Sensors and Actuators, 10 (1986) 239 - 248 239

SMART SENSORS

JOSEPH M GIACHINO
Ford Motor Company, Electncal and Electromcs Division, Dearborn, MI 48018 (US A )
(Received June 20,1986, accepted August 28,1986)

Abstract

The avalablllty and wide range of apphcatlon of low cost sensors have
encouraged a demand for improved sensor performance Integrated sensors
are being developed to meet the designer’s need for simpler systems Smart
sensors are becoming mtegral parts of systems performmg functions that
previously could not be performed or were not economically viable

Introduction

The phenomenal growth of low cost microprocessors has reduced the

computational expense of control systems This, coupled with the growth m
low cost sensors, especially s&con-based devices, has slgnlflcantly reduced
the cost of control systems The use of silicon for a variety of sensors [I] has
led to the natural extension of mtegratmg the transduction element and
electronics [2 - 41 The natural extension of mtegratmg electronics ISto turn
the simple sensor mto a ‘smart’ sensor It should be noted that an integrated
sensor need not be smart and vice versa A sensor, as used m this dlscusslon,
consists of three basic elements, a transduction element or sensing element,
signal detection and condltlonmg electromcs and a package
This paper will discuss the defmltlon of mtegrated sensors and smart
sensors, review examples of these sensors and discuss what impact these
features may have on future sensor desrgn and utlhzatlon

Integratedsensors

Integrated sensors are sensors that have electronics on the transduction

element itself To date most of these devices have been sltlcon based [5 - 71
and are commercially dommated by Hall sensors and plezoteslstlve pressure
sensors Sensors are now being developed that mclude other materials, such
as thm metal films, zmc oxide and polymeric films on slhcon, to produce an
integrated sensor, this utlhzes the transduction properties of one class of

Elsevler Sequola/Prlnted In The Netherlands

240

materials and the electromc propertres of slhcon, Integrated sensors have

also been fabricated with other matenals, most notably GaAs [ 81
These Integrated sensors offer slgmflcant advantages m overall size and
the ability to use small signals from the transduction element The effort to
integrate electronics on sensors 1s berg spurred on by this ability to detect
small signals, or more precisely to obtam the signal when the signal-to-noise
ratio IS optimum Optlcal sensors have especially made use of integrated elec-
tronics to produce commercial devices These include a number of sensors
for measuring exposure in cameras, optical angle encoders and optical arrays
Other sensors are also bemg commercially integrated, mcludrng pressure
sensors that have had electronics combmed with the transduction elements
There IS also research [ 9] gorng on today to utlhze three-dlmenslonal struc-
tures that are IC compatible
Integrated sensors sometimes have advantages other than the obvious
one of more units per batch A maJor cost of sensors LSthe housing Where
expensive materials or costly fabncatmg processes are bemg used to house
the sensor, reduced size may offer a slgmflcant cost reduction Integrated
electronics offer excellent opportumtles to reduce the overall transduction
element and electromcs clrcult land area A slgmflcant portion of most area
reductions 1s due to the ehmmatlon of mterconnects, which generally also
improves the rehabllrty For many blomedlcal apphcatlons, integrated elec-
tronics are a necessity to allow the device to be implanted
Commercial requirements for accurate and mamtenance-free sensors
have encouraged mdustrlal work on mtegrated sensors For accurate sensors
m many &con-based devices, knowledge of the sensmg element tempera-
ture IS important While this mformatlon can be obtamed m many ways,
integrated electromcs ensures that any temperature-compensatmg clrcultry
1s at the same temperature as the transduction element that 1s being com-
pensated
An example of a high-volume commercial sensor that has made rapld
strides m becommg integrated 1s the s&on plezoreslstlve pressure sensor
Plezoreslstlve elements, where many factors are heavily temperature depen-
dent, utilize mtegrated electromcs to offer a commercially viable device that
1s worth the reduced yield nsk Honeywell, Inc was one of the first compa-
nies to offer an mtegrated plezoreslstlve pressure sensor commercially
Today, many other compames offer plezoreslstlve pressure sensors with
varymg degrees of mtegrated electronics [lo]
Not all slllcon-based pressure sensors have gone to mtegratlon as quickly
as piezoresistive dmts An example are slhcon capacltlve pressure sensors
[ll - 131 Capacltlve sensmg elements offer improved temperature perfor-
mance, long-term stab&y and sensltlvlty over plezoreslstlve devices [14]
Many of these capacltlve devices have a full-scale output of only a few plco-
farads, yet there is, to date, no high volume commercially avalable mte-
grated capacitance sensor The technology 1s there, but the benefits to be
derived from mtegratlon are limited Capacltlve elements have a very repro-
ducible and low thermal coefficient of capacitance compared to the clrcults
 241

