can be as high as 3.
Curves for the I-V equation are shown in Fig. 5.39 for various values of Io. A
large Io yields a large reverse current and a small forward voltage drop and vice
versa.

Figure 5.39 Current-voltage curves for a p-n junction with various Io.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 190 Chapter 5

The expression for the reverse saturation current Io is made up of terms

involving the minority carrier concentrations, the Einstein diffusion constants,
the minority carrier diffusion lengths, and the minority carrier lifetimes:

 n p Dn pn D p 
Io = q  +  Ad , (5.103)
 Ln Lp
 
where

q = electronic charge,
np = minority carrier (electrons) concentration in the p-region,
Dn= Einstein diffusion constant for electrons,
Ln = minority carrier (electrons) diffusion length in the p-region,
pn = minority carrier (holes) concentration in the n-region,
Dp = Einstein diffusion constant for holes,
Lp = minority carrier (holes) diffusion length in the n-region, and
Ad = detector area.

The Einstein diffusion constant D is defined as

kT
D= μ, (5.104)
q

and has units of cm2/s with μ being carrier mobility. Like D, it may be
subscripted with n or p to specify electrons or holes. The minority carrier
diffusion length is

L = D τl , (5.105)
REVERSE SATURATION CURRENT vs. TEMP

0.0001

1E-05

1E-06

1E-07

1E-08

1E-09
260 270 280 290 300 310 320 330 340 350 360 370 380 390 400
TEMPERATURE

Figure 5.40 Reverse saturation current versus temperature for a typical p-n junction.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 191

where τl is carrier lifetime. The reverse saturation current is strongly temperature

dependent as shown in Fig. 5.40.
Note that Figs. 5.39 and 5.40 combine to show that as the temperature
increases, the saturation voltage decreases. Thermalization losses will heat the
diode or PV cell. The result is that the voltage goes down, the cell current goes
up slightly, and the power goes down.
Incident optical radiation generates a current through the diode, which adds
to the dark current and shifts the I-V curve downward, as seen in Fig. 5.41. The
expression for the total current becomes

 qV 
I = I o  e βkT − 1 − I g , (5.106)
 
 

where Ig is the photogenerated current, expressed as

λ
I g = η qΦ q = η q Φ. (5.107)
hc

As seen from Eq. (5.107), the current generated is directly proportional to

photon flux and reduced by the detector’s quantum efficiency. At a given
wavelength, the current is also directly proportional to the incident power, and
also directly reduced by quantum efficiency. A typical set of I-V curves for
various incident power values is shown in Fig. 5.41.
There are several ways in which one can operate a photovoltaic detector. The
first, and one of the most important for radiometry and photometry, is called
short-circuit-current mode. In a practical sense, this is accomplished by
connecting the detector to a transimpedance amplifer, which is easy to make with

Figure 5.41 I-V curve for a photodiode at several light levels.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 192 Chapter 5

Figure 5.42 Output voltage versus current for Ig >> Io.

an op amp and feedback resistor. This mode is represented mathematically by

setting V = 0 in Eq. (5.106). The result is that I = –Ig, in which current generated
is a linear function of incident radiant power. This linearity is easily
demonstrated to seven decades in radiometric quality silicon photodiodes, though
unsubstantiated claims place the number of decades at 14.
Open-circuit voltage is another operating condition for a photovoltaic
detector. This is represented mathematically by setting I = 0 in Eq. (5.106) and
solving for V:

β kT  I o + I g 
Voc = ln  . (5.108)
q  Io 

If Ig >> Io, which is almost always the case, then Voc is logarithmic with
radiant power as seen in Fig. 5.42. If Ig << Io, then Voc becomes linear with
incident power as shown in Fig. 5.43. If Io ~ Ig, operation is intermediate between
linear and logarithmic.
Photovoltaic detectors are also commonly used as power generators. Placing
a load resistor RL directly across the detector causes the I-V curve to enclose an
area, as seen in Fig. 5.44. Short-circuit current, graphically depicted as the
location where the I-V curve crosses the current axis (V = 0), and open-circuit
voltage, where the I-V curve crosses the voltage axis (I = 0), form the two
extreme points of the (inverted) I-V curve; in between, both current and voltage
are available simultaneously. This is the requisite condition for power generation,
and devices operating in this region are commonly called solar cells. The lower
curve in Fig. 5.44 is the power versus voltage characteristic for a particular cell,
in which the maximum power is achieved at about 0.48 V.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 193

Figure 5.43 Output voltage versus current for Ig << Io.

The open circuit voltage Voc is a result of the forward-biased rolloff seen
generically in Fig. 5.39. For silicon, this voltage is close to 0.6 V as in Fig. 5.44.
The short-circuit current Ig is proportional to the incident radiation Φ as shown in
Eq. (5.107), and the radiation is in turn proportional to the area of the cell. The
cell in Fig. 5.44 with an Ig of 50 mA is quite small with an area of about 1 cm2.
Cells for typical commercial solar panels or modules have an area of around 100
or more cm2, producing a short-circuit current of around 5 A. A typical panel has
perhaps 72 cells connected in series to produce an open-circuit voltage near 40 V
and a short-circuit current near 5 A. The ratio of the maximum power to the
product VocIg is called the “fill factor,” (not to be confused with fill factor in
CCDs, below) which is typically on the order of 0.75 or so.

SOLAR CELL OPERATION

60

50

40
POWER x 2
MILLIAMPS

30

20

10

0
0 0.1 0.2 0.3 0.4 0.5 0.6
VOLTS

Figure 5.44 Photovoltaic solar cell operation (lower curve) and I-V characteristic (upper
curve).

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 194 Chapter 5

Figure 5.45 Reverse-biased photodiode I-V behavior.

The photovoltaic detector may also be operated in reverse-bias mode, in

which I = –Io – Ig. This mode of operation requires an external voltage source and
a load resistor, both of which contribute to noise. An I-V curve for this mode of
operation is seen in Fig. 5.45, showing the load line of the resistor and the bias
source VB. In dark conditions, the diode resistance is very high, and most of the
voltage drop occurs across it. With the addition of light, the diode resistance
drops, and a higher fraction of the applied voltage is found across the load
resistor. Note, however, that the curves in Fig. 5.45 are idealized, and that a real
diode will eventually break down for sufficiently negative voltages due to the
Zener effect. There are advantages and disadvantages of operating the
photodiode in reverse-bias mode. Some of these are detailed in Table 5.19.
Most manufacturers refer to reverse-bias operation as operation in
“photoconductive mode.” This term is confusing, as photovoltaic and
photoconductive detectors operate very differently. More accurate terms are
“reverse-biased” or “back-biased.”

Table 5.19 Advantages and disadvantages of reverse-bias operation.

Advantages Disadvantages
Better long-wavelength response due Presence of Io and its temperature
to less recombination dependence
Increased speed due to E-field, which Requires relatively stable, quiet voltage
sweeps carriers out source and low-noise RL
Increased speed due to lower junction Lower SNR due to noises from RL and
capacitance C=const (V–0.6)n, VB
–1.2 < n < –1.3

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 195

A common approach to operating a photovoltaic detector in the short-circuit

mode is to connect it to a transimpedence amplifier, which is easy to make using
an op amp and a feedback resistor.

5.6.4.2 Responsivity and quantum efficiency

Figure 5.46 depicts the ideal and real current responsivities for a photovoltaic
detector. Recalling that the current responsivity is equal to the current out of the
detector divided by the input power, or equivalently ℜi = ηqλ(hc)–1, it is apparent
that nonideal (less than 1) quantum efficiency η drives the nonideal response.
There are a number of reasons why the quantum efficiency is less than 1,
including:

(1) Unabsorbed photons of wavelength λ > λc (i.e. photons at longer

wavelength than the cutoff wavelength)
(2) Unutilized electron-hole pairs created beyond diffusion length (depletion
region can be widened via reverse biasing the photodiode)
(3) Surface recombination of carriers (can be reduced with a dielectric
coating)
(4) Optical losses due to transmission and reflection (can reduce with
antireflection coating)
(5) Heating of the device due to the fact that most photons have more energy
than needed to create an electron-hole pair
(6) All efficiencies are less than one.

0.8
RESPONSIVITY (A/W)

0.6

0.4

0.2

0
0 0.2 0.4 0.6 0.8 1
WAVELENGTH (um)

Figure 5.46 Ideal (straight line) and actual (curved line) current responsivity versus
wavelength.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 196 Chapter 5

5.6.4.3 Noises in photovoltaic detectors

Noise sources in photovoltaic detectors include the following:

(1) shot noise from current flow across the potential barrier,
(2) Johnson noise, and
(3) 1/f noise.

The most important noise source is shot noise. Shot noise current arises from the
signal, background, and the device’s own dark current. The rms shot noise
current is

ishot = [ 2q (is + iback + id ) B ] .

1/ 2
(5.109)

Johnson noise can be minimized by choosing a detector with high dynamic

resistance Rdyn which varies with voltage. Cooling the detector reduces Johnson
noise due both to the decrease in temperature and the increase in Rdyn with
decreasing temperature:

4kTB
iJ = . (5.110)
Rdyn

The dynamic resistance is low at forward bias and high at reverse bias. It is
calculated as

dV βkT − qV /( Ad kT )
Rdyn = = e . (5.111)
dI qIo

At the important zero-bias operating point it becomes

dV βkT
Rdyn = = . (5.112)
dI qI o

Although it appears that Rdyn is linear with temperature, the exponential

nature of I as a function of temperature predominates, and the dynamic resistance
increases with decreasing temperature. Longer wavelength photodiodes have
higher Io, lower Rdyn, and are noisier.
The insidious 1/f noise is troublesome, as always. It cannot be avoided due to
signal current, but can be minimized by operation at a higher frequency due to
the following relation:

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 197

1/2
 ( constant )( I dc df ) 
i1 f =   . (5.113)
 f 

To calculate D*BLIP for the photovoltaic detector, we need an expression for

signal current. Recall that is = Φℜi where ℜi is the current responsivity.
Expressed in terms of power,

λ
is = ηq Φ. (5.114)
hc

Noise current, which is shot-noise dominated as stated above, may be

expressed as a function of the signal current, in the visible portion of the
spectrum:
ishot = ( 2qis B ) ,
1/ 2
(5.115)

or as a function of the background, in the infrared portion of the spectrum:

1/ 2
 λ 
ishot =  2ηq 2 ΦB  . (5.116)
 hc 

Inserting the expression for the signal current and the shot-noise current Eq.
(5.116) (for the infrared) in Eq. (5.8) for D* results in

* ηλ
DBLIP (λ , f ) = , (5.117)
2 Eback hc

which is the background-limited expression, applicable in the infrared, where

Eback refers to the background irradiance. Similar logic may be followed for
photon flux, beginning with Eq. (5.114). In that case,

* λ η
DBLIP (λ , f ) = . (5.118)
hc 2 Eq ,back

Comparing Eqs. (5.89) and (5.117), and (5.90) and (5.118), we see that the
photovoltaic detector is better (higher D*) than the photoconductive detector by a
factor of 2. The physical reason for the difference is the absence of G-R noise
in the photovoltaic detector. Recombination takes place within the
photoconductive detector itself; in the photovoltaic detector, it occurs in an
external ohmic circuit where carriers are not statistically correlated. An exception
to this rule arises when photoconductive detectors are operated in “sweepout”

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 198 Chapter 5

mode. In this case, carriers are swept out of the detector before they recombine,
and the D* is equal to that of a photovoltaic detector.

5.6.4.4 Photovoltaic detector materials and configurations

Table 5.20 lists some materials that have proven useful for photovoltaic
detectors. There are a number of configurations for photovoltaic detectors. For
high-speed operation, a conventional photodiode requires a large reverse bias to
decrease device capacitance and create a large drift field. Table 5.21 lists several
common configurations; more detailed information about these may be found in
the references at the end of the chapter. Moreover, photovoltaic materials are
increasingly finding application in solar-cell technology. Some of these materials
are listed in Table 5.22.

Table 5.20 Some photovoltaic detector materials.

Material Bandgap (eV) Center wavelength (μm)

GaP 2.4 0.52
GaAs 1.4 0.93
Si 1.12 1.1
InGaAs 0.73 1.7
Ge 0.68 1.82
InAs 0.28 3.5
InSb 0.16 5.5
HgCdTe variable variable
PbSnTe variable variable

Table 5.21 Common photodiode configurations.

Configuration Comments
p-i-n Built-in depletion region, with an intrinsic layer of Si
between p and n materials
Avalanche Operate under a large reverse bias; fabricated from Si, Ge,
and InGaAs. Can be cooled to 77 K and biased beyond
breakdown point.
Schottky barrier Created by depositing a thin semitransparent metal electrode
on top of a semiconductor material. Particularly useful for
large-area UV detectors.
Inversion layer Created by doping top layer on p-type silicon with a material
having a positive charge to form an n-type material.
Quantum efficiency approaches 1 at short wavelengths.
Ultraviolet Include Schottky barrier, front-illuminated PtSi, and
AlxGax–1N
Infrared Include InGaAs, InAs, HgCdTe, and PbSnTe. Require
cooling beyond 3 μm.
Position sensing Output as a function of position on the detector, often used in
photodiodes (PSPD) tracking applications.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 199

Table 5.22 Photovoltaic materials for solar cells.

Material Comments
Monocrystalline silicon n/a
Poly- or multicrystalline silicon n/a

Ribbon silicon Formed from flat thin films of molten silicon

New structures based on silicon Special silicon arrangements to improve efficiency
Cadmium telluride (CdTe) Large-scale manufacturing and delivery possible
CIGS Copper indium gallium diselenide
CIS CuInSe2
GaAs multijunction High efficiency; have been used in satellites
Amorphous silicon n/a

5.7 Imaging Arrays

5.7.1 Introduction
Many applications require spectral and spatial information that the use of a
single-element detector cannot reasonably provide. In order to obtain this
information, some imaging systems use single detectors along with scanning
optics and other components. More commonly, linear or area (2D) arrays of
detectors are used, with the area array in widest use. A comparison of the
functional differences of single and multiple detectors is shown in Table 5.23.
Array imagers are relatively recent attempts to emulate human vision. Other
innovations throughout history are listed in Table 5.24.
This section will review concepts important to array detection, including
history, basic array parameters, device architecture, and specific array types
along with their applications. Due to the large and growing amount of material on
this subject, the reader is strongly encouraged to investigate the references at the
end of the chapter.

5.7.2 Photographic film

5.7.2.1 History
One of the earliest man-made optical detectors was photographic film. The
treatment here is brief because film has been supplanted in most scientific and
consumer applications by solid-state electronic detectors, as described in this
chapter. It is still used, however, in some specialized applications. For more
information on photographic film, the reader is referred to two thorough books
on the application of photography in science and engineering (Ray, 1999) and the
theory and technology of photography (Stroebel et al, 2000).

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 200 Chapter 5

Table 5.23 Comparison of single- and multiple-detector capabilities.

Single detectors Multiple detectors

Gather information from one Gather simultaneous measurements in
wavelength or field of view at a time multiple spectral bands, multiple
fields of view, or both
Provide instantaneous measurement Utilize integration over time for all
pixels in array
Require a simple electronic interface Require more complex electronic
interface (serial or parallel)
Need complex scanning scheme due to Simple or no scanning possible; array
the presence of a single element with multiple elements can be
utilized as a staring array
One detector, one response Magnitude of response varies due to the
presence of multiple detector
elements and nonuniformities in the
array

Table 5.24 Vision emulation efforts.

Imaging mechanism Time period

Film photography 19th to 20th century
Mechanical scanning systems 20th century
Early television
Remote sensing systems

Imaging vacuum tube devices 1930s and onward

Solid-state array imagers 1970s and onward

Johann Heinrich Schulze discovered that silver nitrate (AgNO3) darkened

when exposed to light and published a scientific paper describing the
phenomenon in 1727. The technologies behind photographic film were
developed over several decades in the 19th century. The earliest “permanent”
recorded film image is generally credited to Joseph Nicéphore Niépce in 1827.
The photograph required an eight-hour exposure to sunlight to form. Various
pioneers, including Louis Jacques Mandé Daguerre and Henry Fox Talbot,
further developed the technology, eventually allowing image recordings onto
paper. Toward the end of the 19th century, George Eastman invented roll film and
factory processing, thus commercializing the technology for routine consumer
use. Photographic film was the most prevalent image recording technology for
the next 100 years or so, when electro-optical detectors began to dominate in
most fields of science and consumer applications.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 201

5.7.2.2 Physical characteristics

Film is a “sandwich” of materials, whose primary layers are a stable base of
plastic or glass topped by a gelatin emulsion layer containing silver-halide
crystals. The base thickness is in the range 50 to 100 μm, and the emulsion layer
is about 5 μm thick for black and white film and about 15 μm thick for color
film. Various refinements are included by additional layers to protect the
emulsion, adhere the emulsion to the base, and reduce backscattering from the
rear surface of the base. The film is exposed to light by a camera shutter, and
subsequent chemical processing (development) converts the silver-halide crystals
to silver, thus darkening the image in proportion to the amount of exposure (for a
negative film).
The spectral sensitivity characteristics are controlled by the chemical
characteristics of the emulsion. The radiometric calibration characteristics of the
final image are determined in part by the emulsion, but also by the chemical
processing of the exposed film.

5.7.2.3 Spectral sensitivity

Black and white (panchromatic) films typically respond to light wavelengths of
about 0.4 to 0.65 μm. The sensitivity for a particular type of film depends on the
details of its construction, chemical compounds, and processing chemistry.
Kodak has developed films sensitive to near-IR wavelengths out to about 0.9 μm;
such films are able to record the high reflectance of vegetation in the near IR.
Color films have three different sensitive layers in the emulsion that respond
to red-, green-, and blue-light wavelengths. Development of the exposed film
results in red, green, or blue transparent dye in the corresponding layer of a
positive (slide) film, or cyan, magenta, or yellow in the corresponding layer of a
negative film.
The fairly broad spectral response of film cannot be easily refined by
chemical modifications, so lens spectral filters are commonly used to limit the
spectral response. For example, a yellow (haze) filter is used to remove blue
scattered skylight, or a blue-green blocking filter is used to restrict IR film to the
red and near-IR regions.

