Crosstalk model of a deformable-mirror-based

infrared scene projector

Kenneth J. Barnard, MEMBER SPIE Abstract. An analytical model for crosstalk is developed for an infrared
Memphis State University scene projector that utilizes a deformable-mirror device as an infrared
Department of Electrical Engineering spatial light modulator. Partial-coherence effects that result from the
Memphis, Tennessee 31852 physical layout of the projection system are included. Crosstalk is de-
E-mail: barnard @eosms.aa.wpafb.afmil termined by defining a crosstalk ratio in the projected image. Crosstalk-
ratio calculations are carried out for monochromatic and blackbody pro-
Glenn D. Boreman, MEMBER SPIE jection sources over the 3- to 5-rim wavelength band. The results predict
University of Central Florida crosstalk ratios of 260:1 , 1200:1 , and 2400:1 for deformable-mirror pixels
Center for Research in Electro-Optics and of 50, 100, and 1 50 i.m when a blackbody source is used in the projec-
Lasers tion system. Crosstalk performance is improved at shorter wavelengths
Department of Electrical Engineering and for larger deformable-mirror pixels. The results of a proof-of-principle
Orlando, Florida 32826 experiment are included to verify the feasibility of a deformable-mirror
device in the projection system.
Dennis R. Pape, MEMBER SPIE
Subject terms: infrared technology; infrared scene projection; deformable mirrors;
Photonic Systems Incorporated
1800 Penn Street, Suite 4B spatial light modulators; partial coherence.
Melbourne, Florida 32901 Optical Engineering 33(1), 140—149 (January 1994).

1 Introduction increasing the magnitude of the address voltage of the de-

The goal of infrared (IR) scene projection is to provide re- vice.9 The mirror elements are coated with aluminum as the
alistic testing and evaluation of ihfrared-imaging sensors in reflective surface, making the device truly broadband. In ad-
a controlled laboratory environment. For a valid test, a dy- dition, the DMD can achieve high frame rates of the order
of 10 kHz. In this paper, we present the optical design of an
namic spatial and temporal scene that accurately represents
real-world targets over a specified wavelength band must be IR scene projector based on the DMD and include the results
generated. Requirements for scene projection are determined of a proof-of-principle experiment to verify the validity of
by the sensor under test and can be broken down into several the design. A crosstalk model is developed for the system
incorporating a DMD with square pixels that tilt along a
criteria: spatial resolution, dynamic range, contrast ratio,
spectral distribution of in-band flux, frame rate, and spatial 45 deg torsion axis. Partial coherence effects are considered
uniformity. ' The ultimate objective for IR scene projection in the model.
It is shown that the system behaves coherently under quasi-
is hardware-in-the-loop (HWIL) testing of thermal-imaging monochromatic conditions. Also, in the case of blackbody
systems. In HWIL testing, an imaging IR seeker would be
mounted on a computer-controlled gimbal and an IR scene illumination, the spectral characteristics of the source may
generator/projector would present a realistic scene to the be approximated as a summation of narrow-band quasimon-
seeker, just as would appear in actual use.4'5 The main dif- ochromatic sources. Each of these sources produces an in-
ficulty in IR scene projection lies in producing complex dependent image-irradiance function in the image plane. The
scenes of sufficient quality and complexity to characterize summation of image-irradiance functions gives the total im-
age irradiance, and a crosstalk ratio is determined based on
today' 5 high-performance IR systems.
The Texas Instruments digital micromirror device6'° the projected image. Under blackbody illumination, the
model predicts crosstalk ratios of 260:1, 1200:1, and 2400:1
(TI DMD) has the potential to satisfy current requirements for DMD pixel sizes of 50, 100, and 150 m.
for IR scene projection when used as a spatial light modulator
(SLM) for IR radiation from a thermal or laser source. The 2
TI DMD consists of a matrix of discrete reflective elements Optical Layout
that can independently tilt in response to an input video signal. Figure 2 is a layout of the projection system incorporating a
A sketch of a portion of the DMD is shown in Fig. 1. Although blackbody source. Two optical subsystems comprise the
the DMD can be operated in a bistable mode where the mirror complete system: an illumination system and an imaging
pixels tilt at angles of + 10 and — 10 deg, in this paper, we system. If a laser source is used, the layout is the same except
describe the device as having an off state represented by an a collimated laser beam replaces the blackbody and colli-
untilted pixel and an on state represented by a 10 deg pixel mator. The imaging system is a two-lens relay system that
tilt. This particular mode of operation can be achieved by images the DMD plane onto the focal plane of the sensor
under test. Relay lens 2 is the lens system of the sensor under
test. Together, the two relay lenses produce an overall mag-
Paper 13013 received Jan. 19, 1993; revised manuscript received May 22, 1993; nification of m = —f2/f1, wheref1 andf2 are the focal lengths
accepted for publication May 23, 1993. of relay lenses 1 and 2. The lens spacing is such that the
1994 Society of Photo-Optical Instrumentation Engineers.
0091-3286/94/$6.00. source is imaged at the entrance pupil of relay lens 2. This

140/OPTICAL ENGINEERING /January 1994/Vol. 33 No. 1

Downloaded from SPIE Digital Library on 23 Nov 2010 to 132.170.6.32. Terms of Use: http://spiedl.org/terms
 CROSSTALK MODEL OF A DEFORMABLE-MIRROR-BASED IR SCENE PROJECTOR

Fig. 1 Sketch of a portion of TI DMD with 45-deg torsion-beam pix-

els.