that they are utlhzmg Therefore, there 1s little to gam from mtegratlon
except m unique sltuatlons where physical size 1s crucial (e g , biomedical
implants) and/or where the signal ISexceptionally small

Integrated s&con-based sensors

While there has been extensive material research to support sensor

development m ceramics, metals and semiconductors (e g , GaAs), nothing
has approached the research effort m silicon
Because of the tremendous knowledge and expertise that have been
developed over the last twenty-five years m slhcon technology, there are
many fabncatlon advantages m making integrated sensors with &con single
crystal as the base This 1s true even when the transduction element ISnot a
silicon single crystal
An example of an integrated sensor ISthe infrared detector array devel-
oped at the Solid State Laboratory of the Umverslty of Michigan [15] A
broad-band &con mfrared detector structure was developed using poly-
silicon-Au thermoplles as the infrared sensing elements, and thm film dlelec-
tnc diaphragms to support the thermoplles Standard sllrcon gate MOS pro-
cessing was used to fabricate on-chip multlplexers The detectors operate
over a temperature range of 0 “C to 100 “C with a 10 ms response time and a
responsiveness of 12 V/W
The 16 X 2 element staggered linear array has one lead from each detec-
tor connected to a common ground lme and the other connected to one
input of a 16 to 1 analog multiplexer Off-chip clrcults supply binary address
signals for the multiplexer decoders The chip also contams a separate call-
bratlon thermoplle, polyslhcon resistors, diodes and MOS transistors to allow
direct measurements of both the cold Junction temperature and the thermo-
electnc power of the polyslhcon lines The overall dimensions of the chip
are 5 5 mm X 110 mm using 10 pm rules
An accelerometer fabncated at the IBM Research Laboratory at San
Jose, Cahfomla [16] 1s an example of a unique sensor made possible by
integrating the sensing element and electronics on silicon The accelerometer
itself 1s a metal-coated S102 cantilever beam that 1s fabricated on a &con
chip, where the capacitance between the beam and substrate provides the
output signal The beam and its associated capacitance, typically 3 5 femto-
farads, 1s so small that the operation of the device 1s dependent on integrated
electronics Without on-chip detection electronics, the parasltlc effects would
swamp the signal The total sensor area 1s 24 square mlls, mcludmg the
detection crrcultry This device uses the established &con IC technologies,
mcludmg epltaxlal growth and amsotroplc etching, to fabricate an integrated
slllcon-based sensor that could not be otherwise realized While many funda-
mental effects (e g , Hall effect, Seebeck effect, plezoreastance, etc ) exist
m sdlcon, there are many sltuatlons where other matenals are more suitable
as the transduction element This has led to the development of complex
242

integrated sensors where the base transduction element 1s not slhcon An

example of an integrated sensor that uses slhcon as a substrate and a dlffer-
ent set of materials as transduction elements 1s a chemical reactlon sensor
[ 17 ] This device, fabncated at the EECS Mlcroelectromc Laboratory, Um-
verslty of Cahfornla, Berkeley, 1s composed of a thm film of platinum sup-
ported on an etched s&on membrane coated with a pyroelectrlc zmc oxide
thm f&n 1 pm thick Heat liberated m a chemical reactlon at the metal ftim,
CO chemlsorptlon m this mstance, mduces surface char@;ethrough the pyro-
electric effect m zinc oxide This charge ISdetected by an on-chip MOSFET
amphfler The mtegratlon of the electromcs m this apphcatlon makes possible
the detection of very low signals From the noise observed m the expenmen-
tal work at Berkeley, the lower limit for temperature sensltlvlty, where the
slgnal-to-noise ratio 1s umty, 1s calculated to be 180 K

