A calibration method for the measurement of IR detector spectral

responses using a FTIR spectrometer equipped with a DTGS

reference cell
O. Gravrand.*a, J. Wlassow a, L. Bonnefond a
a
CEA-LETI Minatec campus, 17 rue des martyrs, 38054 Grenoble
* [email protected]

ABSTRACT

Various high performance IR detectors are today available on the market from QWIPs to narrow gap semiconductor
photodiodes, which exhibit various spectral features. In the astrophysics community, the knowledge of the detector
spectral shape is of first importance. This quantity (spectral QE or response) is usually measured by means of a
monochromator followed by an integrating sphere and compared to a calibrated reference detector. This approach is
usually very efficient in the visible range, where all optical elements are very well known, particularly the reference
detector. This setup is also widely used in the near IR (up to 3µm) but as the wavelength increases, it becomes less
efficient. For instance, the internal emittance of integrating spheres in the IR, and the bad knowledge of reference
detectors for longer wavelengths tend to degrade the measurement reliability. Another approach may therefore be
considered, using a Fourier transform IR spectrometer (FTIR). In this case, as opposed to the monochromator, the tested
detector is not in low flux condition, the incident light containing a mix of different wavelengths. Therefore, the
reference detector has to be to be sensitive (and known) in the whole spectral band of interest, because it will sense all
those wavelengths at the same time. A popular detector used in this case is a Deuterated Triglycine Sulfate thermal
detector (DTGS). Being a pyro detetector, the spectral response of such a detector is very flat, mainly limited by its
window. However, the response of such a detector is very slow, highly depending on the temporal frequency of the input
signal. Moreover, being a differential detector, it doesn’t work in DC. In commercial FTIR spectrometers, the source
luminance is usually continuously modulated by the moving interferometer, and the result is that the interferogram mixes
optical spectral information (optical path difference) and temporal variations (temporal frequency) so that the temporal
transfert function of the DTGS has to be qualified and taken into account. The usual way is to measure it directly by
means of an optical shopper and a locking amplifier for different shopping frequencies. We present here an alternative
method to estimate this DTGS transfer function, based on the fact that a FTIR continuous scan interfergram contains the
different spectral frequencies of interest. Such a calibration method doesn’t need a specific setup as it can be performed
in standard configuration, playing only with spectrometer parameters. It allows for the precise estimation of detector
spectral shapes. However, this measurement is not absolute and the peak response needs therefore to be estimated using a
calibrated black body cavity. The method, its results and limits is presented and discussed for a set of different DTGS
cells.
Keywords: FTIR, DTGS, IR photodiode, calibration, Spectral response

1. INTRODUCTION
The measurement of IR quantum detector quantum efficiency (QE) is of first importance for a lot of applications. Indeed,
as soon as the experiment requires photometry, accurate knowledge of the spectral response of photodectors is
mandatory. Usually the photodiode delivers a photocurrent and the user wants to get information about the input photon
flux onto a given band. Indeed, a photodiode is intrinsically wide band from the cuton to cutoff. On one hand, the cuton
is usually defined by antireflection coating (AR) properties or substrate transparency, or even cryostat window
transmittance. On the other hand, the cutoff wavelength usually corresponds to the absorbing material energy gap:
photon with lower energy than this gap are not absorbed and therefore do not contribute to the detection. Potentially, the
photodiode can exhibit a uniform QE between cuton and cutoff. However, this spectral QE may vary in different ways:
the overall spectral shape may depend on general properties of AR coating or substrate, but also on the electrical

High Energy, Optical, and Infrared Detectors for Astronomy VI, edited by Andrew D. Holland, James Beletic, Proc. of SPIE
Vol. 9154, 91542O · © 2014 SPIE · CCC code: 0277-786X/14/$18 · doi: 10.1117/12.2060253