DM1)
Collimator lens
//
Collimator axis 45 out
of the plane of the drawing

Thermal source
Focal plane array
Fig. 3 Geometrical-optics model of DMD operation.

2O

Relay lens #1 Relay lens #2

"Off" pixe1'"-

Fig. 2 Optical layout of projection system.

_llIll
iEiiiiE
n:n::.::nuDD :n:Du
configuration produces a uniform flux distribution across the
image plane and minimizes vignetting effects. However, it
would be possible to increase the lens spacing and separate
the aperture stop from relay lens 2 if necessary. In the case
of a coherent source, such as a laser, a Fourier transform
plane would exist at the rear focal plane of relay lens 1 and
an aperture stop located at this plane would ensure shift-
invariant imaging in terms of field amplitude.
Figure 3 is a geometrical model of the projection system
operation. The illumination system consists of a thermal
source and a collimator. Collimated rays are directed at the
: :::t::::-
DMD at an angle of 20 deg. Rays reflected from the DMD
pixels in an untilted off state are directed at an angle of 20 deg Fig. 4 Near-IR projected scene using a DMD with 17-gm-square
from the optical axis. For pixels tilted at an angle of 10 deg, pixels and a wavelength passband of 0.830 to 1 .0 rim.
the rays are directed along the optical axis and the pixels are
imaged at the sensor focal plane. The arrangement shown is
referred to as dark-field illumination because the DMD pixels for a binary input signal. This figure demonstrates that high-
appear black until tilted to a state that redirects the input contrast image projection is possible using the DMD as a
radiation along the optical axis of the imaging system. By SLM in an IR scene projector. Note that the DMD used was
applying a pulse-width modulation scheme during the inte- a demonstration device and did possess pixels that were stuck
gration period of the sensor, gray-level operation can be in either the + 10- or — lO-deg positions.
achieved.
As a verification of operation for the given projector de- 3 Partial Coherence Effects
sign, a proof-of-principle experiment was performed. The The image-forming process for JR imaging sensors is typi-
DMD used was a 1000 X 1000 array with 17-pm-square pix- cally an incoherent process because thermal sources produce
els and 45 deg torsion hinges. A filtered tungsten-halogen essentially incoherent radiation. In many optical systems,
source with a passband of 0.830 to 1.0 xm was used as the such as projection systems with both an illumination system
thermal source. A silicon CCD camera was used as a near- and an imaging system, coherence effects become impor-
infrared imager. For 3- to 5- or 8- to 12-rim operation, a tant." This is the result of the Van Cittert-Zernike theorem,
device with larger pixels would be necessary. Figure 4 shows which states that for an extended quasimonochromatic
the sensor output for a magnified portion of the DMD array source, spatial coherence improves with decreasing angular

OPTICAL ENGINEERING/January 1994/Vol. 33 No. 1 / 141

Downloaded from SPIE Digital Library on 23 Nov 2010 to 132.170.6.32. Terms of Use: http://spiedl.org/terms
 BARNARD, BOREMAN, and PAPE

