Phase calibration and applications

of a liquid-crystal spatial light modulator

Alain Bergeron, Jonny Gauvin, François Gagnon, Denis Gingras, Henri H. Arsenault, and
Michel Doucet

A simple phase-characterization method for spatial light modulators is proposed. The low-cost method
permits high-precision measurement and provides data for the setting of the spatial-light-modulator
operating point in the phase-modulation mode. The dynamic phase response is used to perform efficient
kinoform recording. In order to record the kinoform, we modify the global iterative coding to compute
phase holograms. Finally, modified phase–phase correlation is introduced. The phase–phase correla-
tor permits sharper correlation peaks, better energy transmission, and higher discrimination than an
amplitude–phase correlation. Optical experimental results are presented.
Key words: Phase-only filter, interferometric phase measurement, phase–phase correlation, kino-
form, spatial light modulators, pattern recognition.

1. Introduction
Since the work of Horner on phase-only filters,1,2
phase processing has been widely used in optical
correlator architectures. Phase-only filters yield
sharper correlation peaks, higher light efficiency, and
better discrimination capabilities than matched fil-
ters.2–4 This interest has increased with the develop-
ment of low-cost spatial light modulators 1SLM’s2 that
permit phase pattern to be displayed at video frame
rates. Such SLM’s offer constantly improving perfor-
mance and higher resolutions. These performances
open the way to reprogrammable optical architec-
tures dedicated to artificial vision and information
processing.
One of the challenges imposed by phase SLM’s is
the precise characterization of the dynamic phase
response. The knowledge of the SLM phase re-
sponse permits better control of the filtering opera-
tions, which leads to experimental results closer to
theoretical predictions. Much work has been done
A. Bergeron, J. Gauvin, F. Gagnon, D. Gingras, and Michel
Doucet are with the Institut National d’Optique, 369 Franquet,
Sainte-Foy, Québec G1P 4N8, Canada. H. H. Arsenault is with
the Université Laval, Centre d’optique, photonique et laser, Sainte-
Foy, Québec G1K 7P4, Canada; A. Bergeron is also with the
Université Laval.
Received 20 April 1994; revised manuscript received 13 February
1995.
0003-6935@95@235133-07$06.00@0.
r 1995 Optical Society of America.
on SLM phase characterization.6–10 A Mach–Zehn-
der interferometer is often used to measure the SLM
phase response. However, this method is sensitive
to environmental conditions and usually gives only a
rough idea of the real behavior of the device. In this
paper a simple low-cost method showing high-
precision measurement capability is proposed. This
method is based on Young’s fringes and permits a
rapid selection of the operational conditions of the
SLM.
One of the applications suitable for a phase SLM is
kinoform recording. The results obtained from phase
characterization can be used to compute such kino-
forms10 and improve the results. However, in order
to compensate for the absence of amplitude modula-
tion, one needs an iterative kinoform-coding approach.
We recently proposed a global iterative coding11 algo-
rithm 1GIC2 as a tool for the generation of amplitude
holograms. This method has been modified in order
to code kinoforms as well. Some experimental re-
sults with gray-level objects are presented.
Finally, a simple method to perform modified phase–
phase correlation is proposed in Section 6. Many
experiments have been reported on the phase-only
filters. Correlations that involve phase are gener-
ally useful in improving the detection of energy in the
output plane.12–15 Some researchers also proposed
quasi-phase correlation architectures.4 The modi-
fied phase–phase method permits better discrimina-
tion and higher energy transmission than phase-only

10 August 1995 @ Vol. 34, No. 23 @ APPLIED OPTICS 5133

filters alone. Numerical simulations and optical re- output will have the form
sults are presented.
1 Dx 2 1@2

