1228
`
`OPTICS LETTERS I Vol. 17, No. 17 I September 1, 1992
`
`Experiments on a multichannel holographic optical switch with
`the use of a liquid-crystal display
`
`Hirofumi Yamazaki and Masayasu Yamaguchi
`
`NTT Communication Switching Laboratories, 3-9-11, Midori-cho, Musashino-shi, Tokyo 180, Japan
`
`Received March 16, 1992
`A new configuration of holographic switches is proposed and verified for multichannel optical switching. Ex(cid:173)
`perimental! X 64 and 2 X 32 switching is achieved by using real-time binary phase-only holograms generated
`by a twisted nematic liquid-crystal display. This holographic free-space switching is applicable to photonic
`switching systems and optical interconnections.
`
`Most N X N optical switches, whether they are
`waveguide types 1 or free-space types, 2 consist of
`multiple 2 X 2 switches. When large-scale switches
`are made, the cascade connections of many 2 X 2
`switches cause the accumulation of loss and cross
`talk. On the other hand, holographic N X N opti(cid:173)
`cal switches are constructed from 1 X N switches
`arranged in parallel and do not need cascade con(cid:173)
`nections. Therefore they have no accumulation of
`loss and cross talk and thus are suitable for large(cid:173)
`scale switches.
`Holographic switches that use photorefractive
`materials or acousto-optic Bragg cells have been
`4 These switches, however, have
`demonstrated. 3
`•
`drawbacks. Generation of real-time holograms
`with high diffraction efficiency and fast speed in
`photorefractive materials generally requires high
`voltages (of the order of a kilovolt). 5 Acousto-optic
`Bragg cells can steer light beams only one dimen(cid:173)
`sionally and require cascade connections for two(cid:173)
`dimensional beam steering. On the other hand,
`liquid-crystal displays (LCD's) can renew holograms
`with low driving voltages (5 V in our experiments)
`and can steer light beams two dimensionally. But
`the LCD's also have a drawback, i.e., low spatial
`resolution owing to the pixel size, which results in a
`smaller angle for beam steering. The pixel size is,
`however, expected to be reduced to an acceptable
`size (e.g., 10 p,m) in the future. We previously
`demonstrated 4 X 4 holographic optical switching
`by using an LCD as a binary phase-only modula(cid:173)
`tor.6 In this Letter a new configuration of holo(cid:173)
`graphic switches is proposed for easy implementation
`of multichannel switching, and 1 X 64 and 2 X 32
`optical switching is demonstrated experimentally.
`Figure 1 shows a schematic diagram of the pro(cid:173)
`posed switch. The LCD is divided into sections,
`and each section is assigned to a corresponding in(cid:173)
`put port. Holograms are calculated and stored in
`the memory of a controller before the switch is oper(cid:173)
`ated. A hologram is written independently on each
`section with electrical signals from the controller.
`A phase hologrmn on each section bends an input
`beam in the desired direction, and the lens feeds the
`deflected beams at the same angle to the correspond-
`
`ing output port. A signal light is switched by re(cid:173)
`newing the hologram on the corresponding section.
`The configuration that we used for 4 X 4
`switching requires different holograms for different
`input ports to be connected with an output port.
`Therefore strictly nonblocking N X N switching is
`achieved with N 2 types of holograms in general. As
`the positions of unwanted diffracted lights (e.g.,
`minus-order or higher-order diffracted light) on the
`output plane vary according to the input positions,
`the optical design for removing the unwanted lights
`becomes difficult when the number of inputs is
`large. Our proposed configuration associates a
`hologram with an output port as mentioned above.
`It requires only N types of holograms for N X N
`switching and makes it easy to remove unwanted
`diffracted light.
`The 1 X 64 switching with the LCD was studied
`to determine the factors restricting the number of
`output ports.
`In the experiment, optical switching
`was achieved by using binary phase-only modulation
`with a twisted nematic LCD. The display used was
`an active-dot-matrix type for an overhead projector.