5.7.2.4 Radiometric calibration

A scientific procedure for calibration of film was established by Hurter and
Driffield in 1890. They proposed a series of regular exposure steps (total
exposure is irradiance multiplied by exposure time) onto film and measurement
of the resulting optical density D, defined as the negative logarithm of the film
transmittance T:

D = − log(T ) . (5.119)

A plot of D for each step versus the logarithm of the corresponding exposure
log(E) is known as the “D-logE” or “H-D” (for Hurter Driffield) curve. The H-D

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 202 Chapter 5

curve essentially shows output (optical density) versus input (log exposure)
calibration. It depends not only on film type, of course, but also on processing
conditions (chemical temperatures, processing time), exposure spectral
characteristics, density measurement procedures, and many of the variables
associated with photography. The main characteristics of the H-D curve are a
saturation toe at low exposure (the “base + fog” level, typically a D of about 0.1),
a linear segment for moderate exposure, and a saturation shoulder at high
exposure (typically a D of about 3.0). The equation of the linear segment is

D = γ log( E ) + D0 , (5.120)

where D0 is the projected intercept (usually negative) on the density axis (not the
“base + fog” level), and gamma γ is the gain of the film, synonymous with its
contrast; i.e. a low gamma means a low-contrast film. The H-D curve describes
“macrocalibration” for relatively large-area measurements of a millimeter or
more. If calibration of film is attempted for “micro” conditions of measurement
of a tenth of a millimeter or less, other factors come into play, for example, the
adjacency effect where nearby exposure affects the density at a point of interest.

5.7.2.5 Spatial resolution

The spatial resolution of film is not defined geometrically, as with electronic
detector arrays. Film resolution is affected by light scattering within the
emulsion, size of the developed silver grains, thickness of the emulsion, etc.
Consequently, it has been traditionally measured by exposure of a specific target
(for example, three high-contrast bars in a series of decreasing spacing and
width) and determination of the smallest target that can be visually identified.
A modulation transfer function (MTF) of film can be measured by exposing
a series of sine wave targets of varying frequency and contrast. However, the
intrinsic nonlinear characteristics of film are problematic for the application of
MTF, which assumes a linear imaging system. For example, the adjacency effect
causes the MTF to go above one for low- to midrange spatial frequencies.

5.7.2.6 Summary
Film remains a unique image recording mechanism in that the recorded and
developed image is the archival medium itself, i.e. no additional processing is
necessary to save the original image for long periods of time. It is also an
efficient detector in terms of its combination of large format and high resolution.
For example, a large-format aerial photograph 10 × 10 in. with a resolution of 10
μm contains some 645,160,000 resolution cells, or pixels. At the time of this
writing, such large monolithic electronic detector arrays are not possible
(although mosaics of individual arrays can achieve this size).
However, the inconvenience, delay, and cost of chemical processing and
conversion to a digital format by scanning and the associated quantitative
difficulties in image measurements have seriously disadvantaged film relative to

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 203

electronic detectors for applications requiring quantitative measurements. Steady

improvements in the spatial density of detector arrays and computer processing
capabilities continue to erode the role of film in scientific imaging.

5.7.3 Electronic detector arrays

5.7.3.1 History
One of the earliest array devices was the linear photodiode array marketed by
EG&G Reticon. It used silicon photodiodes in the photon flux integrating mode
coupled with field-effect transistor (FET) switches. These were driven by
clocked shift registers and stored electrons in the device capacitance until
readout. During the readout phase, electrons were transferred to an output signal
amplifier. Similar linear-array devices today find application in spectroscopy,
astronomy, grocery and department store scanners, and many other products.
The first large-area imaging arrays were built in the late 1960s using x-y
addressable photodiodes, phototransistors, and photoconductors. They were not
particularly successful due to responsivity nonuniformities and spatial noise. In
addition, only instantaneous readouts were employed, and the signal could not be
integrated over time. Advantages were good area utilization and random access
to pixels within the array.
Charge-coupled devices (CCDs) allowed integration of photosensitive
elements with complex readout mechanisms. Photodetectors within a CCD
operate in integration mode, with the outputs serially clocked at high rates to a
single readout circuit. CCDs having on the order of 100 megapixels (MP) have
been fabricated; this number will increase with technology development over
time. Some of the parameters characterizing imaging detector arrays are listed in
Table 5.25.

5.7.3.2 Device architecture description and tradeoffs

Although CCD technology has achieved focal plane arrays with high pixel
counts, high fill factors, high sensitivities, and charge-transfer efficiencies greater
than 0.99999 (the fraction of stored charge transferred out of the array),
complementary metal oxide semiconductor (CMOS) technology competes
effectively with CCDs in the large-format array market. This is due primarily to
the ability of CMOS manufacturing to take advantage of already-existing
fabrication lines developed for commercial microprocessors. As a result, the cost
of manufacture is lowered and the production yield of CMOS-based imagers is
improved, relative to imagers based on CCD technology.
CMOS devices can be active pixel sensors, employing signal processing
functions, including amplification, on chip. This configuration requires more
hardware, however, and reduces the fill factor from approximately 80% (in a
passive-pixel CMOS) to between 30 and 50%. To increase the amount of light

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 204 Chapter 5

Table 5.25 Array parameters.

Number and size of individual detector elements

Overall size of array
Format (aspect ratio) of array (e.g. 4:3, 16:9)
Instantaneous field of view (IFOV) and array field of regard
Fill factor (portion of an array sensitive to light)
Scanning or staring array
Spatial resolution (MTF)
Responsivity
Spectral characteristics
Dark current
Integration time
Noise amount and type (fixed pattern, readout, etc.)
Dynamic range
Linearity
Pixel data rate, frame rate, bandwidth
Noise-equivalent temperature difference (NETD) for infrared arrays
Pixel response uniformity
Information capacity
Charge transfer efficiency (for CCDs)
Long-term storage potential of detector elements
Auxiliary equipment necessary to provide maximum utilization of
information

detected, and minimize the possibility of light falling onto nonphotosensitive

areas of the chip, arrays of microlenses that focus the incoming radiation are
often used.
Charge-injection devices (CIDs) are also finding increasing use in large-
format arrays. They’ve been around since the 1970s, but exploitation of their
unique properties has only recently gained momentum. One of these properties is
the capability to stare at scenes for long exposures without spillover of charge
into adjacent pixels (“blooming”).
CID pixels are individually addressable, with charge dumped (injected) to the
substrate after each read. By contrast, CCD readouts involve transfer of charge
from site to site during the readout process.

5.7.3.3 Readout mechanisms

CCD. CCDs can utilize one of several readout mechanisms; the three described
below are most common. In a full-frame architecture, the simplest to make and
operate, serial and parallel shifts move the charge through the array to an output
amplifier. This is shown schematically in Fig. 5.47. Rows of charge are shifted

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 205

Serial clocks

SERIAL
REGISTER
Analog
Output

FULL
Parallel ARRAY
clocks

Parallel
shift direction
Figure 5.47 Full-frame readout.

upward to the serial-shift register in time with the parallel clock signals depicted
at left. Upon reaching the serial-shift register, charge is read serially into the
output amplifier. The process is repeated until all rows are read out. It requires
some form of shuttering of the array, so that light collection does not occur
simultaneously with charge transfer.
Frame transfer architecture, depicted schematically in Fig. 5.48, is similar to
the full-frame architecture with the inclusion of a storage array. As in the case of
the full frame, rows of charge are shifted toward the serial register, where charge
from each element is read out sequentially in accord with serial clock signals.
The storage array seen in Fig. 5.48 is not photosensitive, and light collection and
integration in the image array can occur simultaneously with integration in the
storage array. As a result, there is no need for a shuttering scheme.
A weakness of this mechanism is that integration within the image array is
still occurring while transfer to the storage array takes place, resulting in image
smear. Because twice the silicon is used as in the full-frame device, a frame-
transfer device costs more. On the other hand, higher data rates are achieved.
In the interline transfer architecture, shown schematically in Fig. 5.49,
photosensitive areas and readout registers are arranged in successive columns.
After detection and integration takes place in the photosensitive area, charge is
transferred to column registers, which are then clocked to the serial register as in
the previous two architectural techniques. This architecture allows very fast
response time, but is comparatively difficult to fabricate. Although image smear
as in the frame transfer device is not entirely eliminated, the amount of smear is
reduced.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 206 Chapter 5

Serial clocks

SERIAL
REGISTER
Analog
STORAGE Output
Parallel ARRAY
clocks (MASKED)
(storage)

Parallel
clocks IMAGE
(image) ARRAY

Parallel
shift direction
Figure 5.48 Frame-transfer readout.

CID. Pixels are individually addressable in a CID. The charge remains intact in
each pixel after its signal level has been determined, making for a nondestructive
readout. Row and column electrodes are shifted to ground to make way for
collection of the next image frame, and the charge is “injected” to the substrate.
This capability for individual pixel control is valuable in many imaging
applications. For instance, long exposures to low-light-level sources can allow
optimum exposure of a particular target, such as in astronomical use. If other
objects of interest within the sensor’s field of view appear during the long
integration, the pixels containing such images can be read out while the low-
light-level target continues to integrate on the array. At brighter target levels,
blooming is less likely to occur than in the CCD, due to the fact that charge
overload is confined to a single pixel.

Figure 5.49 Interline-transfer readout.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 207

CMOS. Like CIDs, CMOS devices are x-y addressable, allowing individual pixel
information to be read out without reading out the entire array. A disadvantage of
CMOS is high readout noise unless an amplifier is part of the configuration of
each pixel (that is, unless the device is an active-pixel sensor) which, as stated
above, reduces the device fill factor. Additional signal processing circuitry
providing functions such as thresholding, edge detection, and motion detection
can be incorporated into the CMOS chip, allowing CMOS to compete on a
performance level with scientific CCDs.
In addition to the common monolithic CMOS structure (detector and readout
together on one chip) CMOS devices are available in hybrid form (two chips
bump-bonded together). The latter structure allows a 100% fill factor in the
visible spectrum.

5.7.3.4 Comparison
When presented with a choice of a camera based on one of these three
technologies—CCD, CID, or CMOS—one must consider one’s application
relative to the technology and its cost. As in most radiometric system and
measurement problems, there are tradeoffs to be made. While not purporting to
offer the final word on technology selection, Table 5.26 presents a comparison of
some of the characteristics of these technologies. This comparison is by necessity
general; many specific tradeoffs can be made to enhance performance and reduce
cost, depending upon application.

5.7.4 Three-color CCDs

CCD imagers employing three-color arrays having red, green, and blue filters can
provide color imagery in the visible portion of the spectrum. Though this
technique is not as effective as film photography in rendering true color images,
it finds wide utility in digital photography, in which color images may be quickly
available. Typical array formats are large, though not nearly as large as standard
CCDs: Kodak, for example, has developed a 16.6-MP array camera, and larger
arrays feature 25 MP.12
A four-square “checkerboard” filter pattern is often used on the focal plane to
provide the red, green, and blue response channels. These schemes typically
include two out of four pixels green, and one each blue and red. The sensor
therefore gathers 50% of the green light over the area of the filter pattern, and
25% each of blue and red light. Substantial post processing is required to provide
all three colors at all pixel locations; arguably the most famous technique
employed for this purpose is the Bayer algorithm (U.S. patent No. 3971065)
which may also be utilized in cheaper CMOS-based cameras in addition to other
devices.
As one might expect, radiometry performed on data that have undergone
significant processing will be complex due to the number of error sources. At this
time, absolute radiometry performed using three-color CCDs is problematic at

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 208 Chapter 5

Table 5.26 CCD, CID, and CMOS comparisons.

CCD CID CMOS

Array fill Can reach 100% 80% or better Can range from
factor 30% to 100%
depending on
configuration
Sensitivity Highest High when Not as high as CCD
successive
frames are
integrated
Pixel-to- High Not as high as Can be high if
pixel CCD signal processing
uniformity is on chip
Blooming Can be a problem Excellent Needs improvement
control
Fabrication Requires specialized Uses standard Takes advantage of
processing CMOS existing
processing semiconductor
fab techniques
Advantages Well-established Can be used in low- Many on-chip
technology, well and high-light- signal-processing
characterized, level applications; features possible
many scientific excellent
uses blooming control

Problems Blooming control Noisy Readouts can be

noisy

best. The situation is changing, however, as digital cameras are being developed
in competition with film cameras for photogrammetric purposes. Because of the
application, these digital cameras would have to offer better radiometric
sensitivity than that obtained using digitized film images and would provide the
capability for self calibration.13
Alternatively, three separate CCDs may be used to provide color imaging,
with filters designed to pass radiation in the red, green, and blue portions of the
spectrum. This approach involves more hardware than the one discussed above,
but will become more cost effective as the price of this technology comes down.

5.7.5 Ultraviolet photon-detector arrays

Ultraviolet detector arrays typically utilize hybrid architectures. They commonly
employ arrays of photodiodes, though this is not always the case. Many arrays

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 209

utilize aluminum gallium nitride (AlGaN) as the detector material due to its
sensitivity in the UV and insensitivity at longer wavelengths. Flame detection,
astronomy, and undersea communications are some applications that benefit from
these devices. Current technical challenges include tailoring the cutoff
wavelengths of AlGaN p-i-n photodiodes by controlling alloy composition,
allowing AlxGa1–xN arrays to cut off between 227 nm and 365 nm.

5.7.6 Infrared photodetector arrays

Infrared photodetector arrays are utilized over all regions of the infrared
spectrum, providing detection in the near infrared (NIR), short-wave infrared
(SWIR), and long-wave infrared (LWIR). LWIR devices have been successfully
used for many years in sensitive military applications, as have SWIR-based
systems.
Indium gallium arsenide (InGaAs) arrays are used in the NIR to provide
alternatives to more traditional night-vision devices based on image intensifier
technology, as well as many other applications. InGaAs arrays may be comprised
of p-i-n photodiodes or avalanche photodiodes (APDs.) Varying the chemical
composition of the alloy (InxGa1–xAs) allows optimization at a desired
wavelength for maximum signal strength.
Detectors made of lead selenide (PbSe) and lead sulfide (PbS) operate in the
1 to 3-μm and 3 to 5-μm portions of the spectrum; these detectors typically
utilize a monolithic architecture. Another monolithic detector is platinum silicide
(PtSi), operating in the 3 to 5-μm region. PtSi arrays are highly uniform, even if
relatively low response compared to other detectors, and are often used to image
objects with high spatial variation.
Mercury cadmium telluride detectors (HgCdTe) are used in 3 to 5-μm and
particularly in 8 to 12-μm regions. They typically employ a hybrid architecture
and have high sensitivity and fill factor. They exhibit substantial nonuniformity
when tailored for the LWIR and require cryogenic cooling to 77 K, increasing the
size, weight, cost, and power consumption of devices. Indium antimonide (InSb)
arrays are also used in hybrid devices and for scientific and defense applications
in the SWIR and midwave infrared (MWIR).
Signal processing in the element (SPRITE) detectors, commonly used in
forward-looking infrared (FLIR) modules of British manufacture, utilize
HgCdTe, but do not have the many leads and preamps that predominate in
HgCdTe two-dimensional arrays. SPRITEs are typically linear arrays that allow
integration of the incoming signal by scanning along the array at the same rate as
the carrier drift. SPRITE-based FLIR imagers have been deployed on helicopters,
aircraft, and other platforms for tactical military uses in the 8 to 12-μm region.
Quantum-well infrared photodetectors (QWIPs) do not have the
nonuniformity problems associated with HgCdTe. Based on GaAs and InGaAs
technology, their manufacture takes advantage of established wafer productions
processes.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 210 Chapter 5

Table 5.27 Initial approach to uncooled imaging detector array development.14

Staring focal plane arrays based on thermal detection mechanisms

Focal plane response at video rates (30 Hz in the U.S.)
Noise-equivalent temperature difference (NETD) of 0.1 K
Resistive microbolometer staring arrays (Honeywell) and hybrid pyroelectric
detector staring arrays (Texas Instruments)
Use of modern integrated circuit electronic processing technology to enable
development of low-cost multielement arrays

5.7.7 Uncooled thermal imagers

Arrays of thermal detectors operating according to the detection mechanisms
discussed in Sec. 5.5 are finding increasing use in a wide variety of applications.
They are cheaper to manufacture than infrared photon arrays and do not require
cooling.
In the early to mid-1990s, formerly classified research into uncooled thermal
arrays became public. This research, sponsored by the U.S. Defense Advanced
Research Projects Agency (DARPA), the U.S. Army’s Night Vision and
Electronic Sensors Directorate (NVESD), and the U.S. Naval Air Warfare Center
(NAWC) provided an approach to the development of uncooled thermal imaging
detector array–based systems. Two U.S. companies were at the forefront of
efforts to develop these systems: Honeywell Corporation in Plymouth,
Minnesota, and Texas Instruments in Dallas, Texas. In addition, several
organizations worldwide, including GEC Marconi of Great Britain, participated
in developing uncooled infrared imaging detector arrays. Table 5.27 lists some
factors important to the initial approach to uncooled detector development.
Thermoelectric array development was also pursued; linear arrays based on
this technology have been used in several radiometric applications including in-
motion inspection of railroad train wheels.
While early 2D arrays provided 320 × 240 focal planes with individual pixel
elements of 50 μm in size and NETD of 0.1 K (f/1, 30 Hz), developments over
the past decade have significantly improved performance. Sensitivity of the
microbolometer arrays has been improved through improving pixel thermal
isolation and array fill factor. Microbolometer arrays based on vanadium oxide
detector material in a 640 × 480 format are now available, having NETDs of 50
mK or less (f/1, 30 Hz) with element sizes of approximately 25 μm. Packaging
improvements are also in process.
In fact, device-production strategy, effective use of existing fabrication
methods, and development of low-cost fabrication techniques all contribute to
how well a technology will be received in the marketplace. This reception, in
turn, drives research. More “exotic” expensive technology is not always better.
As in so many facets of radiometry, a technology user is wise to trade off costs,
benefits, and level of technical performance required before purchasing a system
or building a product. Table 5.28 lists some applications of uncooled thermal-
imaging devices.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 211

Table 5.28 Applications of uncooled thermal detector array-based systems.