2. Interferometric Phase Characterization

f1x, y2 5 p52a31 x2 4 6
2 2 1 y2

The knowledge of the precise dynamical response of a 1 2 1@2

1 pA 531x 1 2 1 x 4 6B .
Dx
2
SLM permits better control of its properties. For 112
2a 2
SLM characterization, most of the research presented
until now has used a Mach–Zehnder interferome- If
ter.5,10,14 This architecture shows sensitivity to align-
ment and to environmental conditions. Further- r 5 1x2 1 y221@2, 122
more, it often needs precise optical components, which
can be costly. A simpler method that uses diffraction then p is defined as
gratings has also been proposed16; however, it re-
quires highly precise Ronchi components. For these
reasons we propose an alternative configuration, which
p1r@2a2 5 1, r , a, 132

permits exact mathematical relations, to achieve the

measurement of a SLM phase response. This method
shows good stability by using a common-path interfer-
ometer. Figure 1 presents the setup used for the
SLM characterization.
In Fig. 1 the laser beam is expanded and collimated.
The SLM is inserted between the collimated beam
and a lens that performs a Fourier transform. A
CCD camera located in the Fourier plane is used to
measure the phase modulation. The method takes
advantage of a mask with two pinholes placed in the
p1r@2a2 5 0, r . a. 142
In Eq. 112, 2a is the diameter of the pinholes and Dx is
the distance between the holes.
The input wave front is multiplied by the SLM
transmittance pattern, so each hole will observe a
different phase retardation. However, this phase
difference may come with a residual amplitude modu-
lation. The wave front is therefore expressed by
1 Dx 2 1@2

1 2 2p52a31 2 2 4 6
2jf
front of the SLM. The two pinholes have to be as fm1x, y2 5 A exp x2 1 y2
similar as possible in order to obtain the best contrast
in the fringe pattern. The SLM is separated into two jf 1 2 1@2
1 1A 1 dA2exp1 2pA 531x 1 2 4 6B ,
Dx
distinct regions, and each is multiplied by a single
2 2a 2
pinhole. One of the regions is addressed to a uniform
constant value 1voltage2 and is used to modulate the 152
reference beam. On the second region a different
modulation voltage is applied, which can be modified where A is the transmission of the SLM and dA takes
to characterize the phase of the SLM in the required into account the residual amplitude modulation.
dynamic range. Both signals are uniform over the In the Fourier plane the intensity pattern detected is
pinhole surface. When the two regions exhibit the expressed by
same value, the Fourier transform shows a cosinusoi-
a2
dal function modulated by an Airy spot. If the phase
value of one region is varied, the cosinusoidal function
will translate according to the phase difference be-
0F1yx, yy202 5 12 3
3J 12pyra2423 1dA22 1 14A2 1 4AdA2
r 1

1 24 ,
tween both regions. f
The mathematical analysis of the setup is simple 3 cos2 pyxDx 1 162
2
and allows us to see how the fringe pattern may help
to set the SLM operation point. Let a uniform beam where
impinge upon the pinhole mask. The modulated
yx 5 xn@lf. 172

The parameter f is the relative phase transmittance

Fig. 1. Interferometric setup for phase measurement:
liquid-crystal television; FT, Fourier transform.
of one hole with respect to the other. The phase
varies with the voltage applied to each pixel. The
detected signal is composed of a dc level superimposed
on a cosine function. This signal is modulated by an
Airy function. The phase modulation can be com-
puted from the translation of the cosine function.
Also, the phase factor is independent of the amplitude
modulation, the hole diameter, and the distance be-
tween the holes. The phase is computed with the
ratio of the pattern translation to the cosine period, so
LCTV, there is no need to know the exact diameter of the
holes. A difference in the holes’ diameters will sim-

5134 APPLIED OPTICS @ Vol. 34, No. 23 @ 10 August 1995

ply decrease the contrast of the fringe pattern. The
hole dimensions and distances will, however, affect
the fringe and envelope frequencies. These param-
eters can be modified to obtain a signal that is easier
to process. A change of the lens focal length may also
be used to modify the fringe pattern width in order to
obtain higher accuracy.
The residual amplitude modulation only affects the
contrast observed in the fringe pattern. In many
SLM’s the correct operation of the modulator is
achieved by means of a careful alignment of the
polarizers and a proper setting of the bias level of the
liquid-crystal television 1LCTV2. The maximization
of the fringe-pattern contrast will permit minimiza-
tion of the residual amplitude modulation dA, which
varies with the applied voltage.