Integrated sensor concerns

Integrated electronics unfortunately do not only result m advantages

The most obvious concern 1s that the yield may be adversely affected. Less
obvious 1s that now the sensing element and the electromcs must be process
compatible This may result m compromises that are not necessary m a non-
integrated design Integrated sensors also requrrecareful attention to package
design Most electromcs cannot be exposed to the hostlle environments that
the transduction element must see to function Perhaps the most common
hostlle environment 1shigh temperature Not only does performance degrade,
but the overall rehablllty 1s reduced as the temperature increases This 1s
especially true for silicon-based devices
Integrated electronics offer another long-term concern The cost and
time required to make a change m an integrated design will, m most cases,
exceed that incurred m changing a simpler non-mtegrated design The more
complex and inter-related the steps of a process, the more care, thought and
testmg one must do to ensure that no adverse effects result from a ‘sunple’
change
Despite these many potential problems, there are both technical and
commercial forces makmg people examme mtegrated sensors closely A
maJor driving force behmd this effort to mtegrate electromcs on the sensor
element 1sthe production of smart sensors

Smart sensors

The most difficult task m dlscussmg smart sensors 1s m obtammg a defl-

nltlon Many believe that any sensor with integrated electromcs 1s smart
Others believe that only those sensors with integrated microprocessors are
smart For the purposes of this paper, smart sensors are those that include
some lo@c functions and/or make some type of decision This defmltlon
 243

means that those integrated sensors that merely contam signal-condltlonmg

electronics and transform a measurand change into a scaled voltage output
are not smart However, to be smart, a sensor does not have to be integrated
The two functions are, for the purpose of this paper, independent Therefore
a sensor with an ‘off chip’ microprocessor used to commumcate to the user
and modify its response 1s considered smart
The crltena used here for a smart sensor can be summmzed by stating
that a smart sensor must do one of the followmg
perform a logic function,
perform two-way communication,
make a decision
The remamder of this paper will illustrate the application of smart sensors
and the avenues being explored
Smart sensors are enhancmg the followmg apphcatlons
self calibration,
computation
communication,
multisensing

Self cabbrat~on
Most sensors have two parameters that are adlusted durmg fabncatlon
offset and gam The offset 1s the devlatlon of the output from the desved
value when the input 1s at a mmlmum The offset error 1s a common mode
error m that It IS independent of the measurand value The gam establishes
the span or the difference m the sensor output as the maxlmum and mml-
mum measurand values are apphed The gam error 1s usually a normal mode
error m that It 1s a function of the measurand value How well these two
parameter adJustments are done heavily mfluences the overall sensor perfor-
mance and cost
To further comphcate the problem, these adJustments usually change
with tune for a variety of reasons, requlrmg the device to be removed from
service and recahbrated. If there IS no way for the user of the device to
recalibrate the units once they are m service, the manufacturer over-designs
to ensure that the device will be m speclflcatlon dunng its service hfe In
either case there 16added expense
To help solve the problem where sensors are very expensive to service,
a number of companies have avsulable a smart pressure sensor that can be
remotely readJusted The key to these units ISa bull&m microprocessor that
has the correction functions m Its memory The operator sends a known
pressure(s) and the mdlvldual devices correct for the changes that have
occurred with time These sensors also include the capability to take factory
cahbratlon data stored m memory and calculate the appropriate compensa-
tion for the exlstmg temperature and, m the case of dlfferentlal sensors,
static pressure
A sub-category of self cahbratlon 1s dlagnostlcs This 1s the ability for
a sensor to determme If it 1s operatmg properly With system compIexlty
244

contmually mcreasmg, it 1s crucial that malfunctlonmg components be lden-

tlfled as quickly as possible The cost of an moperatlve petrochemical plant,
utlllty boiler or automobile due to an ‘urndentlflable glitch’ 1s no longer
acceptable Nor can the customer, be it a corporation or an mdlvldual, be
expected to have all the resources on hand necessary to trouble-shoot a
variety of complex systems This requires that components can be quickly
ldentlfled as malfunctlonmg by mmmmlly-trained personnel Smart sensors
can be an effective means to this end