Night-driving systems in automobiles

Roofing inspection
Leak detection from pipes and other fluid and gas delivery/storage
structures
Law-enforcement investigations of buildings, trucks, and other
possible concealment locations
Quality control along industrial process lines
Detection of “hot spots” in electrical wiring to determine possible
locations of malfunction
Aircraft inspection to determine weak points in structure based on
thermal information
Border-patrol surveillance

5.7.8 Summary
The above descriptions of array detectors, their uses, and some comparisons are
only an overview. Trading off parameters for a specific application is often
necessary, and there is no particular “right” answer to many detector selection
problems. A thermographer examining a housing structure for sources of heat
leakage can purchase an uncooled thermal instrument for far less cost than a
cooled device and yet receive adequate imagery for that task. An astronomer
observing a distant galaxy or low-light-level astronomical source may find a
cooled photodetector array to be his only reasonable option in the infrared, but
might he choose a CID over a CCD for observations in the visible?
Questions such as these are important for the student or professional wishing
to become adept at systems analysis and systems engineering, two of the most
important disciplines making use of the art of radiometry.

For Further Reading

D. F. Barbe, Ed., Charge-Coupled Devices, Springer Verlag, Berlin (1980).
J. D. E. Benyon and D. R. Lamb, Charge-Coupled Devices and Their
Applications, McGraw-Hill, London (1980).
L. Biberman and S. Nudelman, Photoelectronic Imaging Devices, Plenum, New
York (1970).
I. P. Csorba, Image Tubes, Howard W. Sams, Indianapolis (1985).
G. C. Holst and T. S. Lonheim, CMOS/CCD Sensors and Camera Systems, JCD
Publishing and SPIE Press, Bellingham, Washington (2007).
R. J. Keyes, Ed., Optical and Infrared Detectors, Springer-Verlag, Berlin (1980).
Photomultiplier Tubes: Principles and Applications, Philips Photonics (1994).
Sidney F. Ray, Scientific Photography and Applied Imaging, Focal Press, Oxford
(1999).
G. H. Rieke, Detection of Light: From the Ultraviolet to the Submillimeter,
Cambridge (1994).

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 212 Chapter 5

R. A. Smith, F. E. Jones, and R. P. Chasmar, The Detection and Measurement of

Infra-red Radiation, Oxford, London (1957).
A. H. Sommer, Photoemissive Materials, John Wiley & Sons, New York (1968).
Reprinted by Krieger (1980).
L. Stroebel, J. Compton, I. Current, and R. Zakia, Photographic Materials and
Processes, Second Edition, Focal Press, Boston (2000).
J. D. Vincent, Fundamentals of Infrared Detector Operation and Testing, John
Wiley & Sons, New York (1990).
A. T. Young, Photomultipliers, Their Causes and Cures, Chapter 1 in Methods of
Experimental Physics: Astrophysics, Vol. 12, N. Carleton Ed., Academic
Press, New York (1974).

Several books in the series Semiconductors and Semimetals, R. K. Willardson

and A. C. Beer, Eds., are relevant. Vols. 5 and 12 both treat infrared detectors,
and Vol. 11 deals exclusively with solar cells. Vol. 47 deals with uncooled
infrared imaging array systems.
Similarly, the series Advances in Electronics and Electron Physics is rich in
pertinent articles. Numerous volumes deal with imaging detector conference
proceedings, while others have significant feature articles. Most notable are Vols.
34 and 55.

A number of books have one or more chapters dealing with optical radiation
detectors. These include the following:
R. W. Boyd, Radiometry and the Detection of Optical Radiation, John Wiley &
Sons, New York (1983).
A. Chappell, Optoelectronics: Theory and Practice, McGraw-Hill, New York
(1978).
R. D. Hudson, Infrared System Engineering, John Wiley & Sons, New York
(1969).
D. Malacara, Physical Optics and Light Measurements, Academic Press, New
York (1988).

Several major handbooks offer articles and/or chapters dealing with optical
radiation detectors. These include the following:
J. S. Accetta, and D. L. Shumaker, Eds., Infrared and Electro-Optical Systems
Handbook, Vol. 3, SPIE Press, Bellingham, Washington (1993).
M. Bass, Ed., Handbook of Optics Vol. 1, McGraw-Hill, New York (1995). Parts
5, 6, and 7 contain Chapters 15 through 25, all pertinent to detectors and
detection.
W. L. Wolfe, Ed., Handbook of Military Infrared Technology, Chapters 11 and
12, Office of Naval Research, Washington D.C. (1965).
W. L. Wolfe and G. Zissis, Eds., The Infrared Handbook, Chapters 11–16, ERIM
and SPIE Press (1978).

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Detectors of Optical Radiation 213

Another rich source for detector information, and probably the best for
assessment of the current state of the art, is documented in the Proceedings of the
various conferences of SPIE. SPIE sponsors several major conventions per year,
each having one or more conferences on detectors. Compilations of the best
Proceedings papers, along with other critical papers, are gathered in SPIE’s
Milestones series http://spie.org/x649.xml.

References

1. F. Grum and R. J. Becherer, Radiometry, Vol. 1 in Optical Radiation

Measurements series, F. Grum, Ed., Academic Press, New York (1979).
2. D. M. Munroe, “Signal to noise ratio improvement,” in Handbook of
Measurement Science, Vol. 1, P. H. Sydenham, Ed., Wiley & Sons, New
York (1982).
3. C. D. Motchenbacher and J. A. Connelly, Low Noise Electronic System
Design, Wiley & Sons, New York (1993).
4. H. W. Ott, Noise Reduction Techniques in Electronic Systems, Wiley &
Sons, New York (1976).
5. E. L. Dereniak and G. D. Boreman, Infrared Detectors and Systems, Wiley &
Sons, New York (1996).
6. W. Budde, Physical Detectors of Optical Radiation, Vol. 4 in Optical
Radiation Measurements series, F. Grum and C. J. Bartleson, Eds., Academic
Press, New York (1983).
7. N. B. Stevens, “Radiation thermopiles,” in Infrared Detectors, Vol. 5 in
Semiconductors and Semimetals series, R. K. Willardson and A. C. Beer,
Eds., Academic Press, New York (1970).
8. E. Putley, “The pyroelectric detector,” in Infrared Detectors, Vol. 5 in
Semiconductors and Semimetals series, R. K. Willardson and A. C. Beer,
Eds., Academic Press, New York (1970).
9. E. L. Dereniak and D. G. Crowe, Optical Radiation Detectors, Wiley &
Sons, New York (1984).
10. R. W. Engstrom, RCA Photomultiplier Handbook, Burle Technologies,
Lancaster, Pennsylvania (1980).
11. Photomultiplier Tube: Principle to Application, H. Kume, Ed., Hamamatsu
Photonics K. K. (1994).
12. A. Rogalski, Opto-Electron. Rev. 12(2), p. 235 (2004).
13. F. Leberl, R. Perko, M. Gruber, and M. Ponticelli, “Novel concepts for aerial
digital cameras,” ISPRS Commission I Symposium, ISPRS, Denver (2002).
14. P. W. Kruse and D. D. Skatrud, Uncooled Infrared Imaging Arrays and
Systems, Academic Press, New York (1997).

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Chapter 6
Radiometric Instrumentation
Adapted by William L. Wolfe

6.1 Introduction
Radiometric instruments vary in what they are intended to measure, how they do
it, how complicated and expensive they are, how rugged, and in a number of
other ways. In this chapter, the simplest of radiometers is considered, the
components of radiometers are described, and spectral radiometers are covered.

6.2 Instrumentation Requirements

It surely comes as no surprise that different instruments have different
requirements. For instance, the required accuracy and repeatability vary greatly
with the application. A device meant for routine maintenance in the factory might
have a radiometric uncertainty of about 20%, yielding a temperature uncertainty
of a few degrees, whereas several different bureaus of standards have strived for
uncertainties of 0.01%.

6.2.1 Ideal radiometer

It is hard to describe one ideal radiometer, but, in general, they should all
measure the radiometric property accurately, precisely, and repeatedly. They
should be simple to use, with easily interpreted outputs; of course, the ideal
radiometer should be free of defects. Some need to be quite rugged for field use,
but all should be able to withstand the environment or environments to which
they are exposed. They should be sensitive and have a wide dynamic range, or
they should at least be sensitive enough for the task at hand and have sufficient
dynamic range for their use.

6.2.2 Specification sheet

In the fervent hope that you can obtain the ideal radiometer, the following
specification sheet is presented. It may not be complete for all applications, and
not all applications need all of these specifications, but it is a good start and a
good guide.
215

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 216 Chapter 6

Table 6.1 Radiometer specification sheet.

Radiometric properties Values

Noise-equivalent irradiance, NEI
Noise-equivalent temperature difference, NETD
Accuracy, precision
Linearity, dynamic range
Spatial characteristics
Entrance aperture size, A
Instantaneous field of view, W
Fixed or variable focus
Single detector, linear array, or 2D array
Spectral characteristics
Overall bandwidth range
Number of spectral channels
Spectral bandwidth for each channel
Continuous tuning or discrete wavelengths
Optical characteristics
Out-of-band rejection
Out-of-field rejection
Polarization sensitivity
Temporal characteristics
Spectral scan rate
Spatial scan rate
System response time and/or bandwidth
Operational requirements
Size, weight
Power, cooling requirements
Scanning requirements
Motor drives or manual
Environmental considerations
Interfaces
Mounting
Output presentation

6.2.3 Spectral considerations

These can be divided into three main categories: total radiation, broadband
radiation, and narrowband radiation. The first category involves detecting as
much of the total radiation as you can by using a spectrally flat detector. The
applications are things like measuring the earth’s radiation budget and radiation
temperatures.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Radiometric Instrumentation 217

Broadband applications include making background measurements in the

infrared spectral bands for the design of infrared cameras and similar devices—in
the 3 to 5-μm and 8 to 12-μm bands, for instance. Some measurements are made
in the visible spectrum over the photopic response of the eye; this is called
photometry. Others are radiometric measurements over the visible region, while
still others are trichromatic wherein the three colors are used to obtain color
renditions. In the ultraviolet, both UV-A and UV-B are measured to assess such
things as skin cancer susceptibility.
In these cases, the spectral passband is usually defined by one or more
absorption or interference filters, the detector, the window, or a combination of
all of them. It is important to measure the spectral response of radiometers of this
type.
Narrowband measurements are usually made with some kind of
spectrometer, or in this case, a spectroradiometer. Typically, a spectrometer is a
device that measures the spectral distribution of radiation, but not its absolute
magnitude, whereas a spectroradiometer does both. In some cases narrowband
filters can be used to provide spectral selection. These measurements are usually
made in situations where the radiation varies rapidly with wavelength or a certain
amount of species analysis is desired. They also apply when brightness or ratio
temperatures are to be determined. Since the bands are narrow, relatively intense
sources are necessary to provide appropriate signal strength.

6.2.4 Spatial considerations

There are two primary types of spatial coverage—large area and small area. It
seems fairly obvious that one designs or purchases a radiometer to cover the area
of interest. This may be a spot on an electric generator or it may be an average
over an entire house. The instantaneous field of view (IFOV) is determined by
the angular subtense of the detector, that is, the size of the detector divided by the
focal length, assuming the detector is at the focus. Its size is typically on the
order of milliradians (mrad) for many IR instruments. If the instrument is a
scanning radiometer, then the field of regard, the entire scanned field, should also
be considered. Some instruments have zoom optics, a variable focus, or
interchangeable optics. These allow both small and large fields. Some have linear
or 2D arrays. These have the resolution of the individual pixels but the field of
regard of the entire array. The array must be calibrated pixel by pixel and at
different levels of irradiance.

6.2.5 Temporal considerations

One of the primary considerations in radiometric measurements is the fidelity of
the data, and this relates directly to the signal-to-noise ratio (SNR) obtained by
the instrument for a given source. In turn this is essentially the reciprocal of the
relative uncertainty. So, to obtain a relative uncertainty of, say 5%, one needs an
SNR of 20, and 1% requires an SNR of 100. These requirements are much more
severe than simple detection and tracking.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 218 Chapter 6

Aliasing is an artifact in the data that can be caused by inadequate sampling.

To avoid these false signals, the bandwidth must be at least twice the highest
frequency in the signal. These frequency components can be generated by natural
fluctuations or by the use of a chopper, or perhaps turbulence.
If a chopper or some other device is used to generate square pulses as the
signal, then the bandwidth must be at least equal to the reciprocal of twice the
pulse width. If just a chopper is used, the chopper frequency must be three times
the highest frequency in the signal.
For an uncertainty of 5%, the settling time must be two times the response
time of the detector. Similarly, for 1%, five times, and for 0.1%, seven times.
Since all radiometers have a finite bandwidth, there is noise in that
bandwidth. Section 5.4 addresses noise in detail. To restate, white noise has a flat
power spectrum, and effective noise bandwidth B is defined by Eq. (5.15). For 1/f
noise, the expression in Eq. (5.14) is used.

6.2.6 Make or buy?

The decision to make or buy is one of availability of devices and of money. If
there is a radiometer available that meets your needs or comes close enough and
it fits within your budget, buy it. If the most useful unit needs too much
adaptation, don’t buy it. If you need to train students, don’t buy it. If you do buy
it and need to modify it, do so with as many commercial items as possible. For
example, do not make the monochromator of a spectroradiometer; there are units
for sale.

6.3 Radiometer Optics

6.3.1 Introduction
Chapter 2 introduced basic optical concepts relevant to radiometry as well as
some simple radiometer configurations. They will be reviewed here in
preparation for more detailed discussion of instrumentation. The main issues of
radiometer optics are the placing of stops and pupils, possible reimaging, use of a
field lens, and the reduction of scattered light.

6.3.2 Review of stops and pupils

There are two types of stops: aperture stop and field stop. The aperture stop
controls the amount of light that is accepted; the field stop controls the field of
view from which it is collected. Pupils are images of stops. The entrance pupil is
the aperture stop as projected into object space (by all the optical elements that
are between the aperture stop and object space). The exit pupil is the aperture
stop projected into image space. The entrance and exit windows are the
corresponding images of the field stop. These are illustrated in Fig. 6.1. Notice
that the second lens L2 is the aperture stop because its image in object space is
smaller than that of the first lens. The field stop is C, and its image out in object
space is the entrance window.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Radiometric Instrumentation 219

Figure 6.1 Stops and pupils.

6.3.3 The simplest radiometer: bare detector

A bare detector, shown in Fig. 6.2, can be used as a radiometer with its
associated electronics. Its angular field of view is almost 180 deg, but it is
probably not uniform over that range. Ideally it varies as the cosine of the angle,
but most detectors have some other dependency as a result of coatings, surface
imperfections, partial specular reflectivity, etc. If the detector has a polished
specular window, the angular variation may be calculable from the standard
Fresnel equations for reflection. Again, and it cannot be said enough, check it
out! The bare detector can be used for relatively imprecise quasi-hemispherical
measurements or for a distant source.

6.3.4 Added aperture

A slight improvement is the addition of an aperture, as shown in Fig. 6.3. With
almost no expense, just a case with a hole in it, one can define the field of view.
As a matter of possible interest, one can make a thermocouple by welding two
dissimilar metals, then punch a hole in a number ten tin can to make the business
end of a radiometer. The electronics could be one of those little multimeters you
can buy at the electronics store for about five dollars. The detector could be a
solar cell for visible spectrum use.

Figure 6.2 Bare detector.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 220 Chapter 6

Figure 6.3 One-aperture system.

In this case the detector is the aperture stop, and the hole is the field stop.
Since there are no lenses, there are no pupils. The flux on the detector from a
point source is given by

I s Ad
Φd = , (6.1)
d2

where Is is the source intensity, Ad is detector area, and d is the source–detector

distance. For an extended source (one that is larger than the image of the
detector), the flux on the detector is given by:

Ls Aap Ad
Φd = , (6.2)
d2

where Ls is the radiance of the source and Aap is the area of the aperture.
The field of view is “fuzzy”: that is, different parts of the detector view
somewhat different parts of the field of view because the detector is the field
stop. Note that if you make the detector a little smaller, the field of view is a little
smaller, so the outer portion of the detector views a different portion of the field
than the inner portions.
A reasonable improvement is obtained by the use of two apertures, as shown
in Fig. 6.4. The detector is no longer the field stop. For a point source, and A2
smaller than the active area of the detector, A2 is the aperture stop, A1 is the field
stop, and the flux on the detector is given by:

I s A2
Φd = , (6.3)
d2

and for an extended source,

Ls A1 A2
Φd = , (6.4)
d2

where d is the distance between the two apertures. The field is still a little fuzzy.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Radiometric Instrumentation 221

Figure 6.4 Two-aperture system.

One example of such a radiometer that uses a single detector and two
apertures and baffles but no lenses, is a pyrheliometer (fire + sun, measuring the
sun’s fire) used by the World Meteorological Organization and shown
schematically in Fig. 6.5. The angles made with the detector by A, B, and C
should be about 1, 2, and 4 deg, respectively.

6.3.5 Basic radiometer

Figure 6.6 shows a basic radiometer, just a lens and a detector. A mirror can be
used in place of a lens, and often is, but the diagrams get more complicated than
necessary. The lens is the aperture stop and the detector is the field stop. In this

Figure 6.5 Pyrheliometer.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 222 Chapter 6

Lens

Detector

Figure 6.6 Basic radiometer.

arrangement, the field is well (not fuzzily) defined, but the detector still “sees”
radiation coming at different angles from different parts of the field of view. This
is the case as long as the detector is the field stop. The advantages are that this is
a simple system—one lens, one detector, and associated electronics. The
disadvantage is that the detector averages over different parts of the field of view.
In this system the irradiance on the entrance pupil (the lens) from a point
source is given by

I s τatm
Eo = , (6.5)
d2

where Is is the source intensity, τatm is the atmospheric transmittance and d is the
distance from the source to the entrance pupil. The flux on the entrance pupil is
just the lens area Alens times this, and the flux on the detector is just the lens
transmittance τlens, times that:

I s τatm τlens Alens

Φd = . (6.6)
d2

The corresponding quantities for an extended source are

τatm τlens Ls Alens Ad

Φd = . (6.7)
f2

If the source is extended, the distance from the instrument to the source is
irrelevant; it is the focal length that counts.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Radiometric Instrumentation 223

Aperture
stop

Field
lens

Detector

Figure 6.7 Improved radiometer.

6.3.6 Improved radiometer

The improvement includes the addition of a field lens, as shown in Fig. 6.7. The
field lens images the aperture A onto the detector D, thereby making the flux on
the detector uniform. Although it is not yet obvious, this also helps control stray
light. The radiometry of this radiometer is not much different, but it must be
separated, as before, into the point-source case and the extended-source case.
For an extended source, where the detector is overfilled with the object, the
expression for the flux on the detector is

Φ d = τo Ls Ad Ω fsd , (6.8)

where το is the transmission of the optics, Ls is the radiance of the source, Ad is

the detector area, and Ωfsd is the solid angle the field stop subtends at the detector.