3. Phase Measurement
The setup of Fig. 1 was built to characterize a LCTV
extracted from an Epson@Kodak LC500 video projector.
These LCTV’s are digitally addressed SLM’s that
could be addressed with an 8-bit dynamic range.
However, 5 bits were used for the experimental Fig. 2. Fringe patterns for gray levels 1a2 0, 1b2 10, 1c2 20, and 1d2 31.
purpose. The bias control level 1brightness2 was set
to zero to obtain the maximum phase shift. The
mask was made from a metal plate in which two holes
of the same diameter were drilled. The intercepted the fringe-pattern translation. The horizontal line
LCTV surface for each hole had an area of ,48,000 located above each fringe pattern is the least-square
µm2, which corresponds to ,7 pixels. The gray level, best fit of Eq. 162 performed on the current image.
the phase displayed in one region of the LCTV, was From these images it is clear that the fringe pattern is
varied between 0 and 31, and the other side was kept of good quality and that the measurement method has
constant. The polarizer was adjusted to obtain the sufficient sensitivity to detect the phase variations
maximum fringe-pattern translation for a level varia- generated by the SLM.
tion from 0 to 31. This, in turn, allowed us to operate Figure 3 shows the overall LCTV response for the
the SLM over the maximum phase range. In addi- 32 gray levels. These data clearly show the nonlin-
tion, some tests were made to obtain the maximum ear behavior of the LCTV. It is also of interest to
contrast of the fringe pattern over the different phase note the smoothness of the curve. This reflects the
robustness of the characterization method, i.e., its
values in order to operate the LCTV in the minimum-
low sensitivity to environmental conditions. The
amplitude-modulation mode. This was performed
precision depends on the resolution of the CCD
by choice of an initial orientation for the first polarizer.
camera and the focal length of the Fourier lens.
The second polarizer was then rotated to find the
orientation that permits the maximum pattern trans-
lation, that is, the maximum phase range. The first
polarizer was rotated again, followed by a full rotation
of the second one. At first, both rotation steps were
20° to identify the best range of orientations. Then
more precise iterations were used. The alignment of
the polarizer is, however, not critical, and a range of
610° variations in the polarizer orientation is usually
acceptable.
Once the setting point of the polarizers was deter-
mined, 32 fringe patterns, corresponding to the 32
gray levels available, were analyzed to extract the
LCTV phase response. The fringe patterns were
analyzed by use of a best-fit operation of the analytic
expression of Eq. 162 with the measured data along
the horizontal center line of the fringe patterns.
The patterns obtained for gray levels 0, 10, 20, and 31
are presented in Fig. 2. The vertical white line
indicates the reference and permits visualization of Fig. 3. Dynamic phase response of the LCTV.