Proc. of SPIE Vol. 9154 91542O-1

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 07/10/2015 Terms of Use: http://spiedl.org/terms

 caracteristics of the photodiode itself. Indeed, for a back side illuminated photodiode, high energy photons are absorbed
very close to the rear interface, far from the junction located usually on the front interface. In this case, the
photogenerated carriers have to diffuse up to the junction in order to contribute to the detection. If the carrier recombines
before beeing collected by the junction, it won't contribute to the detection and the spectral shape may decrease for the
lowest wavelengths, approaching the cuton. The reverse example is the case of a thin absorbing layer. For wavelengths
approaching the cutoff, the material becomes less absorbing and the probability that an incoming photon is not absobed
becomes non negligible, resulting also in a QE loss. This results usually in a gradual decrease of the QE approaching the
cutoff. In other words, the spectral response cutoff becomes softer as the absorbing layer becomes thinner. Other
phenomenon might also have an impact on the spectral response of a photodiode, such as interferences in the layered
absorbing medium or substrate. At last, the real photodiode response is usually far from the ideal flat response
photodiode, and its accurate and precise measurement is often mandatory for instruments with photometric requirements.

2. THE ABSOLUTE SPECTRAL QE MEASUREMENT ISSUE

Two ways are usually used to perform this spectral QE measurement:

1. Monochromator based spectrometer
2. Fourier transform spectrometer (FTIR)
The first method is the most widely used in the astronomy community, probably due to the fact that absolute
measurement is more straightforward in this configuration: the source followed by the monochromator forms a
wavelength In a second time, this calibrated flux is sent to the characterized detector. Knowing the input flux and the
output photocurrent, it becomes relatively easy to estimate photodiode QE. The difficulties of such an approach are
manifold:
• The required spectral resolution must not be too narrow. In general, the monochromator resolution is given by
the output slit width and results in a tradeoff between resolution and output power: high resolution results in
poor output optical power, degrading the signal to noise ratio (SNR) of the measurement.
• The measurement is usually very long, depending on the required resolution and spectral band width, ie the
number of acquisition points.
• The reference detector must be very well known onto the wave band of interest, in an absolute way.
• The optical conditions must be identical between reference and characterized detector, or at least very well
controlled.

The second approach is very different. It consists in replacing the monochromator by a Fourier Transform IR
spectrometer (FTIR). This kind of spectrometer is widely used in the IR for transmittance and reflectance measurements,
especially in the pharmacy industry. Therefore, different companies propose ready to use, all integrated instruments [1,
2
]. The idea here is to use the FTIR to send spectrally modulated light onto the characterized detector. By means of
Fourier transformation, the light spectrum seen by the detector may be reconstructed. As for the preceding configuration,
a first step is necessary to measure the spectral luminance of the source and the overall spectrometer transfer function to
be able to estimate the detector spectral characteristics. The advantages of such an approach are:
• The resolution is usually very high (typically of the order of several cm-1), depending on the diaphragm size.
• The output power is usually also very high, giving good SNR
• When using continuous scans, the overall spectral measurement is very fast, which may be useful from a
practical point of view. This allows also the use of averaging to even improve SNR.
However, several difficulties are met in this measurement:
• The characterized detector must be very linear. Indeed, the measurement is based on the Fourier transformation
of the detector output signal and a detector non-linearity may results in a highly distorted output spectrum.

Proc. of SPIE Vol. 9154 91542O-2

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 07/10/2015 Terms of Use: http://spiedl.org/terms

 • As in the monochromator case, the baseline needs to be measured with a well-known reference detector.
• Only one reference detector can be used for the overall required spectrum (as opposed to the monochromator
case where the measurement can switch from one detector to another, depending on the spectral range).