Compu tatron
The ability of a sensor to do computation has permitted sensors that
either would not have been possible or would not have been as accurate to
be fabricated Previously-mentioned smart pressure sensors rely on their
computational ability to compensate for envvonmental changes, such as
temperature, as well as to correct for changes m offset and gam However,
in many instances these same results can be obtamed m ordmary sensors by
the design of more complex cxcultry, elaborate fabrication techniques and
complex testmg In some mstances though, computatlonally smart sensors
enable unique devices to be made Researchers at Case Western Reserve Um-
verslty have fabricated a pH sensor conslstmg of ten ‘ ldentlcal’ sensmg ele-
ments on a single slhcon chip incorporating signal-condltlonmg electronics
[18] By using a microcomputer, they have been able to obtam an improved
sensor by applymg statics to the sensing-element signals This allows one to
obtam the average, vanance and standard devlatlons for the set of measure-
ments Therefore, if an mdlvldual element output 1s determined not to be a
member of the set (e g , it exceeds the confidence interval, usually +3 stan-
dard deviations), It can be discarded
Many tunes these ‘ discards’ are still functlonmg elements but for some
reason have had a shift m cahbratlon A smart computational sensor can
recalibrate the mdlvldual element and after a specified number of readmgs,
readJust that element’s output so that It 1s again vahd Using such a tech-
nique, pH sensors with elements subject to unexplained calibration curve
shifts can be utilized to have unproved performance and extended life
Computational sensor abllltles can also be used to fabricate devices that
utlhze the varymg sensltlvltles of different sensmg elements to a group of
measurands An example of such a device 1s the olfactory sensor developed
by the Hltachr Research Laboratory [19] The sensor 1s composed of a set
of thick film sensing elements and electronics, mcludmg a microcomputer
The sensmg elements are composed of SLX different semlconductor oxides
that are fabricated on an alumma substrate using thick film printing tech-
niques A platmum heater IS on the back to keep the sensing elements at
400 “ C, where it 1s known that semlconductor oxides change their conduc-
tance when exposed to reducing gases Smce each element 1s made from a
different semiconductor, each havmg its own sensltlvlty to vmous reducing
gases, the combined elements can develop speclflc patterns corresponding
to each gas or smell as a histogram of conductance ratios for each element
 245

The microcomputer identifies the scent on the basis of slmllarltles calculated

by comparing standard patterns stored m memory and a sample pattern
developed by the sensing elements The gas is quantlfled by using the ele-
ment with the highest relative sensltlvlty to the gas identified
Computational ability has allowed a sensorto be fabrtcated even though
none of its mdlvldud sensing element was selective enough to function m
the intended environment By applymg computational power, a smart sensor
did the function that a normal sensor, even when integrated, could not have
accomplished It should be noted that the sensmg elements are not slhcon-
based and that the sensor 1s still ‘smart’ by almost all accepted defmltlons

Communlcatlon 1s generally accepted as the process of exchanging

mformatlon The smart pressure sensors mentioned m the Self-Cahbratlon
Sectlon have the ablhty to communrcate This ability 1s used on an ‘as
needed’ basis for calibration purposes There ISextensive work under way m
blologlcal sensor areas to develop units that can communicate [20, 211 This
IS because many studies must last for months Studies usually include a base-
lme period, during which there 1s no treatment and the research period,
dunng which data are collected after treatment has been administered
Extensive work has been done on pressure sensors [22] becuase they
have apphcatlons m cardiovascular, intercranial, mtra-abdommal, gastro-
intestinal, middle ear and mtrauterme areas The most common communl-
cation has been the ability to respond to a ‘keyed’ signal, usually an FM
signal, and to transmit on request