10 August 1995 @ Vol. 34, No. 23 @ APPLIED OPTICS 5135

The setup uses the cosine period of 88 pixels, which
gives a basic precision on the phase value of ,4°.
However, the precision could be even greater because
a best-fit operation is performed. A camera that
could accurately measure an image with 256 gray
levels would permit an overall precision of ,1°. The
smallest change detected between two gray levels was
2.35°, whereas the largest change was 20.7°. The
total phase-modulation range was 2.34p rad. This
value depends on the LCTV used. As a consequence,
each device should be characterized individually be-
fore being used as a phase-modulation device.
Figure 4 shows the amplitude response for a LCTV
operated in the phase mode. Some residual ampli-
tude modulation remains even for a phase operation
of the SLM. This response corresponds to a varia-
tion of ,20% in the transmitted signal. Such behav-
ior can affect the performance of phase filters re-
corded on the LCTV. This architecture permits only
characterization of a small region of the device. The
uniformity of the phase response could easily be
investigated by use of cylindrical lenses combined
with two slits instead of pinholes. With the lenses
imaging in the direction along the slits and the
Fourier transform in the orthogonal direction the
phase response could be measured along a line across
the LCTV.
4. Kinoforms
One application of the LCTV phase operation is the
recording of a kinoform. A kinoform is a phase
recording of a wave front. It can also be seen as a
computer-generated phase hologram. We recently
introduced an algorithm to generate computer-
generated holograms of improved quality.11 Although
the global iterative coding algorithm 1GIC2 was de-
signed for amplitude computer-generated holograms,
it can be generalized to phase hologram computa-
tions.
Because the method is described in detail else-
where,11 only an outline of the method is given here.
Figure 5 represents the general algorithm of the GIC.
An input signal is Fourier transformed and coded into
Fig. 4. Amplitude response for a phase-operation mode.
5136 APPLIED OPTICS @ Vol. 34, No. 23 @ 10 August 1995
Fig. 5. Global iterative coding algorithm: CGH, computer-gener-
ated hologram.
a computer-generated hologram. The resulting re-
construction is analyzed, and the missing information
is computed. The difference between the required
output and the obtained output is then coded and
added directly to the previously generated kinoform.
At each iteration the wave-front correction to be
added is simply the phase content of the object
Fourier transform. The correction is applied to the
frequency-plane kinoform, not to the spatial plane.
After a few iterations, the quality of the reconstruc-
tion improves.
5. Kinoform Experiments
Some experiments were performed with different
objects to test the GIC coded in phase. The kino-
forms were computed to optimize the intensity of the
reconstruction; however, both intensity and phase
could be performed. The phase response curve and
the residual modulation amplitude of the LCTV were
taken into account at each iteration.10 In order to
reconstruct the kinoforms, we illuminated the LCTV
with a laser collimated beam. The phase-modulated
wave front was simply Fourier transformed with the
help of a lens. The reconstruction localized in the
zeroth order of diffraction was acquired by a CCD
camera. The results for images varying from 64 3
64 pixels to 128 3 128 pixels are presented in Fig. 6.
All kinoforms had 200 3 200 points. The results
clearly show the possibility of coding with the GIC an
intensity pattern with a filter that displays only the
phase information with limited resolution. Between
five and ten iterations were necessary to compute the
kinoforms. Note that no dc components are present
on the reconstruction; dc elimination was not possible
when the LCTV response was not taken into account
in the iterative process.
6. Phase–Phase Correlation
Matched filtering is a well-known application in pat-
tern recognition. Many matched filters yield large