3. THE FTIR SPECIFIC CASE: DTGS AS REFERENCE DETECTOR

Basically, a FTIR instrument is made up of a light source, a diaphragm (APT) limiting the geometry of the source, an
interferometer and an internal detector. There exists different kinds of interferometers but we won’t get too much into the
details in the frame of this article. More details may be found in ref [3]. This interferometer may be understood as a
Michelson interferometer: a beam splitter (BMS) separates incoming light from the source into two different optical
paths reflected by two mirrors onto the BMS. Therefore, interference occurs onto the BMS. This interference is then
modulated by the mirrors positions that introduces a controllable path difference. Hence, the output signal of such an
instrument is an interferogram. By varying the optical path difference with a movable mirror, the output intensity
exhibits changing interferences: the interferogram ( ) is sent onto the detector and the Fourier transform of this
interferogram gives the measured spectrum ( ), where is the wave number usually expressed in cm-1:
( )= ( )
As mentioned previously, the output spectrum of the instrument is the product of the source luminance and the optical
transfer function of the rest of the instrument (interferometer, beamspliter, mirrors and spectrometer windows). This
output spectrum has therefore to be precisely characterized, in the so called “baseline measurement”. This is usually
done using a Deuterated Triglycine Sulfate thermal detector (DTGS), generally integrated into the spectrometer itself. In
a second time, the output light is then sent onto the characterized detector. A schematic of this configuration is given in
Figure 1. Being a pyro detector, the spectral response of the DTGS is usually very flat, which is very interesting here.
Moreover, it is also very spectrally wide (from the visible to deep IR) which represents another strong advantage for
FTIR detector characterization. Last but not least, it is not very expensive. However, this kind of detector only works in
alternative mode (the output signal is a temperature variation under variable illumination). Therefore, for the baseline
measurement, the spectrometer is used in the continuous scan configuration. In this configuration, the interferometer
scans continuously the optical path difference with a uniform speed corresponding to a uniform translation of one
mirror of the interferometer. In this case, the interferogram is unfold uniformly in time and is proportional to time
with the mirror translation velocity as proportionality constant. As a consequence, after the Fourier transformation,
becomes proportional to a temporal frequency :
= × ×
where is the sampling frequency of the interferometer and stands for the “retardation rate” of the spectrometer,
relied to the wavelength of the laser used to measure the interferometer displacement = . For a red laser its
numerical value is = 3.164 10 cm.

For that reason, during a FTIR measurement in continuous scan mode, the measured signal ( ) is the product of the
researched spectrum by the temporal transfer function of the detector ( ) = ( ):
( )= ( )× ( )
During a transmission measurement, the baseline and the measurement are carried out using the same detector.
Everything being linear, this transfer function is taken into account and vanishes in the ratio between measurement and
baseline. However, during a spectral characterization of an outside IR detector, the detector transfer functions are
different for the baseline and the detector measurement itself. In the rest of the paper, we will assume that the
measurement is carried out in good conditions for the characterized detector: photodiodes are usually quite fast objects
regarding typical frequencies in the FTIR instrument. As an example, for a = 10 sampling rate, an IR
measurement from 1 to 20µm wavelength corresponds to a temporal spectral band extending from 632Hz (longest
wavelength) to 3.16kHz (shortest wavelength).

Proc. of SPIE Vol. 9154 91542O-3

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 07/10/2015 Terms of Use: http://spiedl.org/terms

 Figure 1: principle of th
he FTIR spectrall response meassurement

(a) Raw data (b) D

Data normalized
d to unity

Figure 2: example of DTGS baseline

b taken at
a different sam mpling frequenccy for Bruker Veertex 70 equipp
ped with a CaF2
2 beam splitter
and a NIR
R halogen lamp (left figure is ra
aw data, and riight figure are data
d normalizeed to unity)

However, this temporal speectral band is never insignificant for DTGGS detectors. As an exampple, Figure 2 sh hows a plot of
DTGS baseliines taken witth a Bruker Vertex
V 70 equuipped with a CaF2 beam splitter and a NIR halogen n lamp. Thosee
different baselines have been
b taken att different saampling frequuencies and because
b of the DTGS tem mporal transferr
function, resuulting spectruums exhibit very
v differentt amplitudes (left plot), spreading oveer more than one order of
magnitude foor sampling freequencies from
m 1.6 to 40 kH Hz. Moreoverr, overall spectral shapes aree different, ass demonstratedd
by right plott showing thee same spectraal data normaalized to unitty. As a conssequence, corrrecting this DTGSD transferr

Proc. of SPIE Vol. 9154 91542O-4

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 07/10/2015 Terms of Use: http://spiedl.org/terms

 function is abbsolutely manndatory for speectral responsse measuremen
nts. The rest of
o the paper w
will therefore be focused onn
the accurate estimation off this transfer function.
f