Multk3ensing
The ability of a sensor to measure more than one physlcal or chemical
variable simultaneously 1s by our defmltlon smart A number of the sensors
previously mentioned have this capablhty Some measure two physical
phenomena such as pressure and temperature, others, such as the olfactory
sensor, measure a variety of chemical vanables
An example of a multlsensor, which emphasizes the potential of mte-
grated silicon-based sensors, 1s the integrated multlsensor chip developed at
the Electronics Research Laboratory, Unlverslty of Cahfomla, Berkeley
[23] This chip, which 1s 8 X 9 mm’, contams conventional MOS devices
for signal condltlonmg together with the followmg on-chip sensors a gas
flow sensor, an infrared sensing array, a chemical reaction sensor, cantllever-
beam accelerometers, surface-acoustic-wave vapor sensors, a tactile sensor
array and an mfrared charge-coupled device lmager
The chip was fabncated using conventional s&con planar processmg,
slhcon mlcromachmmg and thm deposltlon techmques While the slmul-
taneous measurement of seven variables has no immediate application, It 1s a
demonstration of the posslblhtles that the avalable technologies hold
246

Smart sensor apphcation

As can be seen, smart sensors, whatever defmltlon one cares to use, are
already here The ‘smarter’ the sensor, the more complex and costly it 1s
Therefore, increased use of smart sensors requires that the users determine
Just how smart a sensor they want for then particular apphcatlon The use
of a smart sensor for a normal sensor apphcatlon 1s not prudent Systems
must be designed to accept and use smart sensors m order to make smart
sensors viable m the market place

Future developments

The development of smart sensor technology to date has far outpaced

the utlhzatlon of smart sensors Where smart sensors have been effective, it IS
because they could help perform a function where a normal sensor could not
or where the auxlhary computation and mformatlon required to perform
that system function with a simple sensor was more expensive than a smart
sensor As an example, examine the plezoreslstlve pressure sensor There are
still many plezoreslstlve sensors sold that do not have mtegrated electromcs
or are not ‘smart’ Smart pressure sensors are being used where long-term
performance without mamtenance and greater accuracy are required They
are solvmg a problem That 1s the key to the future of smart sensors there 1s
little use for elegant solutions that do not meet a need In many cases this
means that the systems m which smart sensors are used must be reconfigured
and redesigned Unfortunately, m many cases system designers are not aware
what can be done to slmphfy and improve their systems by utlllzmg smart
sensors Therefore, to accelerate the use of smart sensors sensor designers
must become more involved m system design Until the sensor engineer and
the system engineer understand each other’s capablhtles, smart sensor devel-
opment will be lunlted to lmprovmg normal sensor functions
Another factor delaymg smart sensor utlllzatlon has been the envrron-
ment that sensors must be exposed to Electronics, mtegrated or not, do not
like hostile environments, especially heat Unfortunately, it 1s not always
possible to use a remote sensor and therefore for rellablllty many apphca-
tlons require sunple sensors with their mmunum electromcs
Cost has been another maJor deterrent to the use of smart sensors
Smart sensors are more complex and therefore more costly to purchase It 1s
also more costly to modify the basic design, especially when it 1s integrated
Therefore to utilize smart sensors efficiently requires that the system be
designed to maxumze their effectiveness
However, a system designed with and for smart sensors IS more cost
effective than a system with normal sensors The lme between the compo-
nent and control portlons of systems should become blurred d smart sensors
are utilized correctly
 247

