Agilent AN 1316

Optimizing Spectrum Analyzer

Amplitude Accuracy
Application Note

RF & Microwave Spectrum Analyzers

 Table of Contents

3 Introduction
3 Absolute versus relative measurements
4 Factors affecting measurement uncertainty
4 Sources of uncertainty
5 Relative amplitude uncertainty factors
7 Absolute amplitude uncertainty factors
8 Calculating total uncertainty
8 Examples
13 Errors
13 Overall amplitude accuracy
14 Reducing overall uncertainty
16 Summary
16 References
Introduction Absolute versus relative
measurements

This application note covers the factors affecting Amplitude measurements are either absolute
the accuracy (or uncertainty) of amplitude meas- (Figure 1) or relative (Figure 2). Absolute ampli-
urements made with spectrum analyzers, and tude is the power or voltage level of a signal. For
explains how to calculate worst-case uncertainty in example, measuring the power level of a carrier in
general situations. It will help you analyze a meas- dBm is an absolute measurement. Relative ampli-
urement and decide what procedure to follow to tude is the difference between two levels, typically
minimize overall uncertainty. Also, you will learn expressed in dB. Relative measurements use one
the most important accuracy-related features and signal or signal component as a reference. For
specifications to consider when choosing a spec- example, a harmonic distortion measurement uses
trum analyzer. This note does not cover specialized the fundamental as the reference, and the harmon-
measurement procedures for certain types of sig- ic is usually measured in dBc (decibels relative to
nals that introduce additional sources of uncer- the carrier level). Most spectrum analyzer applica-
tainty, such as noise, pulsed RF, and EMI. Refer to tions involve relative amplitude measurements.
the References for additional reading.

Figure 1. Absolute measurement (one signal level Figure 2. Relative measurement (difference between
measured with a single marker) two signal levels measured with the delta marker)

3
Factors affecting measurement uncertainty

Let’s look first at the factors affecting measure- Table 1 lists the amplitude uncertainty factors of a
ment uncertainty. Later sections will show how to typical spectrum analyzer. The range of values for
calculate the uncertainty of an amplitude measure- each factor covers a variety of spectrum analyzers.
ment by considering each of these factors, and how Most spectrum analyzers have specifications for both
to obtain the best accuracy for a measurement. absolute and relative uncertainties. Since relative
uncertainties affect the accuracy of both relative and
Sources of uncertainty absolute measurements, let’s look first at those fac-
The simplified block diagram in Figure 3 shows tors affecting relative measurement uncertainty.
the elements of a typical spectrum analyzer. Some
of these elements can contribute uncertainty to Table 1. Amplitude Uncertainty Factors
amplitude measurements. These contributions are Relative ±dB
called amplitude uncertainty factors in this appli-
Frequency response (flatness) 0.5 to 4
cation note, and are specified for most spectrum Band switching 0.5 to 1
analyzers. Scale fidelity 0.5 to 2
Reference level (IF gain) 0.1 to 1
Input attenuator switching 0.5 to 2
Resolution bandwidth switching 0.1 to 1
Display scale switching 0.0 to 1