4.. TRANSF
FER FUNCT
TION ESTIM
MATION PROCEDUR
P RE

Generally, this transfer funnction is estim

mated using a variable
v frequ
uency shopperr wheel, a giveen light source and a lockinn
amplifier. Hoowever, this direct
d way off characterizinng DTGS pro operties is expperimentally vvery heavy. It
I requires thee
mounting of the
t shopper wheel,
w and thuss the un-mounnting of the orriginal experim
ment.
This temporaal spectral infoormation is hoowever contaiined in the mu ulti-sampling frequency basseline data sett, and the ideaa
here is to exttract this inforrmation withoout modificatiion of the exp
perimental bennch. As mentiioned previou usly, each inn
the baseline corresponds
c too a given wavvelength, but also a given temporal
t frequuency. Hencee, the spectral content beingg
identical fromm one baselinee to the other, the evolutionn seen on Figgure 2a at a giiven may onnly be related d to the DTGS S
temporal trannsfer functionn. Extracting baseline
b valuees at =4565cm-1 for insttance and plootting it in fun nction of =
, we see
s the generaal shape of the investigatedd transfer funcction (see Figure 3, black ddiamond plot points). Blackk
line is a third order polynomial fit of thiss function:
1
( )=

, , and are fitting unnknowns. As can be seen, this function fits very welll with the givven data, demo onstrating thaat
this polynomiial form is suiitable to correectly describe the
t DTGS transfer functionn in this wavellength range.

Figure 3: Transsfert function esstimation from Figure 2 data: black diamondd are 4565cm-1 data,
d black linee is a polynomia
al fit of this data
a
a blue line is the
and t results of th
he overall proceess

Using this sett of coefficiennt, it is then possible to corrrect the basellines of Figuree 2. This is shhown in Figurre 4(a). As cann
be seen, all baselines
b (corrresponding too different sam mpling frequen ncy) are gathered at the saame place, sho owing that thee
correction seeems effectivee. However, the highest frequency
f basseline (whichh undergoes tthe strongest correction) iss
different from
m the others, lying beneatth the other ones as (ie frequency)) increases, thhus demonstrrating that thee

Proc. of SPIE Vol. 9154 91542O-5

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 07/10/2015 Terms of Use: http://spiedl.org/terms

 correction is not perfect. Going furtheer, it is then possible to plot
p the standdard deviationn of the seeven correctedd
baselines as a function of . This is shoown in Figuree 5, in black line.
l This curvve exhibits loww values, below 1% on thee
overall spectrrum. Moreoveer, it exhibits one local minnimum, correesponding to the t normalizattion point of Figure 4a. Onn
both sides off this minimum m, the disperssion tends to increase a litttle, particularlly for higher ie higher frrequencies. At
A
last, on both sides of the spectrum,
s the baselines tendd to zero and so does the reesidual disperrsion. This gen neral shape of
( ) tends aggain to show that
t the correcction is not perrfect and migh ht be improveed given the daata set we havve.

om 4565cm-1 da
(a) Baseline recconstruction fro ata fit (b) Reconstruction
R from the overa
all estimation prrocess
Figure 4: Baseline reconstruction of multi samplling frequency data
d from Figurre 2.

Figure 5: Standard deviation i for 4565cm-1 data fit and blu

ns obtained on corrected baseeline from Figurre 4 (black line is ue line for the
overall processs

Proc. of SPIE Vol. 9154 91542O-6

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 07/10/2015 Terms of Use: http://spiedl.org/terms

 Hence, this dataset containns a lot more innformation thhat only the mo onospectral 45565cm-1 data ppreviously useed. The rest of
the spectrum might be usedd for a more accurate
a calibrration of this DTGS transfeer function. Inn fact, the ideaa is to look forr
the parameterr set { a, b, c, d } that minim
mizes the overrall standard deviation
d of coorrected basellines.
For a baselinne measuremeent set ( ), taken at different
d scann
ner sampling frequencies , the correected baselinee
measurementt may be written as
( )= ( ) ( ) × ( )
And the calibbration proceddure is to find the parameterr set , , , that minimizzes the overalll centered meean square

( )= ( ) ( )