correlation peaks and have poor discrimination capa-
bility.2–4 The phase-only filter has a high optical
efficiency together with an improved discrimination
capability.2 Other researchers have considered the
design of binary phase-only filters for implementa-

Fig. 6. Original input 1left2 and optical reconstruction 1right2 of a
kinoform: 1a2 fire symbol, 128 3 128 points; 1b2 space shuttle,
64 3 64 points.
tion on spatial light modulators.14 Some studies on
quasi-phase correlation were recently carried out to
achieve optical implementation of phase correlations.4
In this section we propose a simple method to perform
modified phase–phase correlation. The method takes
advantage of the phase-modulation capability of liquid-
crystal SLM’s. The basic architecture is an optical
correlator in which two SLM’s are inserted.
In phase-only correlation the SLM in the Fourier
domain is operated in phase. The filter displayed is
the phase of the object Fourier transform. The input
SLM is operated in amplitude, and the object is
displayed in a gray-level format. In modified phase–
phase correlation the input SLM is operated in phase.
Consequently the gray-level image becomes a phase
object:
2pwo1x, y2
wp1x, y2 5 exp j 53 Amax 46 ,
where wo is the object and wp is the phase object.
The input object is fully transmissive. Each point of
the input scene is represented by a phase that corre-
sponds to the normalized intensity of the input image
times the range of available phase, that is, 2p rad.
Because the phase operation of the LCTV is depen-
dent on only the polarizer orientation and the voltage
applied to the liquid-crystal cell, or the bias level,6 no
supplementary processing is necessary to display a
phase image on the SLM. The filter is also displayed
in phase but is computed from the input object phase
representation. The Fourier transform of the phase
object is expressed by
Wp1u, v2 5 Ap1u, v2exp3 jfp1u, v24 5 FT3wp1x, y24. 192
The corresponding coded filter is obtained with the
phase information of the Fourier transform:
Wc1u, v2 5 exp3 jfp1u, v24. 1102
The total energy detected is increased because the
object is represented by a unitary transmission level
instead of different gray levels.
7. Correlation Experiments
The first experiments that we performed were numeri-
cal simulations. The objects were nine gray-level
images of aircraft, typically with 64 3 64 pixels, as
shown in Fig. 7. Figure 81a2 shows the results of the
simulations. In the amplitude–phase mode the high-
est cross correlation is 16.9% of the maximum correla-
tion peak, whereas the phase–phase correlation de-
creases this value to 7.8%. The mean value of the
different cross correlations is 7.9% for the amplitude–
phase correlation, whereas it is 3.1% for the phase–
phase correlation. These results show that the phase-
–phase correlation not only improves the energy
detected at the output but also increases the discrimi-
nation capability of the phase filter.
In order to verify the feasibility of phase–phase
optical correlation, we used a standard correlator17
setup. Both LCTV’s were characterized and used in
the phase configuration. The filters were mapped
according to the phase dynamical response of the
LCTV. However, as the filter already exhibited uni-
tary amplitude over the whole plan, these phase
filters did not require the use of GIC coding. Figure
7 presents a scene with all the input images. The
shuttle was used as a target to compute the phase
filter. The results obtained with the optical correla-
tor are presented in Fig. 81b2. In these results the
maximum correlation value is 1.0 1for the shuttle2.
The top of the chart is truncated to permit better
visualization of the results.
In the amplitude–phase mode the maximum cross-
correlation peak is 29.8%, and the mean of the
different cross correlations is 12.7%. The phase–
phase operation mode allows us to decrease the
182
highest cross-correlation peak to 15.7%. The mean
Fig. 7. Gray-level airplane images for the correlation experi-
ments.
10 August 1995 @ Vol. 34, No. 23 @ APPLIED OPTICS 5137