Afs
Ω fsd = , (6.9)
d2

where d is the distance from the detector to the field stop. Thus, one can also
write the power on the detector as

τo Ls Ad Afs
Φd = . (6.10)
d2

If the image of the objective underfills the detector, then

τo Ls Ao A fs
Φd = . (6.11)
f2

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 224 Chapter 6

If the image of the objective exactly fills the detector, either of Eqs. (6.10) or
(6.11) is accurate.

6.3.7 Other methods for defining the field of view

The two other principal methods for defining the field of view are the use of
diffusers and of integrating spheres. One may think of an integrating sphere as a
special diffuser. A simple diffuser is just a good scattering material. In the choice
of such a material, the grit and the spectral characteristics need to be considered.
Usually a good absorbing material, a diffuse black, is preferred. And remember,
just because it looks black does not mean it is black in the infrared or ultraviolet.
The black can be spread on the detector by an appropriate method, or it could be
used to reflect to the detector, in which case, it should be a white! There is no
substitute for measuring its spectral and angular responses. Such a diffuser is
usually a good depolarizer as well as a spatial averager.
The integrating sphere is coated with a highly reflecting diffuse material. It
has an entrance aperture (hole) and an exit aperture. The light enters the first and
exits the second to the detector. The multiple reflections inside the sphere are
what make the exiting light both uniform and unpolarized. Again, the spectral
characteristics need to be considered. There are several good references on the
calculations of the performance of integrating spheres. Try entering the search
criteria “integrating spheres” on the Internet. There are several good references
that give the throughput equations and offer a variety of types and sizes for sale.
Spheres provide an excellent cosine response, which is nice for
hemispherical measurements, but they are notoriously inefficient.

6.3.8 Viewing methods

It is usually very helpful to be able to see exactly what is being measured. Thus,
people have developed a variety of viewing schemes to accompany the
measurement instrumentation. One simple scheme is to put a telescope on top or
to the side of the radiometer. This scheme is simple, straightforward, and
removable, but it suffers from parallax: the two fields of view will not coincide at
all distances. Coaxial methods are preferable from this standpoint. One way to do
this with obscured systems is to use the folding mirror, as shown in Fig. 6.8.
Another popular way is to use a reflecting chopper. Then the fields are coaxial
and alternating in time. The persistence of the eye takes care of the interruptions.
Such a scheme is shown in Fig. 6.9. The mirror of Fig. 6.9 could be a large
chopper in the incoming beam, and then the eye and the reference can be in either
position. Other schemes may incorporate a fiber-optic pickoff someplace in the
system, but these do not usually work out very well, mostly because they are not
coaxial.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Radiometric Instrumentation 225

Detector

Mirror

Eye

Figure 6.8 Viewing with an obscured (Newtonian) optical system.

Detector
Reflective
chopper

Reference

Mirror

Eye
Figure 6.9 Viewing with a chopper.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 226 Chapter 6

6.3.9 Reference sources

Although all radiometers must be calibrated before and after use, it is also useful
to have an internal reference source. In radiometers with choppers, covered in the
next section, the measurement signal is the difference between the two detector
outputs: when the detector views the object and when the detector views the
reference source. The ideal situation is when there is no difference between the
signals. So the reference source should be designed to give approximately the
same flux on the detector as the object to be measured. This is a “null” reading.
The fundamental requirement is that the reference must be known. Sometimes it
is a blackened chopper, but this can be difficult since the spinning chopper may
vary in temperature and may have temperature gradients. Of course the chopper
temperature and emissivity must be known at the time of the measurement. A
good alternative is for the chopper to be highly reflective. Then it reflects an
internal reference to the detector. The internal reference can then be a well-
controlled source with high emissivity. Its temperature can be measured
continuously, and, since it is stationary, it is likely to be more stable. Most of
these internal references are blackbody simulators. They are usually conical in
shape with a relatively small aperture. Extensive research has been done on the
shapes and sizes to get the highest possible emissivity. The critical thing is that
the interior length should be much larger than the aperture. However, the ultimate
is not necessary for these internal references since they are secondary calibration
sources. Other shapes have included waffles, cylinders, spheres, and the like.
Basic design techniques have been published by Gouffé1 and DeVos2 and are
also described in The Infrared Handbook.3 Many commercial units can be found
by surfing or using the Photonics Spectra catalogs (on disk or in hard copy).
These are not usually designed to fit into equipment but to be laboratory
standards. You can surf for “blackbody simulators,” and find many commercial
devices.

6.3.10 Choppers
Most good radiometers use choppers, or radiation modulators. Of course there
are advantages and disadvantages. The ac signal from a chopped radiometer
provides a discriminant against a static background. It allows the use of drift-free
ac amplifiers. It avoids the low-frequency part of the 1/f noise region. It provides
the ability to use a synchronous detector, and it provides an a-b type of
comparison measurement. On the other hand, choppers reduce the available flux
by a little over 50%; they can be noisy both electrically and acoustically, they can
have reliability problems, and there can be phase noise if things are not exactly
right.
Most choppers are just spinning blades, driven by an electric motor. In this
form they do not use much energy. There are many types available on the market.
Some can be resonant devices that oscillate in and out of the beam, perhaps
tuning forks. These, too, are readily available from companies listed on the
Internet. Browse for “optical choppers”; both types are advertised.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Radiometric Instrumentation 227

6.3.11 Stray light

Stray light on the detector is a false signal. Therefore, stray light should be
eliminated or reduced to a level that makes it insignificant. There are basically
two types of stray-light problems: those that come from stray light that is in the
field of view of the radiometer (background light) and those that come from
outside the field of view. The latter can be the sun or another intense source that
is near the edge of the field of view. It can be greatly reduced by careful design
of stops and baffles. The proper use of baffles can attenuate these out-of-field
sources by many orders of magnitude.
The stray light from out-of-field sources can be reduced by the following
procedures. First, place the field stop or its image as close to the front of the
system as possible. Place the aperture stop or its image as close to the detector as
possible. Avoid obscured systems like Newtonians and Cassegrains (and this
may make a viewer more difficult). Make any baffles angled and black. There is
still an argument as to whether they should be specular or diffuse. The diffuse
baffles have better attenuation; the specular ones control the direction of the light
better. Obscured systems will provide their own scatter, but they can be used.
Figure 6.10 shows how to design a simple baffle. The baffle is designed to
protect against a source, like the sun, that is at an angle θ off axis. The first step

is to make the baffle long enough to achieve this. Then draw line ab at that angle

from the tip of the baffle to past the mirror tip. Then draw line cd parallel to
 
ab and to the end of the baffle tube. Then draw ef to intersect that line; that is
where the first vane is inserted, as shown. Then draw j to intersect at f. Place the
next vane where f intersects the line of baffle tips. This line is parallel to the tube
at the edge of the mirror. Now repeat the process for the other vanes.

Figure 6.10 Baffle design.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 228 Chapter 6

The final step is to analyze the system with one of several programs that can
calculate the amount of radiation on the detector from out-of-field sources as a
function of their angle out of field. And there is no substitute for a measurement
if the application warrants it. A number of stray-light-analysis systems are listed
on the Internet. Search for “stray light optical techniques.” The reduction of stray
light is an iterative technique that is shortened by experience. The analysis
programs tell you how you are doing.
The Lyot stop is a special kind of stop useful in controlling scatter. It was
used by Lyot to measure the solar corona. He imaged an interior disk inside the
system to the front. The image of the disk matched the solar disk, allowing him
to see the light outside the main disk of the sun—the corona and flares. The
image of the disk became the entrance pupil, an image of the aperture stop. This
is one application of using an entrance pupil to control stray light.

6.3.12 Summing up
The design and use of radiometric instruments is simple in principle, but difficult
in practice. It is a science of precision and care. One of the maxims is to think of
everything. One way to do this is to write the radiometric equation for the
radiometer:

V = R(λ, t , x, y , z , α, β, moon, RH, T , ...) . (6.12)

This means that the output voltage (or other electrical signal) is a function of
wavelength, time, spatial coordinates, angular coordinates, the phase of the
moon, the relative humidity, the temperature (of the radiometer and the source
and the background), and everything else you can think of. Then test the
radiometer against all of these variables. The devil is in the details—and so are
good results!

6.4 Spectral Instruments

6.4.1 Introduction
Spectral instruments include both those that make relative measurements and
those that make absolute measurements. The first type measures how much
radiation there is in each part of the spectrum, but only on a relative basis. They
are useful for chemical analysis, for instance. The absolute instruments, generally
called spectroradiometers, measure the spectrum, but also give information about
how many watts (or equivalent radiometric quantity) are at each wavelength.
They are calibrated and more difficult to use. This section is about various types
of spectroradiometers. Another way to view them is that they are radiometers
with several, or many, narrow spectral bands. The different types vary from each
other according to the way they obtain the spectrum. That can be with prisms,

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Radiometric Instrumentation 229

gratings, filters, by scattering, interference, and, in general, any phenomenon that

is spectrally dependent.
A typical spectroradiometer (or spectrometer) is shown in Fig. 6.11.
Foreoptics on the left bring the light to the entrance slit of a monochromator (the
device that generates the spectrum). The output of the monochromator goes to a
detector, and then to electronics, and then to some kind of display or recording.
The monochromator singles out a specific, usually narrow, wavelength band at a
time, and usually then sequences through several bands. With multiple detectors
all wavelengths can be sensed at the same time.
The monochromator consists of an entrance slit, collimating optics, a
disperser (usually a prism or grating that spreads the light of different
wavelengths), focusing optics, and an exit slit. Although the optics shown here
are lenses, most monochromators use mirrors to avoid problems with chromatic
aberration.
Spectroradiometers are characterized by how much light they collect, and
therefore their sensitivity, and by how narrow their spectral bands are. The
amount of light they collect is specified by their throughput (also called optical
extent, étendue, and AΩ product.) The narrowness of the spectral band is usually
specified by the resolving power, that is, the central wavelength divided by the
spectral width of the band. This is the same as the Q of an electronic filter. In
some instances the descriptor is the resolution, which is just the width of the
spectral band, usually specified as the full width at half maximum (FWHM). The
free spectral range is a description of the spectral region over which there is no
interference of other spectra, perhaps by overlapping of orders. Other descriptors
include the multiplex advantage and the throughput advantage. These latter two
are also called the Felgett and the Jacquinot advantages. The throughput
advantage relates to how much light you can get through the system. The
multiplex advantage refers to the measurement of many wavelengths of light at
one time, rather than sequentially. If there is a multiplex advantage, the
bandwidth can be narrower and the sensitivity therefore greater.

Figure 6.11 Typical spectrometer.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 230 Chapter 6

To get the optimum response from a spectrometer, one should make the
width of the entrance and exit slits the same. One should also just fill the entrance
slit with the foreoptics. Underfilling robs you of input; overfilling gives no
advantage and even increases stray light.

6.4.2 Prisms and gratings

Spectroradiometers based on prisms and gratings are probably the most common.
Each has its advantages and disadvantages. The resolution of a grating
spectrometer is constant across the spectrum and usually more than that of a
prism system. The throughput of the grating is also a little larger than that of the
prism. The resolving power of a grating is generally higher than that of a prism.
The grating equation is

mλ = sin θi + sin θd , (6.13)

where m is the order number, an integer, λ is the wavelength of the light, and
θi and θd are the angles of incidence and diffraction, respectively. For a given
wavelength and angle of incidence there are many maxima at various diffraction
angles that correspond to different values for m and θd. The 0th order includes all
wavelengths and is just regular reflection (or transmission). For a given incidence
angle, the first order, m = 1 for a wavelength λ, has the same diffraction angle as
the second order and half the wavelength. This is overlapping of orders and limits
the free spectral range.
The resolving power Q is given by

λ
Q= = mN , (6.14)
dλ

where m is the order number and N is the number of grating lines illuminated.
The throughput is the slit area times the projected area of the used portion of
the grating divided by the focal length of the optics:

Aslit Agrating cos θ

T = AΩ = . (6.15)
f2

The free spectral range of the prism is unlimited, while that of the grating is
limited by multiple orders, the different maxima corresponding to different
values of m. Gratings can be transmissive, reflective, and even concave to
incorporate some of the focusing properties of the system. The resolving power
of a grating is approximately constant across the spectrum, whereas the prism’s
Q varies. The resolving power of a prism is given by

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Radiometric Instrumentation 231

λ bdn
Q= = , (6.16)
dλ dλ

where b is the length of the base of the prism (actually the length of the beam
near the base) and dn/dλ is the change in refractive index of the prism with
respect to wavelength.
The throughput is dictated by the slit area and the focal ratio of the focusing
optics:

Aslit cos θAprism cos θ

T= . (6.17)
f2

Although it is a generalization, gratings are generally better: they have

greater throughput and resolving power than prisms.

6.4.3 Monochromator configurations

Since the monochromator is the heart of generating spectra and measuring them,
several variations are described here. They can be divided generally into prism
devices and grating devices, according to the type of disperser that is used. Prism
systems consist of several different mounts: Littrow, Wadswoth, and Amici.
There are more, but these are the main ones. The Littrow, as shown in Fig. 6.12,
has a mirror at an angle behind the prism. Thus the light enters the prism, exits to
the mirror, and returns through the prism, thereby creating a double pass for
twice the dispersion. The mirror can be adjusted for the degree of separation of
the input and exit beams.
The Wadsworth, as shown in Fig. 6.13, is a single pass. The light is refracted
at minimum angle and exits the mirror parallel to the input beam. The mirror is
an extension of the prism base.
The Amici, which is shown in Fig. 6.14, allows the light with the central (or
any chosen wavelength) ray undeviated and undisplaced (with careful design). It
requires the use of prisms of at least two different materials (with different
dispersions). More than three prisms can be used at the designer’s discretion.

Figure 6.12 Littrow mount.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 232 Chapter 6

Figure 6.13 Wadsworth prism mount.

Figure 6.14 The Amici prism.

Mirror systems have more variability, but generally there is an entrance slit,
an exit slit, one or two mirrors, and a grating. The Czerny-Turner, Fig. 6.15, was
developed by Marianus Czerny and Francis Turner, a onetime student of
Czerny’s and later a researcher at Eastman Kodak and professor at the College of
Optical Sciences, The University of Arizona. There are two versions; one might
be called straight and the other crossed, as in Fig. 6.16. The crossed version uses
the two paraboloids closer to on axis and should therefore have smaller
aberrations. Of course, if aberrations are not a problem, the paraboloids can be
replaced by spheres. In that case, and depending upon the speed of the optics,
spherical aberration rather than coma will be dominant. It is shown with a
reflective grating, but it is not hard to imagine how this setup can be used with a
transmissive grating. The essence of the design is that both the collimating optic
and the focusing optic are off-axis paraboloids. However, they are off axis in
opposite directions so that the comatic aberrations offset each other.
The Fastie Ebert mount is shown in Fig. 6.17. It has the advantage of using
only one mirror, although the mirror has to be larger than either of the mirrors of
the Czerny-Turner system. Again, the mirror can be either a sphere or a
paraboloid, and again the comatic aberrations tend to offset each other.
Obviously, this cannot be used with a transmissive grating.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Radiometric Instrumentation 233

Figure 6.15 Czerny-Turner (laid out) configuration.

Figure 6.16 Alternate (cross) Czerny-Turner configuration.

Figure 6.17 Fastie-Ebert configuration.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 234 Chapter 6

Figure 6.18 Seya-Namioka configuration.

The Seya-Namioka is shown in Fig. 6.18. It has the advantage of having no

mirrors at all. Thus, it is very compact and usually cheap. The focusing properties
are all in the convex reflective grating. However, this grating can be difficult to
make so that is has enough concavity and also enough lines.
These configurations are also discussed in the first of the online references
below.

6.4.4 Spectrometers
These devices come in several different varieties. They can be single or double
pass through the dispersers, like the Wadsworth (single) or the Littrow (double).
They can be single or double beam, as shown below. The double beams were
developed to get an automatic referencing system. In some older single-beam
systems, one had to wrap a cord around the screw that rotated the prism in just
the right way. This was to “program” the slit to account for the variation in
intensity from the source. One also had to run a calibration run, and then a
sample run, and manually do the ratioing. The double-beam system eliminates all
this. There are many ingenious variations on the several simple systems shown
here. One beam passes through the sample, whatever it may be. The other is the
reference beam and is just in air in the spectrometer. The output of the sample
beam is divided by the output of the reference beam to give the transmission of
the sample. There are several ways to do this, but the best way is to use the same
detector for both beams, perhaps with an appropriate chopper. This avoids the
obvious problem of detector matching, initially and repeatedly. The generic basic
two-beam instrument is shown in Fig. 6.19. The output coming from the exit slit
of the monochromator is collimated and divided into two beams by a divider. It
can be a (semitransparent) beamsplitter or a bladed chopper. The one beam goes
straight through; the other is diverted by two mirrors and then combined with the

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Radiometric Instrumentation 235

Figure 6.19 Generic double-beam spectrometer.

first beam by a beam combiner just like the beam divider—a synchronized
chopper or a plane-parallel plate. If the divider and combiner are choppers, the
electronics can just take the ratio of signals. If beamsplitters are used, some
technique must be used to “tag” each of the beams. The beams need not be
collimated, but could be relayed by lenses for compactness.
Other commercial instruments can be found on the internet by entering the
search term “spectrometers.” Many will show up with diagrams, prices, and
advertising!
There is one subtlety about double-beam systems that I (Wolfe) encountered
quite by accident. I was participating in a military study devoted to the detection
of land mines. Someone suggested that soil was transparent around 4 μm. I
scoffed, but I went home to prove my point. I took one inch of certified backyard
dirt into our double-beam spectrometer. Lo and behold, there was a transmission
peak at 4.3 μm! This could not be! So I took one inch of aluminum plate and
made the same measurement. Same result. I pondered this for a while. Clearly
aluminum is not transparent at 4.3 μm. The answer was in the reference beam.
The atmosphere is very absorbent at 4.3 μm due to carbon dioxide. What I was
measuring was the absorption in the reference beam that translated to apparent
transmission in the overall measurement. I have since seen this phenomenon in
other double-beam measurements. If there is extra transmission at 4.3 μm in the
scan of an optical material you have ordered, be skeptical. This illustrates at least
one principle of radiometry: do not simply believe whatever you measure; make
sure it makes sense. By the way, this problem can be obviated by filling the
reference tube with nitrogen.