Absolute
Frequency response 0.5 to 4
Calibrator 0.2 to 1

Preselector or
input filter Mixer

Log Envelope
amp detector

Input Resolution
IF gain bandwidth
attenuator
filter Video
filter

Local
oscillator
A/D

Ramp
generator

Display

Figure 3. Elements of a typical superheterodyne spectrum analyzer

4
Relative amplitude uncertainty factors uncertainty is often greater than the absolute fre-
Frequency response (flatness) quency response specification.
Frequency response uncertainty, or flatness, repre-
sents a window of uncertainty over the specified fre- A low-frequency RF analyzer might have a relative
quency range (sometimes called a band). See Figure 4. frequency response uncertainty of ±0.5 dB. A micro-
wave spectrum analyzer tuning in the 20-GHz
Spectrum analyzer frequency response is often the range could well have an uncertainty in excess of
single highest contributor to uncertainty. It is a ±4 dB. In our calculations, we assume the worst-
function of input attenuator flatness, mixer conver- case situation, where the frequency response
sion loss, and preselector flatness (if applicable). varies over the full amplitude uncertainty range.
Frequency response affects the displayed ampli- As shown in Figure 4, the worst-case response
tudes of signals at different frequencies. It is fre- varies from plus 1 dB to minus 1 dB for a specifi-
quency range dependent and specified as ±n dB cation of ±1 dB. We will see practical application
over a frequency range (typically a harmonic band) of these specifications later.
at a given attenuator setting.
Some spectrum analyzers require you to “peak”
Frequency response uncertainty is usually speci- the preselector (“preselector centering” in the
fied for both relative and absolute measurements. Agilent ESA series analyzers) for the frequency
Relative frequency response uncertainty describes response specifications to be valid. On spectrum
the largest possible amplitude uncertainty over a analyzers that require peaking of the signal ampli-
given frequency range relative to the midpoint tude, such as the Agilent 8593E, 8595E, and 8596E,
between the amplitude response extremes within using preselector peak prior to measuring a signal
that frequency range. yields the most accurate amplitude reading at the
specified frequency. The preselector peak function
The relative frequency response specification for a automatically adjusts the preselector tracking to
given frequency range tends to be lower than the peak the signal at the active marker. While prese-
absolute frequency response specification (defined lector peak maximizes the peak response of the
below) for the same frequency range. However, to signal of interest, it may degrade the frequency
obtain the frequency response uncertainty for rela- response at other frequencies. So, it is important
tive amplitude measurements within a band, the to set the preselector-peak values to their defaults
relative frequency response specification must be before measuring a signal at another frequency.
doubled to reflect the peak-to-peak frequency Some spectrum analyzers, such as the Agilent 8560
response. This peak-to-peak frequency response series, do not require preselector peaking.

+1 dB

–1 dB
Band 1
Specification: ± 1 dB
Uncertainty: ± 2 dB

Figure 4. Typical in-band frequency response uncertainty

5
Band switching For log mode, scale fidelity is better over small
Mixing the input signal with harmonics of the local amplitude differences, and ranges from a few tenths
oscillator (LO) allows measurements over a very of a dB for signals close together to 2 dB for large
wide frequency range. Each LO harmonic provides amplitude differences. A typical scale fidelity speci-
a different harmonic frequency band within the fication is ±0.4 dB/4 dB to a maximum of ±1.0 dB.
analyzer’s overall frequency range. When signals in The ±0.4 dB/4 dB specification is applied when the
different harmonic bands are measured, additional two signals are close in amplitude; the cumulative
uncertainties arise when the analyzer switches specification applies to larger amplitude differences.
from one band to another. On some spectrum ana-
lyzers, the band-switching points are visible as dis- The remaining relative factors in Table 1 involve
continuities in the displayed noise floor. To see if controls that can be changed during a measurement.
your measurement involves more than one band,
refer to your analyzer’s specifications for the fre- Reference level (IF gain)
quency ranges related to each harmonic band. The amplitude represented by the top line of the
Although not always specified, a typical band- graticule is the reference level. It is a function
switching uncertainty is ±1 dB. of the input attenuation and the IF gain. The
reference level control determines the IF gain.
Scale fidelity Uncertainty in the amount of IF gain at a particu-
Scale fidelity uncertainty applies when a signal lar reference level setting affects the accuracy of
at one vertical position on the display graticule is the reference level amplitude. Any changes in the
measured with respect to another signal at a dif- reference level control introduce uncertainty. On
ferent vertical position; for example, when delta spectrum analyzers that can be calibrated in linear
markers are placed at the peaks of two signals of voltage, the reference level control determines the
unequal amplitude. Scale fidelity depends upon amplitude per division scaling from the bottom line
the detector’s linearity, digitizing circuit linearity, of the graticule. For 8560 series portable spectrum
and the log/linear and vertical amplifiers’ ability analyzers, the specification for this factor is “IF gain
to transform different signal voltages into their uncertainty,” but for ESA series and 8590 series
appropriate relative power (log) or voltage (linear) analyzers it is “reference level accuracy.” An example
levels on the display. With most spectrum analyz- of a reference level uncertainty specification is
ers, the accuracy of the log amplifier degrades near ±0.3 dB at –20 dBm, with some incremental uncer-
the bottom of the measurement range. The display tainty the further you are from –20 dBm.
itself is not a factor for analyzers using digital
techniques and offering digital markers because
the marker information is taken from trace
memory, not the display.