subtense of the source. In a projection system, the source will

produce a finite-coherence area at the object plane. If the
resolution spot of the imaging system (imaged at the object
plane) is smaller than the coherence area of the source at the
object plane, the imaging will be approximately coherent.
For a resolution-spot size much greater than the coherence
area, the imaging will be incoherent. If neither condition is
satisfied, the imaging will be partially coherent.
In this section, we determine which imaging regime (co-
herent, incoherent, or partially coherent) describes the IR
scene projector. This analysis is based on a quasimono-
chromatic source assumption. For the actual projection sys-
tem, the source is a blackbody and polychromatic source
effects are included. Based on these results, we develop an
—60 —40 —20 0 20 40 60
imaging model for the projector. X (tim)
Starting with a quasimonochromatic source, we determine
the mutual-intensity function (MIF) at the object plane. The Fig. 5 Image-irradiance function for a 5O-pm slit object in the par-
MIF can then be used to obtain the coherence area. Com- tially coherent model (X=3.O tim).
paring the coherence area with the diffraction-limited
resolution-spot size ofthe imaging system gives the condition
for approximately coherent imaging as In Sec. 6 we derive an approximation of the mutual-
coherence function for a polychromatic source based on a
3.8 lf1D summation of narrow-band quasimonochromatic mutual-
z> (1) coherence functions. In imaging applications, each separate
D mutual-coherence function produces an independent image.
where z is the distance between the collimator and the DMD, The polychromatic image is then given by the summation of
D is the collimator diameter, and D, is the diameter of the the quasimonochromatic image-irradiance functions. The
aperture stop located at relay lens 2. Details of the derivation mutual-coherence functions are evaluated at discrete fre-
of Eq. (1) are given in Sec. 6. Because of the functional de- quencies centered about narrow bands over the passband of
pendencies in Fig. 2, this requirement is independent of the interest. Using this approach in conjunction with the quasi-
wavelength of the quasimonochromatic source and is a purely monochromatic coherent analysis allows the image irradiance
at the focal plane of the sensor under test to be calculated by
geometrical constraint. For the given projector design, D
and D will be on the same order and the distance z must be determining several coherent-image-irradiance functions at
greater than approximately 4f1. This condition can easily be discrete wavelengths and then performing a summation over
met for typical values of focal lengths and fl#s. For our the passband to determine the total image irradiance. The
system-design values of f = 50 mm, D = 25 mm, and spectral characteristics of the source and the spectral re-
D1 = 23 mm, the requirement for coherent imaging becomes sponsivity of the focal plane array must also be included to
z > 207 mm. For values of z less than those of Eq. (1), the characterize the performance of the JR projection system.
imaging system will be partially coherent. By using Eq. (1)
as a design constraint, we can force the system to behave 4 Coherent Model
coherently. A coherent system ensures linear imaging in The ultimate performance limitation ofthe JR scene projector
terms of field amplitude and enables standard Fourier optics will be crosstalk in the final image. The crosstalk results from
analysis for modeling. For a partially coherent system, im- diffraction and will cause off pixels to have some nonzero
aging characteristics are nonlinear and object dependent. flux associated with them. An aperture stop in the projection
A one-dimensional partially coherent model of the pro- optics eliminates unwanted diffraction orders and its size
jector system was developed to verify coherent imaging under determines crosstalk performance for a given wavelength.
the condition given by Eq. (1). To simplify the analysis, the The following analysis provides a theoretical prediction of
DMD was replaced by a transparent 5O-pm slit representing crosstalk defined in terms of a crosstalk ratio in the final
a single DMD pixel. The illumination system was such that projected image. Standard Fourier optics techniques13 are
the slit was transilluminated along the optical axis. The dis- used for the case of coherent illumination. Based on the re-
tance z was chosen to be 300 mm for this analysis and the sults of the coherent model, polychromatic illumination is
design values from the previous paragraph were also used. assumed for an optimized optical layout and a polychromatic
A four-dimensional linear-systems approach incorporating crosstalk ratio is then calculated.
the transmission cross-coefficient was used to characterize Jn the coherent model, the imaging system is treated as a
the system.12 The ringing effects in the resultant image- linear system in terms of electric-field amplitude and can be
irradiance function shown in Fig. 5 illustrate the coherent written in the spatial frequency domain as
effects in the projected image and support the coherent-
analysis approach. , (2)
The actual source in the projection system is a blackbody,
filtered over a spectral band that is much larger than that where U(,i1) and U0(q) are the spectrums of the image
assumed in the quasimonochromatic analysis. Polychromatic and object fields, and H(,i) is the coherent optical-transfer
source effects are incorporated to predict actual performance. function. For a coherent system, the optical-transfer function

142 / OPTICAL ENGINEERING / January 1994/Vol. 33 No. 1

Downloaded from SPIE Digital Library on 23 Nov 2010 to 132.170.6.32. Terms of Use: http://spiedl.org/terms
 CROSSTALK MODEL OF A DEFORMABLE-MIRROR-BASED IR SCENE PROJECTOR

is simply the scaled exit pupil of the optical system. The with all mirrors untilted. The second term is a function rep-
object field is given in terms of both the input source field resenting the single tilted center pixel. The last term removes
and the reflectance function of the DMD. redundancy for the center pixel from the first term. Using
We have chosen to define a crosstalk ratio in terms of a standard Fourier optics notation, the reflectance function is
specific DMD input where only the center pixel is tilted. This given by
configuration allows us to calculate on-axis crosstalk result-
ing from a single on pixel. The image-plane crosstalk ratio
is found by computing the amount of flux in the geometrical Ii
r(x,y)= L comb)** rect_7_7)j
( y\l ( y
image of the on pixel and the amount of stray flux in the rect,
geometrical image of an adjacent off pixel. The crosstalk
ratio is defined as the ratio of these two flux values and is F sin(2a) 1'x y\
given by (x+y) rect( —,— I
+exP[—Jk ] \W WI

c=
_1on pixel region
I(x,y) dAimage

I(x,y)
, (3) —rect
(x
— \ (6)