Fig. 8. Correlation 1a2 simulations and 1b2 experiments: ampli-
tude–phase 1shaded2 and phase–phase 1white2.
of the cross correlations is decreased to 7.2%. Al-
though the optical results show cross-correlation val-
ues slightly higher than the simulated results, one
can see that the general behavior of the filter is
respected. It should be noted that this improvement
is valid for gray-level objects. Binary objects may
show a little variation from this behavior. However,
discrimination is already very high for binary objects,
so this type of correlation allows us to increase the
low-discrimination capabilities of gray-level objects
yet decrease slightly those of binary objects. The
improvement in discrimination is roughly a factor of
2. At these low energy levels, part of the variation
between the expected and the obtained results can be
attributed to the residual amplitude modulation, to
the contrast ratio 1which is typically 80:1 for this type
of LCTV2, and to the resampling of the video signal.
The amount of energy detected in the output plane
was measured and is improved by a factor of 3.36.
Fig. 9. Three-dimensional plot of the theoretical amplitude–
phase Space Shuttle correlation.
5138 APPLIED OPTICS @ Vol. 34, No. 23 @ 10 August 1995
Fig. 10. Three-dimensional plot of the theoretical phase–phase
Space Shuttle correlation.
The sharpness of the correlation peak is important
in many applications. An indication of this sharp-
ness is given by the peak-to-total-correlation-energy
ratio 1PCR2, defined as
0C10, 0202
PCR 5 . 1112
` `
ee
2` 2`
2
0C1x, y20 dxdy
An ideal correlation peak would give a PCR of 1.00
according to this definition. An evaluation of the
expected PCR for the shuttle correlation gives 0.048
for the amplitude–phase mode and 0.142 for the
phase–phase mode. The comparison between the
two autocorrelation peaks for the space shuttle is
shown by Figs. 9 and 10. The optical experiments
showed a PCR of 0.033 for the amplitude–phase mode
and of 0.069 for the phase–phase mode. Although
these values are not as high as the predicted ones, the
optical phase–phase correlation increased the value
of the PCR by a factor of 2.1.
8. Conclusion
The phase information plays an important role in
different spheres of optical computing. In order to
control the phase behavior of a SLM correctly, one
needs to achieve a good characterization. The
Young’s-fringes-derived characterization method
yields high precision and provides a simple way to set
the phase operation point of a SLM. This method is
not limited to LCTV’s but can be applied to most of the
phase SLM’s. Such an architecture can be realized
at low cost and permits the implementation of kino-
forms on LCTV’s. The GIC is suitable for generating
such kinoforms. Finally, the phase–phase modified
correlation presented in this paper increases the
transmitted energy, provides a better discrimination,
and yields a better PCR. These results were con-
firmed by means of optical experiments.
This research was supported by grants from the
Natural Sciences and Engineering Research Council
of Canada, the Fonds pour la formation des cher-
cheurs et l’aide à la recherche program of Québec, and
the Japan Science and Technology Fund program of
the Canadian Ministry of Foreign Affairs.
References
1. J. L. Horner, ‘‘Light utilization in optical correlators,’’ Appl.
Opt. 21, 4511–4514 119822.
2. J. L. Horner and P. D. Gianino, ‘‘Phase-only matched filtering,’’
Appl. Opt. 23, 812–816 119842.
3. O. K. Ersoy, Y. Yoon, N. Keshava, and D. Zimmerman, ‘‘Nonlin-
ear matched filtering II,’’ Opt. Eng. 29, 1002–1012 119902.
4. K. Chalasinka-Macukow and C. Gorecki, ‘‘Optoelectronic imple-
mentation of the quasi-phase correlator,’’ Opt. Commun. 93,
11–18 119922.
5. J. C. Kirsch, D. A Gregory, M. W. Thie, and B. K. Jones,
‘‘Modulation characteristics of the Epson liquid crystal televi-
sion,’’ Opt. Eng. 31, 963–970 119922.
6. N. Konforti, E. Marom, and S. T. Wu, ‘‘Phase-only modulation
with twisted nematic liquid-crystal spatial light modulator,’’
Opt. Lett. 13, 251–253 119882.
7. J. C. Kirsch, J. A. Loudin, and D. A. Gregory, ‘‘Hybrid
modulation properties of the Epson LCTV,’’ in Wave Propaga-
tion and Scattering in Varied Media II, V. R. Varadan, ed.,
Proc. Soc. Photo-Opt. Instrum. Eng. 1558, 432–441 119912.
8. D. A. Gregory, J. C. Kirsch, and E. C. Tam, ‘‘Full complex
modulation using liquid-crystal televisions,’’ Appl. Opt. 31,
163–165 119922.
9. S. S. Coturia, S. E. Monroe, and J. Knopp, ‘‘A live input, live
filter, liquid crystal correlator,’’ in Advances in Optical Informa-
tion Processing V, D. R. Pape, ed., Proc. Soc. Photo-Opt.
Instrum. Eng. 1704, 248–256 119922.
10. J. Amako and T. Sonehara, ‘‘Kinoform using an electrically
controlled birefringent liquid-crystal spatial light modulator,’’
Appl. Opt. 30, 4622–4628 119912.
11. A. Bergeron, H. H. Arsenault, J. Gauvin, and D. Gingras,
‘‘Computer-generated holograms improved by a global itera-
tive coding,’’ Opt. Eng. 32, 2216–2226 119932.
12. J. L. Horner, ‘‘Clarification of Horner efficiency,’’ Appl. Opt. 31,
4629–4630 119922.
13. T. H. Barnes, K. Matsumoto, T. Eiju, K. Matsuda, and N.
Ooyama, ‘‘The application of phase-only filters to optical
interconnects and pattern recognition,’’ J. Mod. Opt. 37,
1849–1863 119902.
14. J. L. Horner and J. R. Leger, ‘‘Pattern recognition with binary
phase-only filters,’’ Appl. Opt. 24, 609–611 119852.
15. G. Paul-Hus and Y. Sheng, ‘‘Optical on-axis real-time phase-
dominant correlator using liquid crystal television,’’ Opt. Eng.
32, 2165–2172 119932.
16. C. Soutar, S. E. Monroe, and J. Knopp, ‘‘Complex characteriza-
tion of the Epson liquid crystal television,’’ Optical Pattern
Recognition IV, D. P. Casasent, ed., Proc. Soc. Photo-Opt.
Instrum. Eng. 1959, 269–283 119932.
17. A. VanderLugt, ‘‘Signal detection by complex filtering,’’ IEEE
Trans. Inf. Theory IT-10, 139–145 119642.
10 August 1995 @ Vol. 34, No. 23 @ APPLIED OPTICS 5139