6
Input attenuator switching is small. For example, if the analyzer’s input VSWR
Input attenuation has inherent uncertainty that is 1.3 and the source’s VSWR is 1.2, the maximum
reduces reference level accuracy only if you change uncertainty due to mismatch is ±0.1 dB. On the
the attenuator setting between reference level cali- other hand, if both the analyzer and the source
bration and the measurement. Because the input have VSWRs of 2.0, mismatch uncertainty
attenuator operates over the entire frequency approaches ±1 dB.
range of the analyzer, its step accuracy is a func-
tion of frequency. At low frequencies, expect the Absolute amplitude uncertainty factors
attenuator to be quite good; at 20 GHz, not as good. Frequency response
(Since the IF gain operates at only one frequency, Absolute frequency response uncertainty describes
it is not surprising that its uncertainty is less than the largest possible amplitude uncertainty over a
that of the input attenuator.) A typical input atten- given frequency range relative to the amplitude of
uator switching uncertainty is ±1 dB. an amplitude reference signal (for example, the
50 MHz amplitude reference of Agilent ESA series
Resolution bandwidth switching analyzers or the 300 MHz reference of Agilent 8560
Different resolution bandwidth (RBW) settings series and 8590 series analyzers).
have different insertion loss characteristics. This
may cause amplitude changes when the same sig- Calibrator
nal is measured with different settings. Changing Absolute measurements are measurements made
the bandwidth setting between amplitude measure- relative to a calibration signal, which has a known
ments degrades accuracy. A typical RBW switching amplitude. Most spectrum analyzers have a built-in
uncertainty is ±0.4 dB. calibrator (“amplitude reference” on ESA series
analyzers) that provides a signal with specified
Display scale switching amplitude at one frequency. A typical calibrator
Changing the per division scaling factor—for exam- has an uncertainty of ±0.3 dB. The calibrator pro-
ple, from 10 dB/division to 1 dB/division or to lin- vides absolute calibration for the top line of the
ear—introduces uncertainty associated with the graticule. Since the calibrator source typically
relative calibration characteristics of the log/linear operates at a single frequency, we rely upon the
amplifiers. Avoid adding this uncertainty by not relative accuracy of the analyzer to translate the
changing the scale. A typical linear-to-log switching absolute calibration to other frequencies and
uncertainty is ±0.25 dB at the reference level. This amplitudes.
may not be a factor for spectrum analyzers that
store trace data into memory and then display Since our signal to be measured is usually at a dif-
data from the memory. ferent frequency than the calibrator, we may need
to change the frequency control. Since it’s typically
Impedance mismatch at a different amplitude as well, we may need to
Although not included in Table 1 (since it depends change the reference level to bring the signal to the
on the signal source as well as the analyzer), top graticule line for best accuracy (if IF gain
impedance mismatch is another factor in measure- uncertainty is less than scale fidelity uncertainty).
ment uncertainty. The spectrum analyzer tries to These changes contribute relative amplitude uncer-
measure the signal level across a 50 Ω load (or 75 Ω tainty to the measurement.
load if your analyzer has a 75 Ω input), but the
spectrum analyzer input impedance is not exactly
the characteristic impedance, so uncertainties may
arise. In some cases, uncertainty due to mismatch