I-off pixel region

where d is the center-to-center pixel spacing, W is the on
where C is the defined crosstalk ratio and I(x,y) is the image- mirror width, W' is the off mirror width, and D is the total
irradiance function. This definition of crosstalk ratio is dif- DMD width. The two assumptions made in deriving this
ferent from the usual low-spatial-frequency definition of con- expression are that (1) the hinge structure of the pixels is
trast ratio in terms of all-dark or all-bright fields. The use of ignored and (2) the tilted pixel shape remains square but
crosstalk ratio yields the pixel-to-pixel crosstalk level that reduced in size. A tilted pixel is represented by a skewed rect
the projector is capable of, for a given set of optical param- function and cannot be expressed in a form that is separable
eters, assuming a particular pixel-array geometry in the sensor in x and y. Substituting a square pixel shape with an area
under test. equal to the projected area of the tilted pixel allows the object
Figure 6 is a cross section of a portion of the DMD with field to be written in a separable form and simplifies calcu-
the center pixel tilted at an angle of a and the associated lation of the image field. The area of the square pixel shape
optical fields. The input source field propagates at an angle is
of 2a and reflects off the DMD. Ignoring constant terms, the = W'2 cos(a)
source field is a linear-phase plane wave that can be written , (7)
as where w'2 is the area of the untilted DMD pixel.
The object-field amplitude is related to the source field
F . sin(2a) and the reflectance function by
u(x,y) = exp jk (x + y) (4)
u,(x,y)=u(x,y)r(x,y) . (8)
The reflectance function of the DMD can be written by con-
sidering a small portion of the array, as in Fig. 7. For the Substituting Eqs. (4) and (6) into Eq. (8) gives the object
case of having only the center pixel tilted, the reflectance field as
function can be written as the sum of three terms:
l /xy
r(x,y) = all mirrors off u(x y)=Ecomb ** rect( —,—J
(x y) \W WJ
+ center mirror on —center mirror off . (5)
F sin(2a) 1 rect — y\ /x y\
The first term in this expression represents the DMD structure
XexpLik _
(x+Y)]
—I +rect( —,—)
(xD'DJ \W W/

u(x,y) F. sin(2a) 1 /'xçy\ .

—
exp jk
\/ (x + y) rect —v (9)

The object field, scaled to reflect the overall magnification

of the optical system, is given by

1 /x
u0(x,y) =—u( —,— J , (10)
g(X,y)
DMD m mm,
where m = —f2/f1 for the two-lens relay system. Substituting
Eq. (9) into Eq. (10) and Fourier transforming yields the ob-
Fig. 6 Input optical fields. ject field spectrum

OPTICAL ENGINEERING / January 1994 / Vol. 33 No. 1 / 143

Downloaded from SPIE Digital Library on 23 Nov 2010 to 132.170.6.32. Terms of Use: http://spiedl.org/terms
 BARNARD, BOREMAN, and PAPE

Tw,
1

E
E
>-

I
Fig. 7 DMD array geometry for analysis.

U0(,i) = m3w'2 [comb(md, mdq) sinc(mw' ,mw'v)]

—15 —10 —5 0 5 10 15
X (mm)

Fig. 8 Contour plot of two-dimensional object-field spectrum and

**D2 sinc1mDJ—
• I [ sin(2a)1 F sin(2a) coherent—transfer-function cutoff for a 15-mm-diam aperture stop.
I, mDIi—
L_ L mX\/] L mX\/
+ mw2 sinc(mw,mwq) — mw'2 Calculations of crosstalk ratio for an JR scene projector using
a 128 x 128 DMD with pixel sizes of 50, 100, and 150 m
. I , [ sin(2c)1 [ sin(2a) were made. Discrete wavelengths of 3.0, 4.0, and 5.0 im
Xsinclmw I— I, mw Iq— (11)
L L mX\/Tj L mX\/ were used for the coherent analysis. A contour plot of the
two-dimensional object field spectrum and cutoff spatial fre-
Equation (1 1) consists of three terms corresponding to the quency of the coherent transfer function for the case of 50- im
terms in Eq. (5). The first term produces the overall diffrac- pixels and a wavelength of 3.0 im is shown in Fig. 8. The
tion pattern of the DMD. It consists of a two-dimensional axes of the plot are given in terms of aperture stop dimensions
array of diffraction orders with the zeroth order located off so a direct relationship between cutoff frequency and aperture
axis. The magnitude of the diffraction orders are weighted stop size can be made. In this plot, the effects of the three
by a two-dimensional sinc function corresponding to an un- terms that contribute to Eq. (1 1) can be seen. The diffraction
tilted pixel. The second term is an on-axis diffraction pattern orders produced by the first term are the array of dots in the
that results from the single tilted pixel. The last term is the figure. The zeroth order is located at (12.1 mm, 12.1 mm).
subtracted diffraction pattern of the single untilted pixel. It is not evident from this plot, but the orders along
The coherent optical-transfer function is given by the exit x = 12. 1 mm and y = 12. 1 mm have a much larger magnitude
pupil of the imaging system, scaled by the appropriate pa- than the other diffraction orders shown. The two-dimensional
rameters, as sinc function centered about the origin is the second term in
Eq. (1 1). The subtracted sinc function is shown centered
about the zeroth diffraction order. The circular cutoff fre-
O2+12)½<- quency ofthe coherenttransferfunction for a 15-mm aperture
:' stop is overlayed on the plot. The best crosstalk ratio per-
(22) (12) formance will occur when the transfer function passes as
2k!2
much of the spectrum of the single on pixel while blocking
0, (2+2)½>
2Xf2
out the strong diffraction orders caused by the periodic DMD
array.
An upper limit on aperture stop diameter is determined
Multiplying Eqs. (1 1) and (12) we obtain the image-field by the location of the large-magnitude diffraction orders in
spectrum. A subsequent magnitude-squared inverse trans- the object spectrum. Using Eqs. (1 1) and (12), this limit is
form yields the image-inadiance function. To simplify cal- found from the relation
culation of the inverse transform, only the x axis slice of the
two-dimensional image is calculated. The x axis image slice
is then used to determine crosstalk ratio along the x direction.
D max sin(2i) = (14)
The appropriate Fourier-transform pair for this computation 2Xf2 mXVi
is given by Rewriting Eq. (14) and using m =f2/f1 gives the maximum
aperture stop diameter as
d (13)
D max f1V sin(2a) . (15)