6.4.5 Additive versus subtractive dispersion

Additive dispersion is just that: two prisms or two gratings operating in series to
increase the dispersion of the light. The Littrow-mounted prism gives additive
dispersion. Additive dispersion can be accomplished with multiple dispersers or
multiple passes or both. Subtractive dispersion combines the dispersions in

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 236 Chapter 6

opposite directions, thereby combining the light and eliminating or canceling the
spread of the spectrum.

6.4.6 Arrays
Monochromators can also be used with detector arrays. The arrangement
includes the foreoptics described above, an entrance slit, collimating optics, and
the array, which functions as a set of exit slits. If the array is linear, the system
can be viewed simply as a monochromator with many exit slits, each one sensing
a different wavelength. This system obviously has the multiplex advantage, but
does not have the throughput advantage. The detector elements can usually be
sized correctly for the proper operation, but apertures can also be used with a
concomitant loss of signal.

6.4.7 Multiple slit systems

Several different schemes have been developed to give a multiplex advantage by
using several entrance slits and several exit slits at the same time. Perhaps the
first of these was invented by Marcel Golay;4 perhaps the most popular, the
Hadamard transform, was developed by Harwit.5 Although these are interesting
designs that incorporate the multiplex advantage, they have not supplanted the
other prism, grating, and interferometric spectrometers.

6.4.8 Filters
Filters come in a variety of types. They can be based on absorption, interference,
and even scattering. Absorption and interference filters are the main candidates in
spectroradiometry. Almost any kind of bandpass can be generated by a proper
thin-film design. They can be narrowband, broadband, angle tolerant, multiple
bandpass, etc. Then they can be put in a filter wheel to obtain a spectrum of a
sort, although not a continuous spectrum. They have good throughput but do not
have as good resolving power as prisms and gratings. They can come in
segmented wheels or circular or linear variable filters. These latter devices are
interference filters with layer thickness variation around the circumference or
along the length. They therefore have a spectral band that varies with either angle
or position.
Their characteristics vary by design. The circumference of a circular variable
filter (CVF) is given by

Δλ
C= Do , (6.18)
dλ

where Δλ is the spectral range, dλ is the resolution and Do is the diameter of the
aperture stop, where it is placed. The diameter of the CVF is this value divided
by π. A representative filter has the following characteristics: the spectral range
is 2.5 to 14.1 μm, Q is 67, FWHM is 1.5%, and resolution varies from 0.03 μm at

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Radiometric Instrumentation 237

2.5 μm to 0.2 μm at 14.1 μm. The CVF is divided into three segments that cover
from 2.5 to 4.3, 4.3 to 7.7, and 7.7 to 14.1 μm. It can be operated at the focal
plane, as the manufacturer recommends, but it should be operated at a field stop
rather than the focal plane. Since it would then be in converging light, the
resolution will be affected. Of course other varieties are available, semicircles for
instance. Typically the spectral range is an octave, the resolving power is about
50, and the diameters are about 4 cm. Specific devices can be found on the
Internet by searching for “optical filters” or “variable optical filters.” If you leave
out “optical,” all sorts of electrical filters will appear in the search results.
Optical filters are optically and mechanically simple, cheap, rugged, and easy
to automate, but they have poor stray-light suppression and limited resolving
power.

6.4.9 Interferometers
The main interferometers used for spectral analysis are the Michelson and the
Fabry-Perot, or as some insist, the Perot-Fabry. The former is a two-beam
instrument, the latter a multiple-beam device. The multiple-beam system has
greater resolving power, but smaller free spectral range.

6.4.10 Fourier transform infrared

The Fourier transform infrared spectrometer is a Michelson, or actually a
Twyman-Green, as shown in Fig. 6.20. (The Twyman Green is a Michelson with
collimated light). The light from the source is collimated by the first lens. The
beam is then divided into two by the beamsplitter. One beam goes up, the other
continues to the right. The mirrors then reverse the direction of both beams and
they are combined by the same beamsplitter, which is now acting as a combiner.
The beams interfere, and the intensity of the interfering radiation is sensed by the

Figure 6.20 Twyman-Green interferometer.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 238 Chapter 6

Detector, whose operation can be understood by considering first a single

monochromatic input, then, successively, added inputs of different wavelengths.
When the interferometer is set so that the two arms have exactly the same optical
length, light from the two beams will interfere constructively. Then, as one
mirror is moved, the beams gradually go out of phase until they reach destructive
interference, and as the mirror is further moved, they gradually reach
constructive interference again. So there will be a sinusoidal output based on the
motion of the scanning mirror. If a beam of somewhat different wavelength is
inserted it will do the same, but for somewhat different positions of the mirror. A
third beam, a fourth, and so on will add to the complexity, but each will
contribute a sinusoidal component to the output signal. At any position of the
scanning mirror, the output is the sum of all waves, each at a different point in
phase. The sum total of the scan is called an interferogram—the interference
pattern of whatever the input beam was. The Fourier transform of the
interferogram is the spectrum of the input beam. Because this device is used
mostly in the infrared, it is called the Fourier transform infrared spectrometer. It
has good throughput and the so-called multiplex advantage; i.e., it senses all
wavelengths of light at the same time (as opposed to prism and grating
instruments that see one small sample of the spectrum at a time).
The resolving power is given by

λ σ σ 5000
Q= = = = . (6.19)
d λ d σ 2δ λδ

The expression σ/dσ is the wavenumber equivalent. The final expression

incorporates the fact that the wavelength is usually given in μm while the
wavenumber is in cm–1.
These systems are complex, susceptible to vibration, have a limited spectral
range, and require considerable computing power to perform the transform
operation, but they have the multiplex advantage and good throughput. They are
limited to the infrared, but they are very useful for what they do.

6.4.11 Fabry-Perot
This interferometer is generally used for high-resolution spectroscopy. It is
essentially a pair of plane-parallel plates, as shown in Fig. 6.21. The light from
the source is collimated by the first lens, and passes through the first plate after a
little lost reflection. Light reaches the second plate and is reflected back to the
first, which reflects it back to the second, which reflects it back to the first. This
interferometer does have high resolution because of the multiple-beam
interference. The basic equation for the transmission of the Fabry-Perot is

τo
τ(λ) = , (6.20)
4ρ 2
1+ sin φ
(1 − ρ) 2

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Radiometric Instrumentation 239

and
(φ1 + φ2 )
φ = 2πnd σ cos θi − , (6.21)
2

where τo is the transmission of the plates, ρ is their reflectivity, n is the refractive

index of the medium between the plates, d is the plate separation, θi is the angle
of inclination of the incident beam, which is usually 0, and the φi are the phase
changes on reflection from the plates. It can be shown that the maximum
transmission is when the reflectance is highest (in the limit, 1), which is a rather
strange result. The resolving power is given by

σ λ ρ
Q= = = mπ , (6.22)
dσ dλ 1− ρ

the throughput is given by

T = τo AΩ , (6.23)

and the free spectral range is given by

1
Δσ = . (6.24)
2d

The Fabry-Perot has a relatively poor throughput because of the requirement for
collimation between the plates. But it is better than a prism or grating.

Final note: Acousto-optical tunable filter (AOTF) devices for spectroscopy of all
kinds are relatively new. “Classical” spectrometers employing prisms and
gratings have seen many improvements. There are pros and cons for both. They
are a tool that should be in the radiometrist’s kit and are described further in
Wolfe and in Chang, below.

Figure 6.21 Fabry-Perot Interferometer.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 240 Chapter 6

For Further Reading

P. Baumeister, Optical Coating Technology, SPIE Press, Bellingham,
Washington (2004). Good section on filters.
I. C. Chang, “Acousto-optic devices and applications,” Chapter 12 in Handbook
of Optics, Vol. II, M. Bass, Ed., McGraw-Hill, New York (1999).
D. S. Goodman, “Basic optical instruments,” Chapter 4 in Geometrical and
Instrumental Optics, D. Malacara, Ed., Academic Press, New York (1988).
F. Grum and R. Becherer, Radiometry, Sec. 7.3, Vol. 1 in Optical Radiation
Measurements series, F. Grum, Ed., Academic Press, New York (1979).
G. R. Harrison, R. C. Lord, and J. R. Loofbourow, Practical Spectroscopy,
Prentice-Hall, New York (1948).
R. Kingslake, “Dispersing prisms,” Chapter 1 in Applied Optics and Optical
Engineering, Vol. 5, R. Kingslake, Ed., Academic Press, New York (1969).
R. Meltzer, “Spectrographs and monochromators,” Chapter 3 in Applied Optics
and Optical Engineering, Vol. 5, R. Kingslake, Ed., Academic Press, New
York (1969).
W. J. Potts, and A. L. Smith, “Optimizing the operating parameters of infrared
spectrometers,” Appl. Opt. 6, 257 (1967).
D. Richardson, “Diffraction gratings,” Chapter 2 in Applied Optics and Optical
Engineering, Vol. 5, R. Kingslake, Ed., Academic Press, New York (1969).
R. A. Sawyer, Experimental Spectroscopy, Dover, New York (1963).
R. Willey, Practical Design and Production of Optical Thin Films, Marcel
Dekker, New York (2002). Good reference on filters.
W. L. Wolfe, Introduction to Imaging Spectrometers, SPIE Press, Bellingham,
Washington (1997).
J. Workman and A. W. Springsteen, Applied Spectroscopy, Academic Press, New
York (1998).

References

1. A. Gouffé, “Corrections d’ouvertures des corp-nois artificels compte tenu des

diffusions multiples internes,” Revue d’ optique 24(1) (1945).
2. J. C. DeVos, “Evaluation of the quality of a blackbody,” Physica 20, p. 669
(1945).
3. A. J. LaRocca, “Artificial sources,” Chapter 2 in The Infrared Handbook, W.
L. Wolfe and G. J. Zissis, Eds., U.S. Government, Washington, D.C. (1978).
4. M. Golay, “Multislit spectroscopy,” J. Opt. Soc. Amer. 39, pp. 437–444
(1949).
5. M. Harwit and N. Sloane, Hadamard Transform Optics, Academic Press,
New York (1979).

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Chapter 7
Radiometric Measurement and
Calibration
7.1 Introduction
This chapter deals with the numerous measurements for which we use
radiometers and spectroradiometers. This is the true meaning of the word
radiometry, the measurement of radiant energy. First, we describe the types of
measurements made. Next is a discussion on errors, their sources, and treatment.
The generalized measurement equation and several derived range equations
follow. An introduction to the philosophy of radiometric calibration is presented
next, and a discussion of calibration configurations completes the chapter.

7.2 Measurement Types

Radiometric measurements may be classified into four general types. They are:
(1) detector and radiometer characterization, (2) optical radiation source
measurement, (3) material properties measurement, and (4) temperature
measurement. A fifth measurement type is calibration, which will be discussed
later. Table 7.1 subdivides the first four categories into what is not an exhaustive
list.

7.3 Errors in Measurements, Effects of Noise, and Signal-to-

Noise Ratio in Measurements
A measurement of any kind is incomplete unless accompanied with an estimate
of the uncertainty associated with that measurement. The term error implies a
difference or deviation from a “true” value, while the term uncertainty means an
estimate characterizing the range of values within which the true value of the
measured quantity lies, including all sources of error. Errors come in two primary
flavors, random and systematic, as we will see through consideration of Fig. 7.1.
Systematic (type B) errors are readings that vary in a predictable, hopefully
detectable, way. Systematic errors are repeatable and consistent, with a fixed
bias, the difference between the measured value of x (mean of N measurements)
and the true value of x.

241

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 242 Chapter 7

Table 7.1 Types of radiometric measurements.

Detector and radiometer characterization

Relative spectral responsivity
Absolute spectral responsivity
Noise properties
Detective properties (NEP, D*, D**)
Field of view, out-of-field response
Linearity
Frequency response
Polarization response
Wavelength characterization (for spectroradiometers)
Passband characterization
Measurement of optical radiation sources
Active (self-radiating) and passive (reflective) sources
Source intensity
Source radiance or brightness
Source power or total flux
Light (photometry)
Ultraviolet and infrared sources
Source temperature
Collimated (laser) sources
Measurement of radiometric properties of materials
Specular reflectance
Diffuse reflectance
Transmittance
Scattering properties, BRDF, and BTDF
Indirect measurements of absorptance and emittance
Direct measurement of emittance
Color
Measurement of temperature
Radiation temperature using entire spectrum
Brightness temperature using one wavelength
Ratio temperature using two wavelengths
Color temperature using chromaticity
Temperature using multiple wavelengths
Distribution and correlated color temperature

Examples of a systematic error are an incorrect setting of a calibration

potentiometer of a voltmeter, or a slipped or improperly installed temperature
dial on a blackbody radiation simulator. Systematic errors are not revealed by
repeated measurements. If detected, they may often be corrected or at least taken
into account. The term accuracy is often applied to systematic errors, implying
small systematic errors. This term should be replaced by inaccuracy, inasmuch as

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Radiometric Measurement and Calibration 243

Figure 7.1 Systematic and random errors.

a voltmeter with only 1% accuracy is far less desirable than one with 99%
accuracy. The systematic error in a measurement is closely related to calibration
of the apparatus used to conduct the measurement.
Random (type A) errors are those that vary in an unpredictable manner when
the same quantity is repeatedly measured under identical conditions. They are
revealed only by multiple measurements. Precision is a term often associated
with random errors; a measurement is considered precise if it is repeatable. We
may employ several methodologies to reduce random errors, enhancing the
measurement precision. First, we can use a finer scale division, i.e., more bits, to
reduce granularity in the measurement. Second, we can reduce some of the
inherent noise in the measurement process by filtering, cooling, shielding, etc.
The most important tool to reduce random errors is statistical analysis. We take
multiple readings and perform analysis to reduce the effects of “noise” and
increase our confidence in the measurement.
To understand the total uncertainty in a measurement, we must consider both
the systematic and the random error components. Figure 7.2 shows a set of
“measurements” with various combinations of accuracy and precision. The
average (x and y) of the third pattern lies very close to the center of the target.
Errors can be specified as absolute, the magnitude of the error in the appropriate
engineering units, or as a relative or fractional error, usually in percent.

HIGH PRECISION HIGH PRECISION LOW PRECISION LOW PRECISION

HIGH ACCURACY LOW ACCURACY HIGH ACCURACY LOW ACCURACY

Figure 7.2 Precision and accuracy possibilities.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 244 Chapter 7

Table 7.2 Signal-to-noise ratios and corresponding uncertainties.

SNR (single measurement) Uncertainty

1 1 (100%)
10 0.1 (10%)
100 0.01 (1%)
1000 0.001 (0.1%)

Errors may also be categorized as multiplicative or additive. A multiplicative

error, often referred to as a scale or gain error, is proportional to an instrument
reading. An additive error, often referred to as an offset, yields an absolute
uncertainty that is independent of reading. The error on a typical instrument
specification might read, “±0.5% of reading ±0.2% full scale.” These are the
multiplicative and additive errors, respectively. Note that the symbol ± is
redundant; a 0.5% deviation can clearly go in either direction.
The fundamental error limit in a measurement is random noise. If all other
error sources were reduced to zero, the remaining noise would be Gaussian, most
likely Johnson or shot noise associated with our detector. We certainly desire to
have the limit to uncertainty in a system dependent upon noise, as this indicates
that the systematic errors are understood and under control. In such a system
where noise is the predominant error term, the measurement uncertainty is
inversely related to the signal-to-noise ratio (SNR), or

1
Measurement uncertainty = . (7.1)
SNR

Thus, an SNR of 10 implies a 10% uncertainty (1σ) in a single measurement,

i.e., 1. Table 7.2 shows uncertainties for several values of SNR.
We wish to take repeated measurements to enhance the SNR. With multiple
measurements, the signal is additive, and the average of the noise tends towards
zero. The resultant SNR is proportional to the square root of the number of
independent measurements, (SNR ∝ N ). We then statistically analyze the data
and make the following assumptions:

(1) The data possess a “normal” (Gaussian) distribution of random errors.

(2) The individual measurements are statistically independent (uncorrelated).
(3) The quantity being measured is stationary.

We then apply the analytical tools listed in Table 7.3.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Radiometric Measurement and Calibration 245

Table 7.3 Mathematical tools for dataset analysis.

Tool Formula

Mean
x i
x = i
, N = number of
N
measurements
( x − x)
2
i
Variance (sample)
σ 2= i

N −1
2
σ= σ
Standard deviation
σ
Standard deviation of mean σx =
N

We further explore the three assumptions listed above.

(1) Assurance of a normal distribution is approached by taking a large
amount of data (central limit theorem). Verification of the normal
distribution can be accomplished using a Chi-square (χ2) test.
(2) The measurements must be independent (uncorrelated). In reality, it is
impossible to achieve complete independence because of the exponential
nature of signal changes with time. Spacing the data sampling interval by
one time constant improves the SNR by (N/2)1/2. Note that there is no
prohibition to faster sampling, but the correlation of such measurements
reduces the apparent gain in SNR. The maximum SNR improvement is
proportional to the square root of the observation time and is independent
of the sampling rate.1
(3) If there is a low frequency drift (1/f noise) in the data, normal statistical
analysis may be compromised and further analysis is mandatory. You
may be able to low-pass filter the data to reduce the random noise, then
use regression analysis (linear, exponential, power, etc.) to fit a curve to
the remaining low-frequency (drift) component. A more sophisticated
method is the analysis of Allan variances, developed for drift assessment
of atomic clocks. Data sets are analyzed over different time frames. The
Allan variance is given by:

1 2
σ 2y ( τo ) = 
N −1
( yk +1 − yk ) . (7.2)
2( N − 1) k =1

Compare with the classical variance

1
( )
2
σ 2( N ) = 
N
yk − y . (7.3)
N −1 k =1

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 246 Chapter 7

If σ2 ( N ) / σ2y (τo ) ≤ (1 + 1 / N ) , then white (classical) noise predominates

over drift (1/f), and we can increase the sampling window τo and therefore the
number of samples N. Several examples of this situation are shown in Fig. 7.3.
The bottom curve is the white-noise case, where the SNR improves as
1 / N . The other three curves show the effects of varying drift rates. As the drift
rate climbs (i.e., progressively higher 1/f noise than white noise) the sampling or
integration time shortens, resulting in a lower SNR.
An important contribution to random error (precision) is the granularity of
the instrumentation. If the markings on the analog meter movement are too
widely spaced, interpolation between divisions may be difficult. Reading a ruler
or a micrometer may be difficult for the same reason. Most electrical
measurements are now taken with digital instrumentation, where quantization
takes place via an analog-to-digital (A/D) converter. Quantization noise was
defined in Eq. (5.31) and is

SIGNALmax
LSB = , (7.4)
2n

where LSB is the least significant bit and SIGNALmax is the full-scale reading.
If this quantization error is significant in a measurement, the use of more bits
(larger n) is indicated. If the presence of sharp quantization levels is apparent in
your data and you cannot get more bits, you can artificially smooth the data by
adding noise. This is called dithering, used to “smear” and mask the appearance
of discrete levels.