7
Calculating total uncertainty

Let’s see how to calculate measurement uncertainty minus extreme of the specification window (refer
for a few general situations. to Figure 4). Specification values for three differ-
ent instruments are given in Table 2.
Examples
Relative measurement (delta marker) within a band Table 2. Accuracy Specifications for In-Band
When comparing two signals in the same frequency Relative Measurements
band, the relevant uncertainties are frequency 8593E 8563E E4407B
response (relative) and scale fidelity. Specification Specification Specification
Factor (±dB) (±dB) (±dB)
Figure 5 shows a 900 MHz signal at the reference Frequency 1.0 to 2.0 (relative) 1.0 to 3.3 (relative) 0.5 to 2.0 (relative)
level and its second harmonic at 1.8 GHz with an response
amplitude below the reference level. If the signals Scale fidelity 1.1 max 0.85 max 1.15 max
are in the same band, the frequency response (over 70 dB range) (over 90 dB range) (over 85 dB range;
RBW≥1 kHz)
uncertainty when comparing them is two times the
1.28 max
specification (for example, two times ±1 dB = 2 dB (over 98 dB range,
peak-to-peak), since the amplitude uncertainty at RBW≤300 Hz)
each signal’s position could fall on the plus or

Fundamental

Second
harmonic

900 MHz 1.8 GHz

Figure 5. Fundamental and second harmonic in the same
frequency band

8
In general, unless the signals you are measuring are Table 3. Uncertainty Calculation for an In-Band
above 20 GHz, frequency response uncertainty will TOI Measurement
not be significant unless the signals are separated 8560E
by 50 MHz or more. As an example, let’s apply these Factor Specification (±dB) Uncertainty (±dB)
uncertainties to a TOI (third-order intercept) meas- Frequency response 1.0 N/A
urement for two 2 GHz signals separated by 50 kHz (input tone)
with third-order distortion at –80 dBc (see Figure 6). Frequency response 1.0 N/A
(distortion product)
The uncertainty calculation when making the Scale fidelity 0.85 max 0.85
(over 90 dB range)
measurement with an 8560E is shown in Table 3.
Total uncertainty 0.85

Test
tones Relative measurement (delta marker) across two or
–80 dBc
more bands
If comparing signals in different bands (between-
band measurements), add the frequency response
uncertainties of each band (see Figure 7). If the band-
switching uncertainty is specified, add it to the fre-
50 kHz Distortion quency response. If band-switching uncertainty is
product
not specified, use the absolute frequency response
2 GHz (relative to the calibrator) instead of the relative
frequency response uncertainty for each band.
Figure 6. TOI measurement

+2 dB

+1 dB 0

–1 dB –2 dB
Band 1 Band switching: ±1 dB Band 3
Freq response: ±1 dB Freq response: ±2 dB
Uncertainty: ±4 dB
Figure 7. Frequency response uncertainty for a between-band measurement

9
The relevant specifications for three different ana- Let’s make the calculation for the 8593E. Use
lyzers are given in Table 4. absolute frequency response because this is a
between-band measurement and band-switching
Let’s look at an example of measuring a second uncertainty is not specified. Frequency response
harmonic with two different instruments. The fun- uncertainty is ±2.5 dB at 10 GHz and ±3.0 dB at
damental is at 10 GHz and the level of the second 20 GHz. Notice the scale fidelity specification is for
harmonic is –85 dBc. Using the Agilent 8563E the range 0 to –70 dB. Since we need to measure
specifications for the calculation: –85 dBc, we must use a different measurement
procedure than the one used with the 8563E. After
Frequency response at 10 GHz = ±2.2 dB measuring the fundamental, change the reference
Frequency response at 20 GHz = ±2.5 dB level to –20 dBm (assume the fundamental was
Band-switching uncertainty = ±1.0 dB
Scale fidelity = ±0.85 dB measured with 0 dBm reference level). Though this
introduces IF gain uncertainty, we can now make
Total uncertainty = ±6.55 dB
the measurement. Let us calculate total uncertainty:

Frequency response at 10 GHz = ±2.5 dB

(0 dBm reference level)
Frequency response at 20 GHz = ±3.0 dB
(–20 dBm reference level)
Scale fidelity (–65 dBc) = ±1.05 dB
Reference level uncertainty
(at 0 dBm) = 0.3 dB + 0.1 X 20 = ±0.5 dB
Total uncertainty = ±7.05 dB

See Table 5 for a summary of these calculations.