144/OPTICAL ENGINEERING /January 1994/Vol. 33 No. 1

Downloaded from SPIE Digital Library on 23 Nov 2010 to 132.170.6.32. Terms of Use: http://spiedl.org/terms
 CROSSTALK MODEL OF A DEFORMABLE-MIRROR-BASED IR SCENE PROJECTOR

For an aperture stop diameter less than that given in Eq. (15), magnitude diffraction orders in the object field spectrum.
the zeroth diffraction order and the other large-magnitude Also present in the crosstalk ratio plots is a periodic trend in
diffraction orders will be blocked. the crosstalk ratio as the aperture stop is increased. This is
For the case of a 15-mm aperture, the spectrum of the caused by the lobes in the sinc pattern of the on pixel. De-
image-field profile along the x axis was calculated according pending on where the transfer function cuts through the lobes,
to Eq. (13) and is shown in Fig. 9. Taking the magnitude the crosstalk performance is improved or degraded.
squared of the inverse Fourier transform of Fig. 9 gives the Based on the analysis in Sec. 7, a blackbody source used
image-irradiance function along the x axis in Fig. 10. In this in the projection system can be modeled by a coherent ap-
case, the DMD pixel size was 50 pm, and the total flux was proach. The image irradiance is calculated at many discrete
calculated over the on pixel width and over an adjacent off wavelengths within the passband of the source and then
pixel width. An area fill factor of 95% was assumed for the summed to give the total irradiance. It is necessary that
DMD pixels. The crosstalk ratio given by the ratio of the two enough wavelengths are used to ensure validity of the quasi-
flux values was 170: 1 for this particular case. A plot of cross- monochromatic conditions over each small wavelength in-
talk ratio versus aperture stop size is shown in Fig. 1 1 for terval. In the polychromatic analysis, we used a blackbody
wavelengths of3.O, 4.0, and 5.0 pm. This figure demonstrates of T= 1900 K, with an exitance given by Planck's equation.
that crosstalk ratios greater than 200: 1 can be achieved over A linear spectral responsivity given by
the wavelength band. Crosstalk ratio plots for DMD pixel
sizes of 100 and 150 m are shown in Figs. 12 and 13. A x
dramatic improvement in crosstalk ratio is possible as the (16)
DMD pixel size increases. From the crosstalk ratio plots, we
can see that the crosstalk ratio generally increases until a where X = 5 rim, was assumed for the detector focal plane
point where it rapidly falls off. Referring to Eq. (15), this fall array. A wavelength interval of 0. 1 m was chosen and 20
off occurs when the transfer function begins to pass the large- image irradiance samples over the 3- to 5-rim band were

0
U

—60 —40 —20 0 20 40 60 0 5 10 15 20 25 30

(cyc/mm) Aperture Stop Dometer (mm)

Fig. 9 Spectrum of monochromatic image field along the x axis for Fig. 1 1 Crosstalk ratio as a function of aperture stop diameter and
a system with a 15-mm-diam aperture stop. wavelength for a 50-p.m DMD pixel size.

:
0.8 -

0.6 0

0.4 0
U
"on'• Pixel • Off Pixe'

0.2

nfl
—150 —100
. —50 0 50 100 150 10 15 20 25 30
X(jm) Aperture Stop Diameter (mm)

Fig. 1 0 Monochromatic image irradiance along the x axis for a sys- Fig.12 Crosstalk ratio as a function of aperture stop diameter and
tem with a 15-mm-diam aperture stop. wavelength for a 1OO-m DMD pixel size.

OPTICAL ENGINEERING /January 1994 /VoI. 33 No. 1 I 145

Downloaded from SPIE Digital Library on 23 Nov 2010 to 132.170.6.32. Terms of Use: http://spiedl.org/terms
 BARNARD, BOREMAN, and PAPE

0.8
\iw
0.6 -

0.4 ' -
00
On" PxeI Off" Pixel

0.2 -

O.l
—300 —200
I
—100 0
I.,,, I .i 100 200 300
10 15 20 X(jm)
Aperture Stop Diometer (mm)

Fig. 13 Crosstalk ratio as a function of aperture stop diameter and

0.01(j 'l
wavelength for a 150-p.m DMD pixel size.
0.008 • -

0.006 On' Pixel 'Off' Pixel .

0.004 -

0.002 -

0.oon
—300
....