Figure 7.3 Drift effects on data averaging for a number of cases.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Radiometric Measurement and Calibration 247

Table 7.4 Chauvenet’s criteria for data rejection.

Data point deviation

Number of readings Standard deviation
4 1.54
6 1.73
8 1.86
10 1.96
15 2.13
20 2.24
25 2.33
50 2.57
100 2.81
200 3.02
500 3.29
1000 3.48

One is often tempted to reject data that appears to be “out of line.” The
general recommendation is DON’T. The data has warts, leave them alone. If you
really must, there are reasonable guidelines for data rejection. Don’t make the
mistake of expunging data from your data set, just set the outliers aside. After all,
there might be something really interesting there. The most popular tool is the
Chauvenet criteria for rejection of outliers. Here we reject a data point if the
probability of observation of the suspect point is less than ½N where N is number
of observations. Then we can recompute mean and standard deviation. The
criteria are shown in Table 7.4 above.
Application of Chauvenet’s criteria affects standard deviation more than
mean. It is interesting to test the dataset with and without data rejection to see if
it really matters! Usually it doesn’t. A final note: even if tempted, do not apply
this technique more than once. If you do, the data will probably be skewed. If one
keeps rejecting data, eventually one will have only a single data point remaining.
To minimize systematic errors, calibrate frequently with suitable apparatus,
as noted above. Table 7.5 lists several sources of systematic error.
The evaluation of systematic error is, in practice, a judgment call, based upon
an investigator’s familiarity with previous data, instrument behavior,
manufacturers’ specifications, handbook reference data, and calibration. Current
recommended practice is to estimate systematic errors at the 1σ (68%
confidence) level.
Systematic errors in optical radiation measurements tend to be large because
the quantities involved are a function of everything in the world; the following
factors are the most common:
(1) Wavelength (broadband or monochromatic)
(2) Power or energy level
(3) Linearity

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 248 Chapter 7

Table 7.5 Sources of systematic error.

Environmental factors (temperature, humidity,

pressure, electromagnetic interference, dust and
dirt, power line fluctuation)
Dynamic (response time and slew rate) errors
Offset error (additive)
Gain or scale error (multiplicative)
Hysteresis
Nonlinearity
Faulty procedures
Personal bias
Calibration frequency
Calibration standards

(4) Modulation frequency or pulse characteristics

(5) Position, direction
(6) Polarization, diffraction, coherence, “phase of the moon.”

There are two secondary error categories: illegitimate errors and model
validity. Illegitimate errors include blunders, mistakes, computational errors
(roundoff, etc.), and chaotic errors. Model validity has to do with our
preconceived ideas about the results of a measurement, which influences the way
we conduct the measurement. While these errors may be significant or even
overwhelming in a particular situation, none are essential contributors to the
limits of radiometry and can be eliminated with proper care.
When all of the various error sources have been identified, they must be
combined into an overall uncertainty estimate. Errors propagate according to
well-known formulae as shown in the following tables. Table 7.6 presents the
formulas for single-valued functions.
For combinations of functions, the worst-case ultraconservative treatment is
to directly add the errors. This method would be correct if all of the errors were
correlated. Table 7.7 presents the formulas for functions in which independent
variables are combined.

Table 7.6 Error expressions for single-valued functions.

Constant Reciprocal Power Exponential Logarithmic

(q = Ax) (q = 1/x) (q = xn) q = e ax q = ln(ax)
σq σx σq σx σq σx σq σq 1 σx
= = =n = aσ x =
q x q x q x q q a ln ( ax ) x

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Radiometric Measurement and Calibration 249

Table 7.7 Error expressions for functions of several variables.

Addition and subtraction (q = x + y – z) σq = σ x + σ y + σ z

σq σx σ y σz
= + +
q x y z
Multiplication and division (q = xy/z)
σq σx
Power (q = xn) =n
q x

Most often, the errors are independent or uncorrelated. Under these

conditions, we use the root sum square (RSS) technique shown in Table 7.8.
When all of the random and systematic uncertainties have been assessed, we
combine them into a single uncertainty estimate using the propagation formulas
above. The combined standard uncertainty is the standard deviation of the
measurement. We assess the random and systematic uncertainties independently
at the 1σ level, then add the standard deviations in quadrature (that is, we use
RSS) to get the combined 1σ uncertainty:

σc = σ 2r + σ2s . (7.5)

The 1σ level implies a confidence level that is only 68%. To give more
confidence in the measurements, we multiply sigma by a coverage factor k. A k
of 2 is 2σ, where σ is the standard deviation of the measurement. Use a suggested
confidence factor k of 2, which gives a confidence of approximately 95%.
A recommended format for reporting results shows the measured value with
the standard (1σ) uncertainty and a coverage factor k as follows:

(45.26 ± 0.03) W/cm 2 (2σ) or (k = 2) . (7.6)

In standards work, authors frequently split the random and systematic uncertainty
components and report both. For further details, consult Taylor.2

Table 7.8 Root-sum-square formulas.

Addition and subtraction (q = x + y – z)

+ ( σ y ) + ( σz )
2
(σx )
2 2
σq =

2 2 2
σq  σ   σy   σ 
Multiplication and division (q = xy/z) =  x  +  + z 
q  x   y   z 

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 250 Chapter 7

7.4 Measurement and Range Equations

The general form of the measurement equation relates the observed SIGNAL
from a radiometric instrument or detector having a responsivity ℜ to an input
radiance L. In differential form, it is

dSIGNAL = ℜ(λ) world Lλ (λ) world dAd Ωd λ , (7.7)

which is not particularly useful. We must integrate over area, solid angle, and
wavelength to get the integral form:

SIGNAL =  ℜ(λ ) Lλ (λ )dAd Ωd λ . (7.8)

In both equations above,

SIGNAL = the output radiometric signal,

Lλ(λ) = the input spectral radiance, and
ℜ(λ) = is the radiometric system’s power responsivity.

The “world” (measurement environment) consists of:

θ,φ angular dependences

x,y position dependences
λ wavelength
t,ν time, frequency
s,p polarization components
diffraction
nonlinearities
“phase of the moon”

These equations are valid only for incoherent radiation; another layer of
complexity is added when speckle and interference effects are added.
There are many alternate forms of the measurement equation. For a single
detector with spectral responsivity ℜ(λ), it simplifies to:

SIGNAL =  ℜ ( λ )  [ Φ λ (λ) d λ ] , (7.9)

where Φλ(λ) = spectral radiant power incident on the detector and ℜ(λ) = the
radiometer spectral power responsivity.
For a detector with spectral responsivity ℜ(λ) and a narrow-band source or
filter:

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Radiometric Measurement and Calibration 251

SIGNAL = ℜ ( λ ) Φ λ (λ ) τ(λ ) Δλ , (7.10)

where τ(λ) = spectral transmittance of the filter and Δλ = bandwidth of the source
or filter.
For a laser line where the spectral linewidth is so small that the responsivity
of the detector/radiometer can be considered constant, then

SIGNAL = ℜΦ . (7.11)

Different radiometric configurations have different measurement equations.

For example, the signal from a distant small object that underfills the radiometer
field of view is described as

1
SIGNAL =
d2  ℜ ( λ ) I λ (λ) dAd λ , (7.12)

where

SIGNAL = the radiometric signal,

Iλ(λ) = the spectral radiant intensity of the source,
d = the distance to the source, and
ℜ(λ) = the radiometer spectral power responsivity.

Many other forms can be developed depending upon the chosen measurement
configuration.
Similar to the measurement equation is the range equation, which gives the
distance at which a source will generate a specified SNR. It is useful to visualize
which system parameters are important in particular applications. A general form
is provided by Hudson:3

1 1
1  π Do τo  2
1  1  2
R = ( Is τa ) 2
  ( D *) 2
  (7.13)
 4 ( f #) Ω   SNR B 
TERM 1 2 3 4

where terms 1 through 4 are defined as:

Term 1: Target parameters: intensity Is and atmospheric transmission τa

Term 2: Optical system parameters: f/#, optics diameter Do, FOV Ω, and
transmission τo
Term 3: Detector parameter: D*
Term 4: Signal processing parameters: required SNR and noise bandwidth B.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 252 Chapter 7

This range equation can take many specific forms to determine SNR or range
for various source/radiometer spatial and temporal configurations. Note, also,
that the term intensity implies a point source.
There are several alternate means of assessing the performance of a
radiometric system based on what are called “noise equivalent” quantities, whose
input values produce an SNR of 1 (signal equals noise). Noise-equivalent power
(NEP), for example, was defined in Chapter 5 as

in i
NEP = = Φ n ;or
ℜ is
v v
NEP = n = Φ n ;or (7.14)
ℜ vs
Ad B
NEP = .
D*

It is usually applied to detectors, but may also be useful for systems designed
to measure radiant power.
We can also define a noise-equivalent irradiance or flux density (NEI or
NEFD), which is frequently used to characterize systems for detection of distant
small sources:

E v NEP
NEI = =E n = . (7.15)
SNR vs Ad

Note that the term NEI does not reflect our symbol for irradiance E.
For an unresolved target that underfills the FOV, Hudson3 gives the noise-
equivalent irradiance (W/m2) as

4( f #)Ω1/ 2 B1/2
NEI = , (7.16)
πDo D * τo

which bears a striking resemblance to the range equation.

The noise-equivalent temperature difference (NETD or NEΔT) is used to
predict the performance of thermal radiometers and infrared cameras. It
characterizes a system by its ability to distinguish a small temperature difference
between a resolved target (underfills FOV) and the background. For a single
small detector Holst gives:4

4( f #) B
NETD = λ2
, (7.17)
∂M (λ, TB )
Ad  τoptics (λ) D * (λ ) d λ
λ1
∂T

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Radiometric Measurement and Calibration 253

where TB in the equation is the temperature of the background.

Numerous variations of these system parametric equations can be found in
the literature, each specific for a given application. Each experimenter must
model his or her own system to generate one of these equations for accurate
assessment. Once the basic equation has been generated and tested, we can do
further modeling to include the effects of other variables, such as scene
parameters, environmental influences, degradation, etc.

7.5 Introduction to the Philosophy of Calibration

What is calibration and why do we do it? The format of the raw data from a
radiometric instrument is usually in the form of a digital count or data number for
digital instruments or a voltage, current, or resistance for analog instruments.
These numbers are quite meaningless, inasmuch as the units of radiance are not
volts, and irradiance is not given in digital counts. Therefore, the primary
purpose of calibration is to assign absolute values in engineering units to
measured data according to an accepted standard. A secondary but still important
purpose of calibration is to estimate uncertainties of the acquired data.
Several formal definitions of calibration are provided below.

calibration n: (1) The set of operations which establish, under

specified conditions, the relationship between values indicated
by a measuring instrument and the corresponding known values
of a standard (NASA EOS). (2) The measurement of some
property of an object that yields as an end result a number that
indicates how much of the property the object has (Webster New
Collegiate). (3) The comparison of a measurement standard or
instrument of known accuracy with another standard or
instrument to detect, correlate, report, or eliminate by adjustment
any variation in the accuracy of the item being compared (MIL-
STD-45662A.) (4) A set of operations, performed in accordance
with a definite documented procedure that compares the
measurements performed by an instrument to those made by a
more accurate instrument or standard for the purpose of
detecting and reporting, or eliminating by adjustment, errors in
the instrument tested (Fluke, Calibration). (5) The process of
assigning engineering units and uncertainties to meter
deflections, digital counts, etc., such that an instrument reading
conforms to a recognized standard (Palmer’s definition).

The last of these five definitions is the most useful, as it adds reality to the
measurements that we conduct. The common thread in the definitions of
calibration is the involvement of a standard. A valid measurement is inextricably
linked to the calibration process and therefore to physical standards. How well
we can measure is closely related to the quality of the standards we employ, and

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 254 Chapter 7

future improvements in our measurements will necessitate better standards. A

measurement or physical standard can be defined as an accepted object, artifact,
material, instrument, experiment, or system that stores or provides a physical
quantity that serves as the basis for measurements of the quantity. It is used as a
reference for establishing a unit for the measurement of the physical quantity.
There are several types of standards used by the community. A primary
standard is one that has the highest metrological qualities. It may be realized
from first principles, calculable, or built to plan with no other measurements
required. This type of primary standard is also known as an intrinsic standard. An
example is the degree, which is based upon the triple point of water. Primary
standards may also be established by international agreement as an artifact
standard. An example is the kilogram, a particular artifact stored at the
International Bureau of Weights and Measures (BIPM) in Paris. A secondary
standard is designed to carry and transport a calibration scale. It must be as
repeatable and stable as possible, and is calibrated with reference to a primary
standard. Secondary standards are used to disseminate a scale for widespread
distribution. Finally, a working standard is similar to a secondary standard but is
one generation further removed from the primary standard. These are the
generation that we usually purchase from secondary suppliers of calibration
equipment and standards and use for routine calibrations of our instrumentation.
Note that as one gets further from the primary standard, the uncertainties increase
due to the inevitable errors present in the transfers.
Other terminology is also applied to standards. An international standard is
one which has been adopted based upon an international agreement. A transport
standard is one which has been designed to maintain its calibration through the
rigors of transport via common carrier. A consensus standard is one used by
consenting parties when no suitable standard is available.
U.S. Department of Defense requirements for the use of physical standards
are spelled out in MIL-STD-45662A, which states that they must be:

(1) certified as traceable to the National Bureau of Standards [now National

Institute of Standards and Technology (NIST)], or
(2) derived from accepted values of natural physical constants, or
(3) derived by ratio type of self-calibration techniques.

Traceability is defined as the ability to relate individual measurements to

national standards or nationally accepted measurement systems through an
unbroken chain of comparisons (MIL-STD-45662A). We have a problem here
with requirement (1) because NIST says:

NIST does not define nor enforce traceability except in its NVLAP
laboratory accreditation program. Moreover, NIST is not legally
required to comply with traceability requirements of other federal
agencies; nor do we determine what must be done to comply with
another party’s contract or regulation calling for such traceability.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Radiometric Measurement and Calibration 255

However, NIST can and does provide technical advice on how to

make measurements consistent with national standards.5

There are several possible solutions to this dilemma:

(1) We can keep on doing what we have been doing for many years,
claiming our calibrations and measurements are traceable to an agency
that disavows the word. Standards maintained in this way are called Type
I standards.
(2) There have been other attempts to define traceability, and as the above
quotation implies, they have not been entirely successful. Perhaps
another try is in order:

Traceability n: the demonstration that an instrument or artifact

standard has been either calibrated by NIST (or equivalent) at
appropriate intervals or has been calibrated against other
designated standards via an unbroken chain of comparisons. The
designated standard may be a national standard, an international
standard, or a standard based upon fundamental physical
constants.

This new definition allows us to:

(3) Go offshore to another recognized national laboratory with a reputation
for low-uncertainty measurements.
(4) Purchase or generate a Type II standard given sufficient expertise, time,
and funding. Example: freezing-point blackbody radiation simulator
(requires certified pure material), or electrical substitution radiometry
(requires measurement of electrical power). Here we generate our own
standards from first principles (i.e., standard of length by counting
fringes). NIST is moving in this direction as well, conducting research
and providing information that will allow our self determination.
(5) Implement a Type III standard using ratio and self-calibration
techniques.

There is another type of standard in regular use called a procedural or

documentary standard, also called a protocol in Europe. It is a document
outlining operations and processes to be performed in order to achieve a
particular end. These are generated and maintained by several organizations,
including the American Society for Testing and Materials (ASTM), the
International Electrotechnical Commission (IEC), and the International
Commission on Illumination (CIE), to name a few. A prime example is the
aforementioned MIL-STD-45662A.
Over the years I (Palmer) have developed a calibration philosophy which
may be of interest, consisting of the principles in Table 7.9.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 256 Chapter 7

Table 7.9 The Palmer philosophy of radiometric calibration.

(1) Calibration is the process of assigning absolute engineering units (i.e.,

radiance, temperature, etc.) to acquired data (volts, digital counts); i.e., it is
the determination of the instrument transfer function.
(2) Calibration requirements are driven by science and engineering goals.
(3) For any instrument, include calibration requirements and methodology in the
initial design phase. Among these considerations is the feasibility of
including an on-board calibrator.
(4) Apply the following general principles in the design of the calibration
exercise:
A. make the calibration independent of the specific instrument;
B. calibrate the sensor in the configuration it will be used;
C. take into account every factor that may influence the calibration: the more
that Principle B is violated, the more important Principle C becomes;
D. calibration involves a comparison with primary or secondary standards;
select appropriate standards.
(5) Conduct an error assessment during the calibration planning phase to allow
estimation of uncertainties on acquired data. Give special attention to model
error.
(6) An end-to-end calibration is preferable to summation of individual
component-level calibrations.
(7) Vary relevant external environmental parameters (temperature, pressure,
humidity, etc.) to determine their influence on the transfer function.
(8) Determine the transfer function over the entire dynamic range of the
instrument.
(9) To maximize confidence in the calibration, use several calibration
configurations and compare the results for consistency.
(10) Prior to the final calibration of a flight instrument, conduct the entire
calibration procedure on a dummy, prototype, engineering model, or
whatever is available, to uncover and fix any problems with the calibrator
and/or procedures.*
(11) Inspect and interpret the results early, while the device undergoing
calibration is still in the test position; this allows for an immediate reality
check and timely fix if needed.
(12) Calibration is the last step on the PERT/CPM/GANTT chart prior to
delivery. Because of this precarious position, it is most susceptible to the old
squeeze play, so plan ahead.
(13) Above all, adhere to the KISS (Keep It Simple, Smarty!) principle.
*
This was done with imager and sounder prior to the launch of GOES-8 and proved invaluable
(BGG).