Where stands for thee mean value over

o and the mean value
v over .
In other wordds, the idea is to look for thhe parameter set { a, b, c, d } that gives the lowest diifference betw ween correctedd
baselines takeen at differentt scanner veloocities, that is to say the most accurate coorrection at alll frequencies. Note also thaat
the first param
meter a may beb fitted… or not. It actuallyy simply represents a norm malization factoor. The fit maay therefore bee
carried out onn the reduced data set { b, c,
c d } using onnly normalized d baseline fromm Figure 2 b.
When done white
w a Lavennberg Markarrt algorithm [4], this minim mization proceess converges easily, provided that the a
priori parameeter set is the output of the monospectrall fit given by black dots in Figure 3. Thee resulting traansfer functionn
obtained is very
v close to the
t a priori one
o and is shoown in blue liine in Figure 3. As can bee seen in Figu ure 4b, all thee
corrected basselines then siit well onto thhe same geom metrical place, even the fasttest one (40kHHz). As a conssequence, thee
residual dispeersion of the corrected
c baseelines is very low, with noo particular sppectral feature as can be seeen in Figure 5
for the blue curve
c (the ( ) remains quuite flat onto the overall sp pectrum, with more disperssion at both en nds due to thee
low signal levvel in these raanges).

5 COMPA
5. ARISON WIITH DIREC
CT ESTIMA
ATION WIT
TH A SHOP
PPER WEEL
L

(a) Baselinee for sampling frequencies

fr betw
ween 1 and 32kkHz (b) Standard
d deviations of tthe corrected ba
aselines
(sttep 1kHz) grreen line: directt estimation
b line: fitting procedure
blue

Figure 6: Comp
parison of the direct
d estimation of the DTGS transfert
t functio
on with a shopp
per and the oveerall fitting proccess.

Proc. of SPIE Vol. 9154 91542O-7

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 07/10/2015 Terms of Use: http://spiedl.org/terms

 This calibration procedure may therefore be compared with a direct estimation of the DTGS transfer function using a
constant source, a variable frequency shopper wheel and a lockin amplifier to isolate the shopper frequency component.
This has been carried out on a DTGS from IFS28 from Brucker equipped with a KBr beamsplitter and globar source.
Shopper frequencies were taken between 400Hz and 6kHz and the scanner sampling frequencies were taken from 1 to
32kHz by steps of 1kHz. Baselines are given in Figure 6a and reconstructed baseline dispresions are given in Figure 6b
for both calibration procedures. At last Table 1 gives the parameters obtained with both calibration procedures.
As expected, the computed parameters fitted from the direct estimation or computed using the overall fitting procedure
are very close (the value of a was fixed and therefore wasn’t a fitting parameter), and the resulting disperesions are very
low for both methods. However, the residual dispersion remains lower when corrected by the overall fitting procedure,
demonstrating a better accuracy of the correction factors.

Direct estimation Fitting procedure

a 0.377 0.377
b 0.0297 0.02366
c 4.902 e-6 7.558e-6
d 4.9461 e-10 7.82 e-10
χ² 3.70e-5 1.40e-5
Table 1: Comparison between parameters obtained with the direct estimation with shopper wheel and the fitting procedure.

6. EFFECT ON MEASURED DETECTOR SPECTRAL RESPONSE

The effect of this DTGS baseline correction on an IR photodiode spectral response characterization may then be
examined. The first investigated photodiode was a backside illuminated MW photodiode fabricated on a HgCdTe MBE
grown on Ge. Characterization is carried out at 77K. Material cutoff is 5.5µm. Due to the Ge substrate, the cuton is close
to 2µm, related to the Ge band gap (backside illumination), and interference fringes are visible, that occur in the CdTe
buffer between Ge substrate and HgCdTe buffer layer. Figure 7 shows the spectral response measured on this devices
with the FTIR spectrometer:
( )
( )=
( )
A thermal detector such as the DTGS gives a quantity homogeneous to the incident optical power. As a ratio between
photonic current from the photodiode ( ) and baseline ( ) from the DTGS the resulting spectral response is a
current response homogeneous to A/W. For sake of simplicity, the plotted responses have been normalized to unity. The
plot shows to spectra: the red plot is the spectrum obtained without DTGS correction whereas the blue one is computed
with the DTGS temporal transfer function correction. As can be seen the blue spectra exhibits a conventional triangular
shape, suggesting a close to uniform ( ), whereas it appears obvious that the red spectral is not correct, because of
the strong error due to the temporal behavior of the DTGS.