Note that the 8593E can make the measurement,
but the wider display range on the 8563E allows
the measurement to be made with less uncertainty,
since the reference level doesn’t have to change.
The Agilent E4407B can make the measurement
with the least amount of uncertainty.

Table 4. Accuracy Specifications for Between-Band Relative Measurements

8593E 8563E E4407B
Factor Specification (±dB) Specification (±dB) Specification (±dB)
Band switching No specification 1.0 None
Frequency response 1.5 to 5.0 (absolute) 1.0 to 3.3 (relative) 0.51 to 3.0 2 (absolute)
Scale fidelity 1.1 max 0.85 max 1.15 max
(over 70 dB range) (over 90 dB range) (over 85 dB range; RBW≥1 kHz)
1.28 max
(over 98 dB range, RBW≤300 Hz)

1. 20 to 30 • C
2. 0 to 55 • C

10
Table 5. Uncertainty Calculations for a Between-Band Relative Measurement
8593E 8563E E4407B
Factor Specification Uncertainty Specification Uncertainty Specification Uncertainty
Frequency response ±2.5 dB ±2.5 dB ±2.2 dB ±2.2 dB ±2.0 dB ±2.0 dB
(@10 GHz) (absolute) (relative) (relative) (absolute) (absolute)
Frequency response ±3.0 dB ±3.0 dB ±2.5 dB ±2.5 dB ±2.0 dB1 ±2.0 dB1
(@20 GHz) (absolute) (relative) (relative) (absolute) (absolute)
Band switching No ±1.0 dB ±1.0 dB No
specification specification
Scale fidelity ±1.1 dB max ±1.05 dB ±0.85 dB max ±0.85 dB ±1.15 dB max ±1.15 dB
(over 70 dB range) (–65 dBc) (over 90 dB range) max (over 85 dB range) (–85 dBc)
±1.28 dB
(over 98 dB range)
Reference level ±0.5 dB N/A N/A
Total uncertainty ±7.05 dB ±6.55 dB ±5.15 dB

1. 20 to 30 • C (±2.5 dB for 0 to 55 • C)

As another example of a between-band measure- Table 6. Uncertainty Calculation for a Between-Band

ment, suppose you need to make a spur measure- Spur Measurement
ment from 22 GHz to 26.5 GHz with a carrier at 8563E
10 GHz. The spurs must be –70 dBc or lower (see Factor Specification (±dB) Uncertainty (±dB)
Figure 8). You want to know how accurately you Frequency response 2.2 2.2
(6.46 to 13.2 GHz)
can measure the spurs from 22 GHz to 26.5 GHz
Frequency response 3.3 3.3
relative to the 10 GHz carrier. The factors to con- (22 to 26.5 GHz)
sider are the frequency response (relative) for the Band switching 1.0 1.0
two bands, band-switching uncertainty, and scale Scale fidelity 0.85 max (over 90 dB range) 0.85
fidelity. Using the 8563E specifications in Table 6, Total uncertainty 7.35
the frequency response (relative) for the 6.46 GHz
to 13.2 GHz band is ±2.2 dB, the frequency response
for the 22 to 26.5 GHz band is ±3.3 dB, band-switching
uncertainty is ±1 dB, and scale fidelity (log) is
±0.85 dB, for a total uncertainty of ±7.35 dB.