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Radiometric Measurement and Calibration 257

Statement 4 is a response to the “de-embedding problem”; i.e., that no

system, device, or component has a unique identity outside its environment.
Embedded items will always differ from their de-embedded counterparts due to
mutual interaction/parasitics among various components. Following the
principles listed in the statement will help minimize, although never completely
eliminate, these interactions.
An example: one of the authors (Grant) had the experience of witnessing
electro-optical instrument calibration and test results and comparing them with
results of similar tests after the instrument was integrated to its platform.
Different results were observed due to the different environments. One take-away
lesson is that when developing instrument specifications at the beginning of a
design process, do not fail to include consideration of the manner/platform in
which the instrument will ultimately be used! Add a little extra margin for
postintegration performance, if you can, or at least identify the factors involved.
Further, in order to approach the requirements of calibration philosophy
Statement 4, Principle B, one must choose between a standard source and a
standard detector. If at all possible, choose a standard source when the unknown
quantity is radiance (extended source) or intensity (point source). Place the
standard next to the unknown and view them sequentially. Examples of standard
sources include blackbody radiation simulators, calibrated tungsten or deuterium
lamps, and calibrated integrating sphere sources. Choose a standard detector
when the unknown quantity is irradiance or a power. Interchange the standard
detector with the radiometer being calibrated. Examples of standard detectors
include electrical substitution radiometers and light-trapping quantum detectors.
The selection of an appropriate configuration to conduct a calibration is
governed both by the desired measurements and by the availability of appropriate
standards. There are several calibration configurations that parallel the
measurement configurations shown in Chapter 6. One of them should suffice for
any radiometric instrumentation calibration. If at all possible, use two or more
configurations to gain additional confidence in the calibration.

7.6 Radiometric Calibration Configurations

7.6.1 Introduction
The selection of an appropriate configuration to conduct a calibration is governed
both by the desired measurements and by the availability of appropriate facilities
and standards. There are a number of different ways to set up a calibration or
comparison source and a radiometer. The five basic calibration configurations
that follow are the most common combinations of radiometer aperture-stop and
field-stop considerations and source distance and size. One of them should
suffice for any radiometric instrumentation calibration. If at all possible, use two
or more configurations to gain additional confidence in the calibration.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 258 Chapter 7

Figure 7.4 DSS calibration configuration.

7.6.2 Distant small source

In the distant small source (DSS) configuration shown in Fig. 7.4, the source is
placed at a distance where the inverse square law is valid. The radiometer need
not be focused, but the image of the source must be entirely contained within the
field stop. Then,
S2
ℜ E = ( SIGNAL) , (7.18)
I
where

ℜE = Irradiance responsivity in SIGNAL/(W/m2),

S = distance from source to radiometer (m), and
I = source intensity (W/sr).

The primary advantage of this configuration is that almost any calibration

source can be used as long as the distance is sufficient to meet the inverse square
law. There are several disadvantages: (1) the signals are typically small, (2) there
may be an intervening atmosphere, (3) one must know S (the error in ℜE is twice
the error in S due to the square term), and (4) a background is present because the
source image does not typically fill the field stop. Examples of DSS sources
include small blackbody radiation simulators and small tungsten lamps at suitable
distances, and the sun and stars.
The DSS configuration can be significantly improved if the small source is
placed at the focal point of a collimator, as shown in Fig. 7.5. As before, the
image of the source must fall within the field stop. The size of the image is equal
to the size of the collimator source multiplied by the ratio of the radiometer focal
length to the collimator focal length. Then

f2
ℜ E = ( SIGNAL) , (7.19)
I

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Radiometric Measurement and Calibration 259

Figure 7.5 Eccentric pupil parabola used in DSS calibration.

where f = focal length of the collimator (m) and I = source intensity (W/sr) = LAs.
The result is again irradiance responsivity ℜE in SIGNAL/(W/m2). The
advantages to this approach include: (1) a controllable atmosphere (vacuum
chamber, if necessary), (2) a controllable background, (3) the distance S is not in
the equation, (4) you can use almost any small source, and (5) the radiometric
signals tend to be larger. The disadvantages include (1) the need to know the
focal length f (typically a one-time measurement), and (2) the collimator/chamber
hardware can get very expensive. Examples include laboratory calibrators and
low-background test chambers. As to collimators, their basic types are refractive
and reflective.
The basics of a refractive collimator are shown in Fig. 7.6. Its advantages
are: (1) relatively simple alignment due to unfolded path (in the visible, only),
and (2) a simple setup, particularly if off-the-shelf components are used.
The disadvantages include: (1) the wavelength range is limited by the τ(λ) of
the lens material, (2) reflection losses occur due to refractive index, (3) ghost
images arise due to reflections from optical elements, (4) antireflection coatings
are wavelength dependent, (5) chromatic aberration from refractive optical
elements occurs, and (6) difficulties in alignment occur if the lens is not
transmissive. However, most of these disadvantages can be dismissed if
operating at a wavelength for which the collimator components are optimized.

Figure 7.6 Basics of a refractive collimator.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 260 Chapter 7

Figure 7.7 On-axis reflective collimator.

Disadvantages of the reflective on-axis collimator include: (1) a central

obscuration, (2) diffraction from the secondary mirror mount, (3) a direct path for
stray light from the source, and (4) difficulties in baffling. On the other hand, an
advantage is that wide fields of view are possible with this collimator, shown
schematically in Fig. 7.7.
Several of these disadvantages can be eliminated by use of an off-axis
collimator, shown schematically in Fig. 7.8. For example, stray light is
minimized and an off-axis parabola minimizes aberrations. The reflective off-
axis design has a narrow field of view, however. As in so many other topics
mentioned in this book, tradeoffs and choices must be made.

7.6.3 Distant extended source

In the distant extended source (DES) configuration shown in Fig. 7.9, the distant
source subtends a larger angle than the radiometer field of view, overfilling it.
For this configuration,

SIGNAL
ℜL = , (7.20)
L

where ℜL = radiance responsivity in SIGNAL/(W/m2sr).

Figure 7.8 Off-axis reflective collimator.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Radiometric Measurement and Calibration 261

IMAGE

FIELD
STOP

FIELD ANGLE
SOURCE APERTURE
STOP

Figure 7.9 DES calibration configuration (adapted from Wolfe and Zissis).6

The advantages of this configuration include: (1) the distance between the
source and the radiometer is not important, and (2) there is no background due to
the fact that the source overfills the field of view. Disadvantages include: (1) an
intervening atmosphere, and (2) need for a large uniform source. Source
examples include White Sands, New Mexico, and a lake of known surface
temperature for remote-sensing applications, and a large integrating sphere,
blackbody radiation simulator, or a white diffuse panel in the laboratory.

7.6.4 Near extended source

In the near-extended-source (NES) configuration shown in Fig. 7.10, an extended
source is placed directly in front of the radiometer undergoing calibration.
Radiation from the source (out of focus) must completely fill the field stop. In
this configuration, Eq. (7.20) applies and the radiance responsivity is in
SIGNAL/(W/m2sr) as before.
Advantages of this configuration include: (1) the distance between the source
and the radiometer are not important, (2) there is no background, and (3) there is
minimal atmosphere. On the down side, you need a rather large uniform source.
Examples include a large-area blackbody radiation simulator, an integrating
sphere, or a transmission or reflection diffuser used with a standard lamp.

SOURCE

FIELD
STOP

FIELD
ANGLE
APERTURE
STOP

Figure 7.10 NES calibration configuration (adapted from Wolfe and Zissis).6

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 262 Chapter 7

Figure 7.11 NSS (Jones) calibration configuration (adapted from Wolfe and Zissis).6

7.6.5 Near small source

Also called the “Jones method” (as seen in Fig. 7.11) after its ubiquitous
inventor, R. Clark Jones, the near-small-source (NSS) calibration provides
radiance responsivity calculated according to

 SIGNAL  A 
ℜL =   , (7.21)
 Ls  As 

where A = aperture area (m2) and As = source area (m2).

In this approach, the source must be contained within the region bounded by
XZ and YZ; both segments make the angle θo with the optical axis. The chief ray
angle is also θo, which defines the field of view. This is simply a scaling of areas,
and the radiometer is focused at infinity. The radiance responsivity ℜL has units
SIGNAL/(W/m2sr).
The advantages of the Jones method include (1) minimal atmosphere, and (2)
the possibility of using a small calibration source. The primary disadvantage is
that you must account for background radiation. An example is the use of a small
blackbody radiation simulator that provides radiation to a system having a large
entrance aperture. Appendix H provides additional information on this method.

7.6.6 Direct method

In the direct-method approach, seen in Fig. 7.12, we use a small narrow beam
that underfills both aperture and field stops (for example, a laser). The beam
power is measured with a calibrated detector or laser-power meter. The beam is
then pointed toward the radiometer and the output is measured. The result is a
power responsivity ℜΦ in SIGNAL/W.
The primary advantage of this method is its extremely simple setup. The
disadvantages include: (1) visible background radiation, and (2) no accounting
for aperture and field-stop nonuniformities. In addition, lasers can be quite noisy.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Radiometric Measurement and Calibration 263

INCOMING BEAM

CIRCULAR
FIELD
STOP
LENS AND
APERTURE
STOP

Figure 7.12 Direct-method calibration configuration.

To minimize effects from laser drift and noise, use a beam-power stabilizer
(expensive) or a beamsplitter and another stable detector to characterize the
beam-power fluctuations during the measurement. In addition, you must ensure
that saturation of either detector does not occur.

7.6.7 Conclusion
The above calibration configurations yield different types of responsivities, but
under many circumstances we can use the simple equation for transfer of radiant
power in an optical system, Eq. (2.47):

Φ = LAΩ .

The AΩ product (T, throughput, étendue) of a radiometer is usually

characterized by the area of its entrance pupil and its field of view. If the
radiometer has both a well-defined aperture A and field of view Ω, we may
convert from one form of responsivity to another using

ℜE ℜL
ℜΦ = = . (7.22)
A AΩ

These conversions permit the use of such a well-defined radiometer to

measure one quantity using a calibration derived from a different calibration
configuration.

7.7 Example Calculations: Satellite Electro-optical System

An example will help to illustrate some of the equations presented in this chapter.
Consider a satellite in space located 200 km from a spherical source of 1-m
diameter that radiates as a blackbody of 2000 K against a background of cold
black space. This is depicted in Fig. 7.13. The satellite contains an electro-optical
system having parameters listed in Table 7.10.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 264 Chapter 7

1 m diameter
T = 2000 K

200 km

Figure 7.13 Source–satellite configuration.

(1) What is the detector noise-equivalent power, NEP?

As we are not given specific information regarding voltage or current, it is best to

use the third of Eqs. (7.14):

Ad B 1cm 2 1 Hz
NEP = = = 1 × 10−10 W . (7.23)
D* 1010 cmHz1/ 2 /W

(2) Does the source represent a point source for this configuration?

The first thing to figure out is if we are dealing with a point source or an
extended source, as we do not wish to use system performance equations
indiscriminately. We know that the source is “small,” given its distance to the
sensor, but we need to determine its relationship to our detector’s size.

Table 7.10 Satellite system parameters.

Primary mirror diameter, Do 0.2 m

f/# of optics f/3
Detector active area Ad 1 cm2
Detector D* 1010 cm Hz1/2/W
Nominal wavelength range of operation 8 μm to 12 μm

Electrical bandwidth B 1 Hz

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Radiometric Measurement and Calibration 265

To do so, we determine the “diffraction-limited spot size” on the focal plane,*

using a wavelength of 10 μm (center wavelength of our band of interest):

Dblur = 2.44λ ( f # ) = 2.44(10 × 10−6 m)3 = 7.32 × 10−5 m . (7.24)

Because the diameter of the (diffraction-limited) source image is less than the
dimension of our detector, 1 cm or 0.01 m, this source qualifies as a point source.

(3) What is the system’s noise-equivalent irradiance, NEI?

Noise-equivalent irradiance, as stated above, is often used to characterize a

system for its ability to detect distant small sources. From Eq. (7.15), the NEI of
this system may be calculated as

NEP 1 × 10−10 W
NEI = = = 1 × 10−10 W/cm 2 . (7.25)
Ad 1cm 2

(4) How does this value compare to the NEI obtained from Eq. (7.16),
above?

First, we have to determine the solid-angle field of view of the sensor. We

need to calculate system focal length f:

f = ( f # ) × Do = 3 × 0.2 m = 0.6 m . (7.26)

Next, the solid-angle field of view is determined as

Ad 1cm 2 1
Ω= 2
= 2
= = 2.78 × 10−4 sr . (7.27)
f (60 cm) 3600

If we assume that the transmittance is unity (there is no atmospheric

component, and we will assume unity optical transmission), then NEI becomes

4(3) ( 2.78 × 10−4 ) 11/ 2 Hz

1/ 2

NEI = 10 1/2 −1
= 3.183 × 10−13 W/cm 2 . (7.28)
π(20 cm)1 × 10 cmHz W

*
Note, although this is not a text on system design, the size of the diffraction-limited “blur” plays a
role in sizing a system’s detector(s). Further information on system design is obtained in
references, some listed below and others in the appendices.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 266 Chapter 7

This result differs from that in Eq. (7.25) by three orders of magnitude.
Rechecking calculations, we find no error causing a discrepancy this large; what
could the problem be?
To answer, look again at Eq. (7.25). NEI was calculated according to the area
of the detector—not the area of the system entrance pupil Ao. If we repeat that
calculation using the correct surface, we obtain

NEP 1 × 10−10 W
NEI = = = 3.185 × 10−13 W/cm 2 , (7.29)
Ao 314 cm 2

which is much better. Conclusion: Make sure you are addressing problems with
the mathematical expression which corresponds to the setup/configuration you
are analyzing.

(5) What is the expected signal-to-noise ratio?

The answer to this question requires an inversion of the range equation, Eq.
(7.13). It is:

I s τa πDo τo D *
SNR = . (7.30)
4 ( f #) ΩR2 B

It also requires that we know source intensity within the particular spectral band.
This is obtained through

I s = Ls Ap , (7.31)

where Ap is the source projected area. From a blackbody radiation calculation

program, Ls = 5.04 × 103 W/m2sr in the 8- to 12-μm band. The projected area of
the spherical source of 1-m diameter is 0.785 m2, so

I s = 5.04 × 103 × 0.785 = 3956 W/sr .

Substituting into Eq. (7.30), assuming unity transmittances and a 1-Hz electrical
bandwidth,

3956 W/sr π rad(20cm)(1010 cmHz1/ 2 /W)

SNR = ≈ 31 . (7.32)
4(3) 2.78 × 10−4 sr (4 × 1014 cm 2 )1 Hz1/ 2

Whether or not a SNR of 31 is adequate depends, of course, on the particular

application.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Radiometric Measurement and Calibration 267

Equation (7.32) may appear, on first glance, to be dimensionally inconsistent,

as it appears to reduce to units of steradians in the denominator. But look at Eq.
(7.30) again, in this way:

I s .....
SNR = .
R 2 .....

Discussing irradiance in Chapter 2, we noted that E=I/R2 can sometimes be

confusing, as to units, and that a different way to consider the expression is

I Ωo
E= ,
R2

where Ωo may be thought of as the “unit solid angle,” having value 1 sr.
Applying this notion to Eq. (7.32) allows for dimensional consistency.
In conclusion, while real-world problems may be different from those
described in this section—for example, they may include source/target
parameters that change with time and sources that differ from black/graybodies—
these equations are basic, yet powerful enough to provide the engineer or analyst
a starting point from which to develop solutions.

7.8 Final Thoughts

I often say that when you can measure what you

are speaking about, and express it in numbers,
you know something about it; but when you
cannot measure it, when you cannot express it in
numbers, your knowledge is of a meager and
unsatisfactory kind; it may be the beginning of
knowledge, but you have scarcely, in your
thoughts, advanced to the stage of science,
whatever the matter may be.
– Lord Kelvin

A measurement of any kind is incomplete

unless accompanied with an estimate of the
uncertainty associated with that measurement.
– James M. Palmer

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 268 Chapter 7

For Further Reading

Y. Beers, Introduction to the Theory of Error, Addison-Wesley, Reading,
Massachusetts (1957).
P. R. Bevington and D. K. Robinson, Data Reduction and Error Analysis for the
Physical Sciences, 2nd Edition, McGraw-Hill, New York (1992).
A. Daniels, Field Guide to Infrared Systems, SPIE Press, Bellingham,
Washington (2006).
Calibration: Philosophy in Practice, 2nd Edition, Fluke Corporation, Everett,
Washington (1994).
J. Mandel, The Statistical Analysis of Experimental Data, Dover, New York
(1984).
S. L. Meyer, Data Analysis for Scientists and Engineers, John Wiley & Sons,
New York (1975).
J. R. Taylor, An Introduction to Error Analysis, 2nd Edition, University Science,
Mill Valley, California (1997).
J. D. Vincent, Fundamentals of Infrared Detector Operation and Testing, John
Wiley & Sons, New York (1990).
H. D. Young, Statistical Treatment of Experimental Data, McGraw-Hill, New York
(1962).

References

1. S. J. Wein, “Sampling theorem for the negative exponentially correlated

output of lock-in amplifiers,” Appl. Opt. 28, 4453 (1989).
2. B. N. Taylor and C. E. Kuyatt, Guidelines for Evaluating and Expressing the
Uncertainty of NIST Measurement Results, NIST Technical Note 1297,
(1994). Available at http://physics.nist.gov/Pubs/guidelines/contents.html.
3. R. D. Hudson, Infrared Systems Engineering, Wiley & Sons , New York
(1969).
4. G. C. Holst, Common Sense Approach to Thermal Imaging, JCD Publishing,
Winter Park, Florida and SPIE Press, Bellingham, Washington (2000).
5. NIST SP-250 Appendix, U.S. Government Printing Office, Washington D.C.
(1998).
6. G. J. Zissis, “Radiometry,” Chapter 20 in The Infrared Handbook, W. L.
Wolfe and G. J. Zissis, Eds., U.S. Government, Washington, D.C. (1978).