This method has also been used to characterize HgCdTe photodiodes, backside illuminated, with substrate removed [5].
In this configuration, the incoming light directly enters the absorbing HgCdTe layer to contribute to the detection and the
resulting spectral QE is very flat in the absence of anti-reflection coating. Measurements performed with the calibration
method on a Brucker Vertex70 showed a very uniform QE down to very low wavelengths (Figure 8). Apart from
interference fringes near the peak wavelength (due to interferences in the absorbing layer itself, this measurement
showed an almost flat QE down to very low wavelengths. The FTIR measurement with a corrected DTGS scanned at a
1kHz sampling rate allowed an accurate measurement down to 0.7µm with a Brucker vertex70 equipped with a CaF2
beam splitter and an halogen lamp source. The plot shows also measurements carried out with UV-Visible dispersive
monochromator equipped with a silicon photodiode reference detector. As can be seen, the overlap between the two
measurements is very good in the range between 0.7 and 1 µm wavelength.

Proc. of SPIE Vol. 9154 91542O-8

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 07/10/2015 Terms of Use: http://spiedl.org/terms

 Figure 7: Effectt of the DTGS baseline correctiion on a MCT photodiode
p MW
W spectral respo
onse (IFS28, =2.5kHz)

Figure 8: Spectral
S responsse and QE meassured with this calibration method on a thinn
ned HgCdTe pho
otodiode and co
ompared to
moonochromator data.
d

CONCLUSIO
ON

We proposedd a new calibraation proceduure to correct forf the temporal transfer fuunction of DTGS reference cells during a
FTIR detectoor spectral response measuurement. The method is based b on the use of a set of baseline measurements
m s
performed ussing different interferometeer velocities inn continuous scan,
s and expploits the tempporal informattion containedd
in these baselline measurem ments. The ressulting correctted baselines are then indeppendent of thee scanner veloocities and thee
correction paarameters appeear very closee the parameteers obtained with
w a direct measurement
m using a variaable frequencyy
shopper wheeel and a lockinn. We believee that this calibbration proced
dure takes advvantage of all frequencies available in thee
required specctral range of the
t final meassurement and allows an opttimum estimattion of the FT TIR baseline and
a therefore a
very accuratee characterizattion of photoddiodes spectrall responses an
nd QE. This method
m has beeen applied succcessfully withh
different FTIIR spectromeeters (IFS25, IFS28, Verttex 70 and vertex80) v to characterize different detectors: MCT T
photodiodes cutting downn to very low w wavelengthhs [6] (which corresponds to very highh frequenciess ie importannt
correction facctors), to veryy long MCT detectors,
d up tot 20µm [7], dual
d band dettectors [8], MCCT APDs, InS Sb focal planee
9
arrays and evven microbolom meter arrays [ ].

Proc. of SPIE Vol. 9154 91542O-9

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 07/10/2015 Terms of Use: http://spiedl.org/terms

 REFERENCES

1
www.brucker.com
2
www.perkinelmer.com
3
P. Griffiths, “Fourier Transform Infrared Spectrometry”, Second Edition, Wiley-interscience (2007)
4
W.H. Press, “Numerical recipes in C, The Art of Scientific Computing”, Second Edition, Cambridge University Press
(1992)
5
O. Gravrand et al, “HgCdTe Quantum Detection : from Long-Wave IR down to UV,” Journal of Electronic Materials,
vol. 40, no. 8, pp. 1781–1784, (2011), DOI: 10.1007/s11664-011-1691-0
6
O. Gravrand et al, “Ultralow-Dark-Current CdHgTe FPAs in the SWIR Range at CEA and Sofradir,” Journal of
Electronic Materials, vol. 41, no. 10, pp. 2686–2693, (2012)
7
O. Gravrand et al, “From Long Infrared to Very Long Infrared Wavelength Focal Plane Arrays Made with HgCdTe n
+ n −/p Ion Implantation Technology,” Journal of Electronic Materials, vol. 36, no. 8, pp. 981–987, (2007)
8
G. Destefanis et al, “Status of HgCdTe Bicolor and Dual-Band Infrared Focal Arrays at LETI,” Journal of Electronic
Materials, vol. 36, no. 8, p. 1031, (2007)
9
A. Touvignon et al, « IR microbolometer array spectral response measurement and angular effect using FTIR”, proc.
SPIE 9071-32, Baltimore (2014)

Proc. of SPIE Vol. 9154 91542O-10

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 07/10/2015 Terms of Use: http://spiedl.org/terms