—200 —100 0 100

L
200 300
x (jim)

j
Fig. 15 Total image irradiance along the x axis for a 1OO-m DMD
pixel size and a 23-mm diam aperture stop under polychromatic il-
—150 —100 —50 0 50 100 150 lumination.
x (&m)

0.010 '- calculated crosstalk ratios for these cases are 260: 1 , 1 200: 1,
and 2400:1.
0.008 - .
5 Conclusions
0ff" Pxel .
A DMD-based JR scene projector is capable of high-contrast
0.006 - On" Pixel
projection over the 3- to 5-lim band. The feasibility of using

L
a DMD in the projector design was verified in a proof-of-
principle experiment. Because of diffraction, the best cross-
talk performance occurs at shorter wavelengths and larger
DMD pixel sizes. For projection over the 3- to 5-m band,
a DMD pixel size of approximately 100 im gave a crosstalk
ratio in excess of 1000:1.
The particular projector design presented in this paper was
—150 —100 —50 0 50 I 00 150 based on the assumption that the optical system of the sensor
X(jm)
under test is unobscured, as was the optical system in the
Fig. 14 Total image irradiance along the x axis for a 50-pm DMD proof-of-principle experiment. Some JR sensor systems use
pixel size and a 23-mm diam aperture stop under polychromatic il- a Cassegrain objective, which has a central obscuration. A
lumination.
modified projector design that uses a pupil diameter smaller
than the central lobe of the DMD diffraction pattern has been
included in the total image-irradiance function. For this case investigated14 for this class of systems.
v/v 30 over the passband. The previous coherent analysis
indicates that the best crosstalk performance is obtained when 6 Appendix A: Derivation of Equation (1)
the transfer function just excludes the large-magnitude dif- The coherence area of the source illumination at the object
fraction orders in the object-field spectrum. Using this cri- plane is found by propagating the MIF from the source to
tenon, we assumed a fixed aperture stop diameter of 23 mm. the object plane.'2 The illumination system is shown in
The total image-irradiance functions for DMD pixel sizes of Fig. 17 with axes labels at the source, lens, and object planes.
50, 100, and 150 pm are shown in Figs. 14, 15, and 16. The In this analysis, the source is assumed to be quasimono-

146/OPTICAL ENGINEERING/January 1994/Vol. 33 No. 1

Downloaded from SPIE Digital Library on 23 Nov 2010 to 132.170.6.32. Terms of Use: http://spiedl.org/terms
 CROSSTALK MODEL OF A DEFORMABLE-MIRROR-BASED IR SCENE PROJECTOR

1. Collimating
Lens
(x,y) Object
(1,)
0.8 Source
(a,)
(\
0.6

0.4
On' Pixel "Off" Pixel
f z

0.2 Fig. 17 Illumination system.

0.( I

—400 —200 0 200 400

Xm)
length of the quasimonochromatic radiation. The c subscript
0.010 -. -. indicates collimator-lens plane parameters. For a lens aper-
ture function P(x,y), the MIF transmitted by the lens is given
0.008 ' - by

J(x1 'Yi ; x2,y2)=P(x1 ,y1)P(x2,y2)

0.006 - "On" Pixel "Off' Pixel .

x +y) - (x
I' 0.004 - . exP{_J[(x +)J}
XJ(x1 'Yi ;x2,y2) . (19)
0.002 .
Substituting Eq. (18) into Eq. (19) and evaluating the Fourier
transform integral results in
—400 —200 0 200 400
X(im) - /Ix y\
J(x1,y1;x2,y2)=P(xi,yi)P(x2,y2)I (20)
Fig. 16 Total image irradiance along the x axis for a 15O-tm DM0
pixel size and a 23-mm diam aperture stop under polychromatic il-
lumination. -
where I is the two-dimensional spatial Fourier transform of
the source irradiance function. The source is assumed to sub-
tend a large enough angle at the lens that the source coherence
chromatic and spatially incoherent. Small angles are assumed area at the lens is much smaller than the area of the lens.
throughout and a thin-lens approximation is used. Ratherthan Under this condition, J will be nonzero only for small values
a coherence-area figure of merit, the coherence parameter is

(
of x and zy, and we can approximate the MIF as
given as the maximum separation of two points in the object
plane.
Ignoring constant multipliers, the incoherent-source MIF _ IP(j) I27
J'(x1 'Yi ; x2,y2) (21)
is given by S\Xf XfJ
J5(a1 , ;a2, I32)=J(ai , 31)(a1 — a2 —
132) (17) where =(x1 +x2)/2 and = (y +y2)/2. Using the Van
Cittert-Zemike theorem again to propagate the MIF leaving
where I is the source irradiance function; a and 3 are the the lens over a distance z to the object plane gives the object-
source coordinates at points P1 or P2, depending on the sub- plane MIF as
script used; and is a Dirac delta function. The s subscript
indicates source plane parameters. Using the Van Cittert-
Zernike theorem to propagate the source MIF over a distance
I gives the MIF at the plane of the collimating lens as
0(1
J(xi 'Yi x2 'Y2) = exP{_J[(x + y) — (x + xJJ IP()I2 exp[J?(+5)] d d

XJJIs(a
) exP[J(ixa + LY)] da d, oo_fz,xzy\ F2ir
expLi(%x+
y)
—

(18) XdLxdiy . (22)

where tx =x2 —x1 and iy =Y2 Yi' and X is the mean wave- The o subscript on the MIF indicates that this is an object

OPTICAL ENGINEERING I January 1 994 I Vol. 33 No. 1 I 147

Downloaded from SPIE Digital Library on 23 Nov 2010 to 132.170.6.32. Terms of Use: http://spiedl.org/terms
 BARNARD, BOREMAN, and PAPE

plane parameter. In arrivingat Eq. (22), the assumption that 3.81 f,D
< X/4 and iXy/z < X/4 was made so that an extra z> (29)
phase term could be eliminated. This approximation can be
recast as This condition is the same as that given in Eq. (1).