Carrier

–70 dBc

Spur

10 GHz 22 to 26.5 GHz

Figure 8. Spur measurement

11
Absolute measurement (single marker) When choosing a spectrum analyzer, you need to
If only frequency and amplitude controls are determine if a particular instrument can make
changed, absolute accuracy depends on these three your measurement. Suppose you need to make a
factors: calibrator accuracy, absolute frequency spur measurement at 10 GHz, where the spur is
response from the signal to the calibrator, and IF offset from the carrier by 10 kHz; the spur must be
gain uncertainty (“reference level accuracy” for at –70 dBc or lower and the carrier level must be
Agilent ESA series and 8590 series analyzers) due measured to ±3 dB. In measuring the absolute level
to changes from the calibrator level. For a signal of the carrier, the factors to consider are calibrator
measured at the reference level of the analyzer, the accuracy, IF gain uncertainty, and absolute fre-
absolute amplitude uncertainty equals the calibra- quency response uncertainty at 10 GHz. For the
tor uncertainty plus the absolute frequency 8563E, the specifications are ±0.3, ±1.0 dB, and
response uncertainty. For a signal at a level other ±2.9 dB, respectively, for a total amplitude uncer-
than the reference level, change the reference level tainty of ±4.2 dB. This is greater than the ±3 dB
control to bring the signal to the reference level for requirement. However, Ampcor (see page 15) can
best accuracy (if IF gain uncertainty is less than reduce the frequency response uncertainty of this
scale fidelity uncertainty). In this case, absolute analyzer to an acceptable level. Since the largest
amplitude uncertainty is the sum of the reference uncertainty contributor is typically frequency
level, absolute frequency response, and calibrator response, steps taken to reduce uncertainty should
uncertainties. first focus on reducing frequency response uncer-
tainty, as described below.
Suppose you need to measure a 2 GHz signal at
–10 dBm. Using the Agilent 8560E specifications in
Table 7 as an example, the calibrator uncertainty
is ±0.3 dB. The IF gain uncertainty of ±1.0 dB can
be left out since the calibrator level is –10 dBm in
the 8560E. The absolute frequency response uncer-
tainty is ±1.5 dB, for a total uncertainty of ±1.8 dB.

Table 7. Uncertainty Calculation for an Absolute

Amplitude Measurement
8560E
Factor Specification (±dB) Uncertainty (±dB)
Calibrator 0.3 0.3
Reference level 1.0 N/A at reference level
Frequency response 1.5 (absolute) 1.5
Total uncertainty 1.8

12
Errors Overall amplitude accuracy
Up to now we have discussed uncertainties. Some modern spectrum analyzers specify overall
Uncertainties differ from errors in that we can absolute amplitude accuracy. This specification
correct for errors in our measurement. The following provides a single number that you add to the
is a source of error that you should be aware of absolute frequency response uncertainty to easily
when measuring amplitude. determine amplitude accuracy when you are mak-
ing a measurement within certain boundary condi-
Signal plus noise tions. For example, the specification for overall
The amplitude at any point on the spectrum analyzer absolute amplitude accuracy of the Agilent ESA-E
display is the sum of all energy present in the IF series analyzers is (±0.54 dB + absolute frequency
passband. Therefore, the displayed amplitude of a response) 1. Table 8 shows the amplitude uncertain-
signal is actually the signal level plus the noise level. ty of this analyzer family using the overall ampli-
That is, the displayed signal is higher than its true tude accuracy specification.
value because of the added noise. Measurements
of signals close to the displayed average noise level Table 8. Uncertainty Calculations for Absolute Amplitude
(DANL) are subject to an amplitude error due to Measurements Using Overall Absolute Amplitude
the addition of noise. This error may be large or Accuracy Specification
insignificant depending on the signal-to-noise ratio, ESA-E series
and can be corrected. See Reference 3 for a discus- Absolute frequency Uncertainty1
Frequency range response specification2 (±dB) (±dB)
sion of measuring signals near noise. For signals
that appear 10 dB or more above the noise, the 9 kHz to 3.0 GHz 0.5 1.04
3.0 GHz to 6.7 GHz 1.5 2.04
error is negligible.
6.7 GHz to 13.2 GHz 2.0 2.54
13.2 GHz to 25 GHz 2.0 2.54
A similar error can occur when measuring closely 25 GHz to 26.5 GHz 2.0 2.54
spaced signals of unequal amplitude. This happens
1. For reference levels 0 to –50 dBm; RBW = 1 kHz; video BW = 1 kHz; log scale;
when both signals pass through the analyzer’s IF 0 to 50 dB log range; coupled sweep time; sample detector; 0 to –50 dBm input
and are detected together. The displayed amplitude signal; span ≤20 kHz; 20 to 30° C.