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Index
1/f, 139 background, 6, 96
1/f noise, 143, 145, 157, 196, 226, test chambers, 259
245, 246, 304 baffles, 227
2D array, 216 bandpass filter, 301
3-dB bandwidth, 136 simple, 303
3-dB frequency, 130 bandwidth shrinkage factor, 300
barrier height, 187
basic
absorptance, 69, 72, 77, 98 radiance, 25
spectral, 69 radiometer, 221
absorption, 6, 61, 69, 79, 104, 314 Bayer algorithm, 207
absorption coefficient, 71 bias voltage, 157
accuracy, 242, 243 bidirectional reflectance distribution
acousto-optical tunable filters, 239 function, 65, 73, 317
active sources, 6 bidirectional transmittance
additive dispersion, 235 distribution function, 63, 73
air mass, 119 blackbody, 77, 121
airglow, 121 curve, 86
aliasing, 218 equation, 100
Allan variance, 245 radiation, 87, 95, 138
Amici, 231 simulator, 33, 83, 132, 226,
amorphous silicon, 199 242, 255, 261
amplifier noise, 143 source, 49
angle of observation, 32 spectral radiance, 88
aperture stop, 23, 24, 40, 41, 218, BLIP, 133
220 blooming, 204
apparent radiance, 51, 52 Bode plot, 129
area array, 199 bolometer, 155, 157
array Boltzmann’s constant, 85, 178, 181,
2D, 127 189, 313
detector, 175 Bouguer, 4, 30
imagers, 199 Butterworth RC filters, 297
atmosphere, 6, 28, 104, 235, 314
atmospheric transmission, 51
loss, 49 calibrate, 247
aurora, 121 a radiometer, 309
361

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 362 Index

calibration, 102, 253, 256 dark current, 181, 191

configurations, 241, 256, 257 detection, 5
philosophy, 257 detective quantum efficiency, 131
camera equation, 54, 57 detectivity, 132
carbon arc, 109 detector, 7
carrier lifetime, 170, 172 array, 236
CCD, 89, 193 D*, 264
noise, 142 field of view, 55
central obscuration, 57, 260 noise-equivalent power, 264
charge injection device, 204, 208, silicon, 128
211 thermal, 128
charge-coupled device, 203, 204, dielectrics, 102, 123
208, 211 sources, 108
charge-transfer efficiency, 142, 203 diffuse, 61, 227
Chauvenet criteria, 247 materials, 317
chief ray, 23 reflector, 75
chopper, 109, 218, 225, 235 surface, 65, 73, 316
reflecting, 224 diffusers, 224
circular variable filter, 236 dimensional analysis, 17
classical variance, 245 direct-method approach, 262
collimators, 259 distant extended source
color films, 201 configuration, 260
compact fluorescent lamp, 107, 116, distant small source configuration,
123 258
complementary metal-oxide Doppler
semiconductor, 207 Gaussian, 110
conduction band, 164, 177 effect, 111
cone half angle, 16, 18, 41 dynamic resistance, 196
configuration factor, 38
cosine3 law, 31, 32, 33, 58
cosine4 law, 33–35 earth
Crooke radiometer, 163 projected area, 51
cross-sectional area, 20 radiance, 122
Curie temperature, 158, 160 reflectance, 51
current responsivity, 195 effective focal length, 19
cutoff effective noise bandwidth, 136, 137,
frequency, 130 174, 218, 297
wavelength, 128, 177 Einstein diffusion constants, 190
Czerny-Turner, 232 electrical
resistance, 152
time constant, 159
D*, 132, 150, 154, 162, 174, 197 electroluminescent sources, 117
photon, 133 electron-hole pairs, 195
D**, 133 electro-optical
D*BLIP, 174, 197 instrument calibration, 257

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Index 363

system, 263, 299 frame transfer, 206

emission, 6, 76, 83, 104, 314 architecture, 205
emissivity, 76 free spectral range, 230, 237, 239
emittance, 76, 98 frequency, 134
spectral, 76, 78 Fresnel
emitter, 18 equation, 103
energy gap, 164, 166, 167 reflection losses, 71
entrance pupil, 19, 23, 24, 54, 218, full frame, 205
222, 263, 266 architecture, 204
entrance slit, 22
equation of radiative transfer, 36
basic form, 58 gain of power, 41
differential form, 46 gases, 103, 123
integral form, 36 Gaussian, 110, 111, 112, 137, 139
error, 241, 244 distribution, 147, 244
assessment, 256 noise, 135
étendue, 20, 229, 263 GE radiation calculator, 100
exhaust gases, 110 generation, 5
exit pupil, 23, 24, 54, 218 generation-recombination noise,
extended source, 34, 222, 223, 264 141, 143, 173, 197
extrinsic geometrical extent, 20
photoconductor, 169 glint, 319
semiconductor, 167 Golay cell, 163
eye damage, 123

half-power point, 298

f/#, 19, 21, 22, 35, 264 Havens limit, 150
system parameter, 53 H-D curve, 202
Fabry-Perot, 237, 238 heat capacity, 141, 147
Fastie Ebert, 232 hybrid architecture, 208
Fermi level, 176
field of view, 6, 220, 222, 251, 263
instantaneous, 217, 318 illegitimate errors, 248
of the detector, 55 illuminance, 8, 57
field stop, 23, 24, 218, 220, 258 illumination engineering, 38
filters, 236 image irradiance, 55
flat-plate solar collectors, 55 impedance, 131
FLIR devices, 315 incandescent light bulbs, 107
fluorescent lamp, 115 incoherent
flux density, 11 radiation, 250
Foote’s formula, 35 sources, 119
forward-looking infrared, 209, 318 index of refraction, 12, 79
devices, 315 infrared, 2
Fourier transform infrared photodetector array, 209
spectrometer, 237 sources, 110

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 364 Index

Infrared Handbook, The 76, 78, 226 Lambertian, 27, 37

instantaneous field of view, 217, approximation, 33, 37, 316
318 disc, on-axis, 43
integrating sphere, 5, 46, 47, 224, disc radiance, 42
261 source, 32, 310
intensity, 31, 40, 48, 266 sphere, 44
interferometer, 237, 238 sphere, on-axis, 45
interline transfer, 206 Langley, S. P., 155
architecture, 205 large-area
International Bureau of Weights and blackbody radiation simulator,
Measures, 254 261
intrinsic imaging array, 203
photoconductor, 169 laser, 118, 262
semiconductor, 164 laser-power meter, 262
invariance of throughput, 21 least-significant bit, 143
invariant, 20 lens transmission, 41
inverse square law, 37, 45, 48, 258 light-emitting diode (LED), 83, 105,
of irradiance, 17, 30 117
inversion layer, 198 linear
irradiance, 25, 28, 30, 31, 121, 201, array, 216
267, 316 photodiode array, 203
at the detector, 42, 48 linearity, 131, 247
inverse square, law of, 17, 30 Littrow, 231, 234
on the detector, 43, 171 long-wave infrared, 209
reduction in, 57 Lorentzian, 110, 112, 113
responsivity, 129 lossless
isotropic source, 58 medium, 44
I-V curve, 191, 194 optical system, 42
I-V equation, 189 LOWTRAN7, 314
lumens, 8
luminance, 8
Johnson noise, 137, 143, 152, 157, luminescence, 69, 110
162, 173, 182, 196 luminescent sources, 83
Jones method, 262 Lyot stop, 228
Jones, R. Clark, 262, 309

magnification, 54, 57
Kirchhoff, 4 majority carriers, 187
law, 77, 98, 99, 149, 314 Marcel Golay, 236
KTC reset noise, 143 marginal ray, 24
matte reflectors, 58
maximum power transfer theorem,
156
Lambert, 4 of electrical engineering, 172
Lambert-Bouguer-Beer law, 70

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Index 365

mean Johnson, 137, 143, 157, 162,

earth–sun distance, 50 173, 182, 196
square noise voltage, 135, 137 microphonic, 142
measurement, 4 power, 298
measurement equation, 250 quantization, 143, 246
metallic sources, 108 shot, 139, 182, 196
metals, 102, 123, 156, 176 temperature, 144
Michelson, 237 fluctuation, 141
microchannel plates, 184 thermal, 135
microphonic, 162 thermal fluctuation, 162
noise, 142 triboelectric, 142
Mie scatter, 119 white, 135, 146, 246, 303
minority carriers, 187 noise-equivalent
concentrations, 190 flux density, 252
model validity, 248 irradiance, 216, 252, 265
modulation transfer function, 202 photon flux, 132
moments normalization, 307 power, 154, 252
monochromator, 218, 229, 231, 234 temperature difference, 210,
moon and sun 216
angular subtense, 51 normal
multiple reflections, 47 distribution, 245
muzzle flash, 110 incidence, 35, 103
numerical aperture, 19

narrowband measurements, 217

national standard, 255 object at infinity, 53
near infrared, 209 off-axis collimator, 260
near-extended-source, 55 Ohm’s law, 136, 138, 173
configuration, 261 OLEDs, 117
near-small-source calibration, 262 optical
negative electron affinity material, axis, 24, 36
180 element, 19
Nernst glower, 108 radiant power, 128
Nicodemus, 66 radiation, 73
noise, 131, 134, 243 detectors, 127
1/f, 139, 143, 145, 157, 196, system, 21, 23, 48
226, 245, 246, 304 systems, 16, 19
amplifier, 143 thickness, 71, 72
bandwidth, 132, 149 transmission, 265
CCD, 142 organic light-emitting diodes. See
factor, 143, 184 OLEDs
Gaussian, 135 overlapping of orders, 230
generation-recombination, 141,
142, 143, 173, 197

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 366 Index

passive sources, 6 point

Pauli exclusion principle, 176 detector, 175
PCTRAN, 314 source, 29, 30, 31, 34, 220,
Peltier 222, 252, 310
coefficient, 151 polarization, 248
effect, 152 sensitivity, 216
phosphors, 115 states, 70
photocathode, 181, 185 polarized, 102, 103
photoconductive, 7, 164 polished metal surfaces, 33
detector, 169, 171, 173, 175 power
gain, 172 gain, 136, 298
mode, 194 generator, 192
photodiode, 5 responsivity, 262
photoemission, 176 spectrum, 134
photoemissive, 7, 164 precision, 243
photoemissive detector, 182 pressure (Lorentzian), 110
photoemitter, 180 primary standard, 254
photographic film, 199, 200 principle of superposition, 37
photometric, 2 projected
photometry, 191, 217 area, 13, 32
photomultiplier tubes, 89, 183 solid angle, 17, 18, 19, 20, 26
photon, 128 projection systems, 56
D*, 133 pure material, 62
detectors, 127 pyrheliometer, 221
flux, 129 pyroelectric
irradiance, 170 coefficient, 158, 160
noise, 142 detector, 157, 159, 161
radiance
spectral, 90
responsivity, 129 quantization noise, 143, 246
photopic response of the eye, 217 quantum efficiency, 176, 178, 195
photovoltaic, 7, 164, 191 quantum-well infrared
cell, 165 photodetectors, 209
detector, 175, 185, 191, 192,
195, 198
solar cell, 193 RA product, 133
physical standards, 253 radian, 13, 15
Planck, 5, 84 radiance, 24, 26, 27, 32, 52, 63, 314
constant, 89, 313 apparent, 52
equation, 84, 86, 89, 91, 92, 93 of the source, 223
expression for blackbody responsivity, 129
radiation, 84 spectral, 63, 87
function, 77 radiant
plane angle, 13, 16, 17 energy, 24
p-n junction, 185, 187 exitance, 25, 26, 27, 38, 44

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Index 367

flux, 61, 63 index, 19, 85, 231, 239

intensity, 25 resolving power, 229, 230, 238
power, 24, 25, 29, 36, 39 responsive quantum efficiency, 131,
radiation 170
contrast, 96, 97 responsivity, 128, 132, 157
geometry, 30 reverse saturation current, 189, 190,
radiative transfer, 36, 38 191
basic equation, 58, 263 reverse-bias mode, 194
radiometer, 21, 215, 241 Richardson equation, 181, 182
optical system, 311 right circular cone, 18
radiometric root mean square, 130
calibration, 17, 241 root sum square, 249
configuration, 6, 251 rough aluminum, 75
equation, 228
instruments, 215, 228, 250
measurements, 217 scattering, 6
system, 252 Schottky barrier, 198
radiometry, 1, 24, 191, 211, 223, secondary standard, 254
248 Seebeck
Raman scattering, 69 coefficient, 151
random effect, 150
errors, 243 semiconductor
noise, 244 detectors, 140
uncertainty, 249 photocathodes, 177
range equation, 251, 252, 266 Seya-Namioka, 234
ray, 11 short-wave infrared, 209
Rayleigh, 84 shot noise, 139, 182, 196, 244
scatter, 119 signal processing, 7
Rayleigh-Jeans, 93 signal-to-noise ratio, 131, 149, 217,
equation, 84, 92 244, 266, 297, 303
law, 92 silicon
RC amorphous, 199
bandpass filter, 300 detector, 128
circuit, 130 simple bandpass filter, 303
low-pass filter, 299 simple equation for transfer of
receiver, 18 radiant power, 263
reflectance, 47, 64, 70, 315 single-element detectors, 127
factor, 65, 67 Snell’s law, 12, 25
spectral, 63, 78 solar
reflecting chopper, 224 cell, 192, 198, 219
reflection, 61, 63, 83 collectors
reflective natural objects, 105 flat-plate, 55
refractive constant, 28, 50
collimator, 259 irradiance, 49
elements, 109 panels, 193

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 368 Index

spectrum, 121 sunglint, 313

zenith angle, 119, 314 sunlit scene, 52
solid angle, 15–17, 25, 66, 223 surface
field of view, 265 normal, 13
source image, 21 reflectance, 317
spatial resolution, 202 systematic
specification sheet, 215 error, 241, 247
spectral uncertainty, 249
absorptance, 69 systems
absorption coefficient, 71 analysis, 211
bandwidth, 216 engineering, 211
directional emissivity, 98
emittance, 78, 102, 106
linewidth, 251 target, 96
photon radiance, 90 temperature coefficient of
radiance, 63, 87, 250, 313 resistance, 155
blackbody, 88 temperature fluctuation noise, 141
radiometer, 215 thermal, 7, 104, 127
reflectance, 63, 78 conductance, 141, 149
response, 201 detection of optical radiation,
of the eye, 8 147
responsivity, 128, 130, 250 detector, 128, 141, 147, 312
sensitivity, 201 emission, 313
transmittance, 62, 251 fluctuation noise, 162
spectrometer, 22, 217, 230, 235 infrared, 319
spectroradiometer, 217, 228, 241 noise, 135
specular, 61, 227, 316 resistance, 147, 148
reflecting surfaces, 33 time constant, 141, 148, 159
reflection, 319 thermistors, 156
reflectors, 58 thermocouple radiation detectors,
surfaces, 73 153
speed of light, 12 thermoelectric
sphere coatings, 47 detector, 150
SPRITE detector, 209 effect, 150
standard, 254–255 thermopile radiation detectors, 153
detector, 257 three-color CCDs, 207
source, 257 throughput, 20, 229, 230, 239, 263
statistical analysis, 243 time constant, 130, 154
Stefan-Boltzmann constant, 84 total emissivity, 98
Stefan-Boltzmann law, 89, 95 traceability, 254
steradian, 15 transmission, 5, 61, 62, 79
stray light, 227, 228, 260 loss, 40, 48, 53
subtractive dispersion, 235 of the lens, 42
sun of the optics, 223
projected area, 51

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 Index 369

transmittance, 70, 72, 315 valence band, 168, 177

total, 62 vector, 11
triboelectric noise, 142 vignette, 41
tungsten lamps, 105, 258 vignetting, 57
tungsten-filament lamp, 105–106 visible, 2
tungsten-halogen lamp, 106–107 voltage responsivity, 160
Twyman-Green, 237
type A errors, 243
type B errors, 241 Wadsworth, 231, 234
wavefront, 11
Welsbach mantle, 109
Ulbricht, R., 5, 46 white
ultraviolet, 2 diffuse panel, 261
detector array, 208 noise, 134, 135, 146, 246, 297,
uncertainty, 241, 244, 253 303
uncooled thermal-imaging devices, Wien, 84, 93
210 approximation, 93, 97
uniform radiance source, 47 displacement law, 86
unit solid angle, 17 work function, 176
unpolarized light, 70 working standard, 254
Zener effect, 194
Zodiacal light, 121

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms

 James M. Palmer (1937-2007) was
a research professor emeritus in the
College of Optical Sciences,
University of Arizona. He received
his AB in physics from Grinnell
College in 1959, and his MS and
PhD degrees in optical sciences in
1973 and 1975, respectively, from
the University of Arizona,
specializing in radiometry and
infrared systems. Prior to attending
the University of Arizona, he worked in industrial positions at Hoffman
Electronics Corporation and Centralab, Semiconductor Division of Globe Union,
Inc. Over a career spanning more than 40 years, he authored or coauthored more
than 60 technical papers on many aspects of radiometry and photometry, and he
was named Fellow of SPIE in 2003. Other awards include a NASA Group
Achievement award for his work on the Pioneer Venus Mission (1979), a Tau
Beta Pi Teacher of the Year Award (1992), and a Non-Traditional Student
Teaching Award from the University of Arizona (1993). He taught numerous
short courses at SPIE conferences, CIE meetings, and conferences of the Optical
Society of America. He served as a consultant on commercial and government
projects. Dr. Palmer was a brilliant lecturer whose former students, worldwide,
have expressed gratitude for the knowledge they gained under his tutelage.

Barbara G. Grant received her BA in mathematics

from San Jose State University in 1983, and her MS in
optical sciences from the University of Arizona in
1989, where her graduate research focused on the
absolute radiometric calibration of spaceborne imaging
sensors. She was subsequently employed at Lockheed
Missiles and Space Company, Sunnyvale, California,
where supported by excellent management, she pursued
problems in infrared sensor calibration and postflight
data analysis of electro-optical payloads. She also
worked as a NASA contractor, overseeing integration
and test of imager and sounder payloads on the GOES weather satellite, for
which she received two NASA awards. She is the author of two book-length
volumes of market research for process spectroscopy instruments. Since 1995,
her consultancy, Lines and Lights Technology, has addressed problems in
systems engineering, infrared imaging and data analysis, UV measurement, and
spectroradiometry, among other areas.

Downloaded From: http://ebooks.spiedigitallibrary.org/ on 02/03/2013 Terms of Use: http://spiedl.org/terms