z>2_
DsDca
(23)
7 Appendix B: Quasimonochromatic
x Approximation of Polychromatic Mutual-
Coherence Function
where D is the maximum dimension of the source and Dca
is the maximum dimension ofthe coherence area ofthe source We begin with the expression for the mutual-coherence func-
at the lens. This requirement can be written in terms of the tion of an extended incoherent source, not necessarily quasi-
focal length of the collimating lens as monochromatic, derived by Born and Wolf.'5 The source
geometry is given in Fig. 1 8 and shows a small source element
8f , dS of the extended source and an observation plane where
z>- (24) the mutual coherence between two points P1 and P2 is to be
calculated. From Section 10.6. 1 , Eq. (7) of Ref. 15, the
by using relations between the defined coherence area and mutual-coherence function at the observation plane is given
the area of the source. by
Returning to Eq. (22), we see that two Fourier transforms
are contained in the expression. The transform of the lens
aperture is a narrow function in (z y z ) compared to the F(P1 ,P2,T) = J dv exp(—j2iivT)
transform that produces a scaled source irradiance function
in (, ). This means that the lens aperture determines the 1
coherence area at the object plane and the source determines x I(S,v) exp[jk(R,—R2)] dS , (30)
the average irradiance distribution over the object plane. Jcr R1R2
The lens aperture function can be written as
where I(S, v) is the spectral exitance of the source and R1 and
R2 are the distances from the source point dS and points P,
O(x2+y2) 2< and P2. The spectral characteristics of the source are ap-
1,
proximated as a sum of narrow rect functions over the ap-
2 2½ propriate passband as
, (25)
PC(xY)cYl[ D ]= (x2+y2)Y2=
"v—V\
0,
yD
(x2+y2) 2> I(S, v) I(S,V)—
Lvrect(
\LV/—a)
1
(31)

where D is the diameter of the collimating lens aperture. where z v is the bandwidth and i is the center frequency of
Substituting Eq. (25) into Eq. (22) and ignoring the phase the spectral bands. Substituting Eq. (31) into Eq. (30) and
term and the average irradiance term gives the MIF at the rewriting the wave number k in terms of temporal frequency
object plane as gives

J (Y J1[(rrD/Kz)(z2 + 2)'/2]
2)2
(irD/Kz) (y2 + 2) /2
(26) rect(n)
where J1 is the first-order Bessel function of the first kind.
If we assume full coherence for 2J,(r)/rO.88, then two x VVR, _R2)] exp(—j2vT) dv
exp[12
points in the object plane with a separation given by

s irDXz
(27) xfI(S,Vfl)exPU(R)(1
—R2)]
dS . (32)

Using the quasimonochromatic assumption that i >> zv and

will be coherently illuminated.'5 The radius of the first zero R, — R2 <<c/v for each of the narrow rect spectral bands,
of the diffraction-limited resolution spot of the imaging optics
at the object plane is given by
1.22Xf1
(28)
D,
where f1 is the focal length of relay lens 1 of the imaging
system and D is the diameter of the aperture stop located at
relay lens 2. The condition for approximately coherent im-
aging then becomes Fig. 18 Extended source and observation plane.

148 / OPTICAL ENGINEERING /January 1994 / Vol. 33 No. 1

Downloaded from SPIE Digital Library on 23 Nov 2010 to 132.170.6.32. Terms of Use: http://spiedl.org/terms
 CROSSTALK MODEL OF A DEFORMABLE-MIRROR-BASED IR SCENE PROJECTOR

the exponential term exp{ —j2'rr{(v — — R2)} in the generation technology: calibration to functional testing,' ' Proc. SPIE
first integral can be approximated as unity. Evaluating the 1050, 130—136 (1989).
6. L. J. Hornbeck, ' '128 x 128 deformable mirror device,' ' IEEE Trans.
resulting Fourier transform integral and using the Van Cittert- E!ectron Devices ED-30, 539—545 (1983).
Zernike theorem to replace the integral over the source area 7. D. R. Pape and L. J. Hornbeck, "Characteristics ofthe deformable mir-
ror device for optical information processing,' ' Opt. Eng. 22, 675—68 1
by the MIF gives (1983).
8. W. R. Wu, R. 0. Gale, L. J. Hornbeck, and J. B. Sampsell, "Electro