of the smaller signal may be incorrect. This error is 2. 20 to 30° C

less than ±0.5 dB and occurs if the smaller signal is
less than 10 dB above the larger signal’s skirt.

13
Reducing overall uncertainty

Minimize changes Characterization

Before taking any data, step through a measurement Characterizing your analyzer can reduce frequency
to see if any controls can be left unchanged. If a sin- response and calibrator uncertainties. If you have a
gle input attenuator setting, resolution bandwidth, more accurate calibration signal, or one closer to
and display scaling can be used for the measure- the frequency of interest, you can use that instead
ment, then the uncertainties associated with chang- of the built-in calibrator. To reduce calibrator uncer-
ing these controls can be ignored. For best accuracy, tainty, use a signal source and a precision power
calibrate the analyzer with the controls in the same meter to calibrate the reference level (top line of the
state in which the measurements will be made. graticule). To reduce frequency response uncertainty,
use a signal source having the same frequency as
Substitute IF gain for scale fidelity the signal-under-test to calibrate the reference level.
If the reference level (IF gain) uncertainty is less Use data from a point-to-point calibration of the
than the scale fidelity, you can minimize the uncer- spectrum analyzer’s frequency response for meas-
tainty of a relative measurement by using the refer- urement correction over a frequency range.
ence level control to measure both signals at the
same vertical position, eliminating scale fidelity. Ampcor
Place the first signal at a reference amplitude using Some spectrum analyzers have a built-in amplitude
the reference level control, and use the marker to correction routine (such as Ampcor) that stores
read the amplitude value. Then adjust the refer- point-to-point frequency response characterization
ence level to bring the second signal to the same data in a table and automatically applies it to the
reference amplitude (measured with the marker), measured data to yield corrected display information.
and calculate the difference between the two refer- Using a spectrum analyzer with Ampcor, a signal
ence level settings to obtain the relative amplitude. source and a power meter, you can also correct
the input signal for unwanted variations due to the
Self calibration cables and adapters between the device-under-test
Many of today’s spectrum analyzers have self- and the analyzer (see Reference 2). Ampcor uses
calibration routines (such as the auto-alignment the table of frequency-amplitude pairs that you
function in the ESA series analyzers). These rou- enter to make a correction curve to offset the varia-
tines generate error coefficients (for example, tions in the input signal. The resulting signal trace
amplitude changes versus resolution bandwidth) represents the actual signal amplitude at the output
that the analyzer uses to correct measured data. of the device-under-test (see Figure 9).
These routines help to ensure accurate amplitude
measurements and give more freedom to change
controls during a measurement. However, the
uncertainty specifications assume self calibration
has been done; therefore, self calibration is required
to achieve specified accuracy.