References

K E Petersen, Slhcon as a mechamcal material, Proc IEEE, 70 (1982) 420

R S Muller, D L Polla and R M White, Integrating sensors and electromcs - new
challenges for sillcon, 1984 IEEE Electrotechnol Rev, 1985, pp 56 - 58
K E Petersen, Slhcon sensor technologies, IEEE Znt Electron Devuxs Meet, Washrng-
ton, DC, Dee 1 - 4. 1985, pp 2 - 7
D L Polla, W T Chang, R S Muller and R M White, Integrated zinc oxlde-on-
sillcon tactile sensor array, IEEE Int Electron Devices Meet Tech Dtgest, 1985,
pp 133 - 136
P W Barth, Silicon sensors meet mtegrated clrcults, IEEE Spectrum (Sept 1981),
33 - 39
S Mlddelhoek, Integrated Sensors, Proc 3rd Japanese Sensor Symp , 1983, pp 1 - 10
J I Golcolea, R S Muller and J E Smith, An integrable slhcon carrier domain
magnetometer with temperature compensation, IEEE Int Electron Devrces Meet,
WashIngton, DC, Dee 1 - 4, 1985, pp 300 - 301
8 D Bursky, Double-Implanted GaAs process integrates Hall sensor with amplifier,
Electron Design, (Dee 26) (1985) 53
9 H Guckel and D Burns, A technology for integrated transducers, IEEE Int Electron
Devices Meet, Washmgton, DC, Dee 1 - 4, 1985, pp 90 - 92
10 F Goodenough, Sensor ICs processmg, materials open factory doors, Electron
Deszgn, (April 18) (1985) 131 - 148
11 J M Glachmo, The challenge of automotive sensors, Proc IEEE Solrd State Sensor
Conference, Hdton Head, SC, June, 1984, p 3
12 W H Ko, B X Shao, C D Fung, W J Chen and G J Yeh, Capacitive pressure
transducers with Integrated clrcults, Sensors and Actuators, 4 (1983) 403 - 411
13 A Hanneborg, T E Hansen, P A Ohlckers, E Carlson, B Dahl and 0 Howlech,
A new integrated capacitive pressure sensor with frequency modulated output, 1985
Int Conf on Solid State Sensors and Actuators Drgest, p 186
14 H L Chau and K D Wise, Scaling hmrts m batch fabricated &con pressure sensors,
1985 Int Conf on Solld State Sensors and Actuators Digest, p 174
15 I H Chow and K D Wise, A linear thermoplle infrared detector array with on-chip
multlplexmg, 1985 Int Conf on Soled State Sensors and Actuators Dtgest, p 135
16 K E Petersen, A Shartel and N F Raley, Micromechanical accelerometer Integrated
with MVS detection circuitry, IEEE Truns Electron Devrces, ED-29 (1982) 23 - 27
17 R T Howe and R S Muller, Resonant mlcrobrldge vapor sensor, IEEE Electron
Devtce Lett , (April) (1986) 499 - 506
18 P W Cheung, Recent Developments m Integrated chemical sensors, Proc 1981 Int
Electron Devwes Meet, Washmgton, DC, 1982, pp 110 - 113
19 A Ikegaml, and M Kaneyasu, Olfactory detection usmg mtegrated sensor, Proc 3rd
Int Conf SolldState Sensors and Actuators (Transducers 85), Phdadelphq PA,
June 11 - 14, 1985, pp 136 - 139
20 J W Knuttl, H V Allen and J D Memdl, Integrated circuit implantable telemetry
systems, IEEE Eng Med Blol Mag , (March, 1983), 47 - 50
21 M J S Smith, M A Prlsbe, L Bowman and J D Memdl, A micropower IC for
blomedlcal pressure transducer, Proc 3rd Int Conf Sohd-State Sensors and Actua-
tors (Transducers 85), Phziadelphua, PA, June 11 - 14, 1985, pp 42 - 46
22 D L Polla, R S Muller and R W White, Integrated multisensor chip, IEEE Electron
Deurce Lett , (March) (1986) 255 - 256
23 D G Fleming, W H Ko and M R Neuman (eds ), Indwellrng and Implantable Pres-
sure Transducers, CRC Press, Cleveland, OH, 1977
248

Biography

Joseph M Gzuchano receive a B S (Eng Physics) and M S (Physics)

from New York Umverslty He ISpresently a supervisor m the Electrlcal and
Electromcs dlvlslon of the Ford Motor Company, with interests m sensors
and the use of &con mlcromachmmg to fabrlcate devices