F(P1 , P2 ,T) sinc(z VT) J( P1, P2 , i) exp( —j2'rri5T) optical performance of an improved deformable mirror device,' ' Proc.
SP1E825, 24—31 (1987).
('2l\ 9. L. J. Hornbeck, ' 'Deformable-mirror spatial light modulators,' ' Proc.
SPIE 1150, 86—102 (1989).
10. R. M. Boysel, ' 'Deformable mirror light modulators for image pro-
where J (P , P2 , V,) is the quasimonochromatic MIF at a fre- '
cessing,' Proc. SPIE 1151, 183—194 (1989).
quency of V,. Because the quasimonochromatic assumption 11. B. J. Thompson, "Image formation with partially coherent light,"
Chap. IV in Progress in Optics, Vol. III, E. Wolf, Ed., pp. 169—230,
also includes L VT << 1, Eq. (33) can be further simplified North Holland Publishing Company, Amsterdam (1969).
to 12. J. W. Goodman, Statistica! Optics, pp. 286—324, Wiley, New York
(1985).
1' A\ 13. J. D. Gaskill, Linear Systems, Fourier Transforms, and Optics, Wiley,
F(P1,P2,T)J(P1,P2,) exp(—j2TrT) 3'+) New York (1978).
14. J. A. Carter III and D. R. Pape, submitted to Opt. Eng.
15. M. Born and E. Wolf, Principles ofOptics, 6th ed., pp. 51 1, 532—534,
In general, the mutual-coherence function under the quasi- Pergamon Press, New York (1989).
monochromatic assumption is given as'2
Kenneth J. Barnard received the BS de-
F(P,,P2,T)J(P1,P2) exp(—j2ITVT) (35) gree in electrical engineering from the Uni-
versity of Michigan in 1986 and MS and
where V is the mean frequency of the radiation and J ( P,, PhD degrees in electrical engineering from
P2) is the MIF. Comparing Eqs. (34) and (35), the poly- the University of Central Florida in 1990
chromatic mutual-coherence function can be written as a and 1992. Dr. Barnard has been at Mem-
phis State University since 1992 and is
summation of independent quasimonochromatic mutual- currently an assistant professor of electri-
coherence functions over some passband. In partially coher- cal engineering. His current research in-
ent imaging applications the time dependence T, appearing terests include infrared scene projection,
in Eq. (35), is zero and the MIF is propagated through the optical processing, and holography. He is
a member of SPIE, OSA, and IEEE.
system to calculate image properties. For polychromatic im-
aging, each of the independent MIFs in Eq. (34) produces a
separate image-irradiance function. The total irradiance func- Glenn D. Boreman is an associate pro-
fessor of electrical engineering in the Cen-
tion is then given as a summation of individual irradiance ter for Research in Electro-Optics and Las-
functions as ers (CREOL) at the University of Central
Florida. He received a BS from the Insti-
IT(P)(Pfl)J(1°i,P2,l)fp1=p2=p (36) tute of Optics, University of Rochester, and
a PhD from the Optical Sciences Center,
University of Arizona. He has held visiting
where I is the quasimonochromatic image irradiance, 'T 5 research positions at IT&T, Texas Instru-
the total irradiance, and P represents the coordinates of the ments, U.S. Army Night Vision Lab, and
McDonnell Douglas. Dr. Boreman serves
image plane. as topical editor for Applied Optics in the areas of radiometry and
detectors and is past-president of the Florida Optical Society.
Acknowledgments
This work was supported by Photonic Systems, Incorporated, Dennis R. Pape received a BA degree in
physics in 1974 from Cornell University
Melbourne, Florida, through a Small Business Innovative and a PhD in physics in 1980 from Duke
Research (SBIR) subcontract from the Naval Weapons Cen- University. From 1980 to 1984 he was a
ter, China Lake, California (Paul Amundson, Contracting member of the technical staff in the Optical
Officer' s Technical Representative), and by the Florida High Information Processing Branch at Texas
Technology and Industry Council. The authors wish to thank Instruments' Central Research Laborato-
ries in Dallas, Texas, where he designed
Texas Instruments and in particular Jim Florence, for the and developed an integrated silicon-based
opportunity to use their facilities and DMD in Dallas, Texas optically addressed membrane spatial light
to construct the proof-of-principle projector described in this modulator. In 1984 he became the group
paper. leader of the Acousto-Optic Device Development Group at Harris
Corporation. At Harris he was responsible for the development and
application of advanced acousto-optic devices for military optical sig-
References nal processing systems. In 1987 he founded Photonic Systems In-
corporated (PSI) in Melbourne, Florida, where he is the president.
1. W. E. Woehi, "Transfer requirements for target radiance simulation in
He is the author of two book chapters and over 25 papers in the
optical projection simulators," App!. Opt. 14, 124-128 (1975). field of optical information processing. He is chairman of the 1992
2. A. P. Pritchard, "Dynamic JR scene generation: basic requirements and
comparative display device design," Proc. SPIE 940, 144—149 (1988). SPIE meeting on Advances in Optical Information Processing, a
3. T. Barnett, "JR transducer technology: an overview," Proc. SPIE 765, member of the technical program committee for the 1992 IEEE Ul-
82—84 (1987). trasonics Symposium, and past president of the Florida Section of
4. J. L. Hester, "Thermal array target simulation technology," Proc. SPIE the OSA. He is also a member of the APS, the Association of Old
940, 150—152 (1988). Crows, and the Air Force Communications and Electronics Associ-
5. D. E. Parker, D. G. Taylor, and C. T. Wallace, Jr., "Trends in scene ation.

OPTICAL ENGINEERING /January 1994/Vol. 33 No. 1 / 149

Downloaded from SPIE Digital Library on 23 Nov 2010 to 132.170.6.32. Terms of Use: http://spiedl.org/terms