14
Figure 9a. Original signal Figure 9b. Signal corrected by Ampcor

Improving sensitivity sitivity depending on its insertion loss and mis-

As mentioned earlier, displayed signal amplitude is match. Furthermore, the frequency response char-
the sum of the signal and noise present in the pass- acteristics of the insertion loss/gain and mismatch
band. For signals close to the analyzer’s noise floor will introduce additional flatness uncertainty into
(within 10 dB), the error between the displayed ampli- the measurement. Using Ampcor, you can correct
tude and the true amplitude is significant (Reference for this uncertainty.
3 gives the equation for correcting this error).
Preamplification
Therefore, when measuring low-level signals, the Because of limited IF gain, small signals cannot be
noise level of the analyzer should be as low as raised to the reference level of a log display. Scale
possible. Since the noise level depends on band- fidelity uncertainty will therefore be a factor when
width, make the measurement with as narrow a measuring these signals. Some spectrum analyzers
resolution bandwidth as possible. Of course, you have a built-in preamplifier that increases the sig-
may not always be able to use a narrow resolution nal level so that smaller signals can be measured at
bandwidth; you will have to weigh this against the top graticule line. However, the preamplifier’s
the longer sweep time. However, digital resolution gain flatness and mismatch may reduce the overall
bandwidth filters allow you to use very narrow flatness. Also, since the preamplifier’s gain must
resolution bandwidths with shorter sweep times be subtracted from the measured signal level to
than analog filters. determine true signal level, preamplifier gain
uncertainty further affects total measurement
When measuring signals close to the noise level, uncertainty. These added uncertainties can be
use video filtering for a more stable, and therefore greater than the scale fidelity uncertainty.
more repeatable measurement. When narrower video
bandwidths are not available, use video averaging Reduce mismatch
to digitally smooth the trace over multiple sweeps. Reduce mismatch uncertainty by attaching a well-
matched pad (attenuator) to the analyzer input.
Any equipment connected between the device- Since an analyzer’s match is worst with its input
under-test and the analyzer input (such as an attenuator set to 0 dB, avoid the 0 dB setting
external attenuator) may reduce the available sen- whenever possible.

15
Summary By internet, phone, or fax, get assistance with all your
test and measurement needs.
Amplitude measurements with spectrum analyzers
Online Assistance
are subject to both relative and absolute uncertain-
www.agilent.com/find/assist
ties. The overall uncertainty of a measurement
Phone or Fax
can be calculated by referring to the analyzer’s United States:
accuracy or uncertainty specifications and adding (tel) 1 800 452 4844
those factors that apply to that measurement. Total Canada:
uncertainty can be minimized by careful attention (tel) 1 877 894 4414
to measurement technique. In general, this means (fax) (905) 206 4120
limiting the number of changes to analyzer control Europe:
(tel) (31 20) 547 2323
settings and substituting smaller uncertainties for (fax) (31 20) 547 2390
larger ones. Uncertainty can be further reduced
Japan:
by characterization and amplitude correction. (tel) (81) 426 56 7832
(fax) (81) 426 56 7840
References Latin America:
(tel) (305) 269 7500
1. Agilent Application Note 150: Spectrum Analysis (fax) (305) 269 7599
Basics, literature number 5952-0292.
Australia:
(tel) 1 800 629 485
2. Agilent Application Note 1286-1: 8 Hints for (fax) (61 3) 9210 5947
Making Better Spectrum Analyzer Measurements, New Zealand:
literature number 5965-7009E. (tel) 0 800 738 378
(fax) (64 4) 495 8950

3. Agilent Application Note 1303: Spectrum Asia Pacific:

(tel) (852) 3197 7777
Analyzer Measurements and Noise, literature (fax) (852) 2506 9284
number 5966-4008E. Product specifications and descriptions in this
document subject to change without notice.
4. Amplitude Accuracy Enhancement Techniques Copyright © 1999, 2000 Agilent Technologies
Printed in U.S.A. 8/00
for the Agilent 8566A/B and 8568A/B Spectrum 5968-3659E
Analyzers (Agilent white paper).

5. Spectrum Analyzer Basics (part of Agilent

Back-to-Basics Seminar), literature number
5965-7920E.

6. Spectrum Analyzer Frequency and Amplitude

Accuracy (Agilent SR112 Fundamentals Product
Training).

7. Agilent ESA-E Series Spectrum Analyzers

Data Sheet, literature number 5968-3386E.

8. Agilent 8560 E-Series Spectrum Analyzers

Data Sheet, literature number 5965-8078E.

9. Agilent 8590 E-Series Portable Spectrum

Analyzers Data Sheet, literature number
5963-6909E.