`It has 640 horizontal and 400 vertical pixels, and its
`pixel size is 0.33 mm X 0.33 mm. Measurements
`showed that the phase-modulation depth of the light
`through the LCD was TT/3. 6
`Figure 2 shows the optical setup that we used. 7
`The light source is a He-Ne laser with a wavelength
`
`Output light beam
`
`Fig. 1. Schematic diagram of the proposed holographic
`switch that uses an LCD.
`
`0146-95921921171228-03$5.0010
`
`© 1992 Optical Society of America FNC 1017
`
`

`

`September 1, 1992 I Vol. 17, No. 17 I OPTICS LETTERS
`
`1229
`
`Liquid-Crystal Display
`I
`
`L2
`
`L1
`
`IJ2
`
`Fig. 2. Experimental setup for free-space optical switch(cid:173)
`ing by using binary phase-only modulation with an LCD.
`
`• • • • •••••• • • • •••• •
`•
`•
`• • ••••••••••••
`•
`• •
`• •
`• • ••••••• • •
`• • •
`• • •
`•
`•
`• • •
`• • •
`• • • • • • •
`• • •
`• • •
`• • •
`• • •
`• ••••• ••• •
`• •
`• •
`" ..............
`• •
`• •
`• • •••••••••••• •
`•
`•
`
`Fig. 3. Output light spots from 1 X 64 switching. The
`light sequentially switched to any one of 64 output posi(cid:173)
`tions is shown by multiple exposure.
`
`of 633 nm. The polarization of the input light is
`aligned parallel with the liquid-crystal molecule di(cid:173)
`rector at the front of the display by using a Glan(cid:173)
`Thompson polarizing prism (GTP) and a A./2 plate.
`The light beam goes through lenses L1-L6 and the
`It is reflected by mirror M2 and returns to
`LCD.
`mirror M3. The light extracted by mirror M3
`reaches the semitransparent screen through lenses
`L7-L10 and spatial filter F. The output light on the
`screen is monitored by the camera.
`The light beam is sent through the LCD twice for
`two purposes. One is to match the polarizations of
`the lights through the off-state and on-state pixels.
`The polarization of the light through the off-state
`pixels is at 90 deg to the light through the on-state
`pixels because the twisted nematic LCD rotates the
`polarization of the light through the off-state pixels
`by 90 deg. This right-angle polarization difference
`between the lights through the off- and on-state
`pixels disables the optical interference that is neces(cid:173)
`sary for the holographic beam steering. However,
`by introducing the double-passage configuration, the
`90-deg polarization rotation through the off-state
`pixels can be canceled, and the lights through the
`off- and on-state pixels can be made to interfere
`with each other. The other purpose is to make the
`phase-modulation depth of the LCD closer to 1r,
`which is necessary to suppress the unwanted dif-
`
`fracted light and to maximize the first-order dif(cid:173)
`fracted light. 8 The phase-modulation depth of the
`light through the LCD becomes 2/37T (doubled) after
`two passes.
`The diameter of the input beam is 3 mm. The
`light beam is expanded by lenses L3 and L4 andre(cid:173)
`duced by lenses L4 and L 7 so that the light passes
`through more pixels to obtain higher spatial resolu(cid:173)
`tion. The diameter of the light is expanded to
`27 mm on the LCD and reduced to 1/70 on the way
`from the LCD to an output port. Lenses L8 and L9
`and the spatial filter F extract the zeroth-order beam
`from the diffracted light generated by the effect of
`the regular grid structure of the LCD. 9 The light
`from the lens L9 is bent in a direction determined
`by the hologram. Lens L10 collects the deflected
`lights at the same angle to an output port.
`Figure 3 shows the experimental result for 1 X 64
`switching. The output light is switched to any one
`of 64 output ports, and the figure shows all of them
`by multiple exposure. There are 129 spots in the
`figure: 64 positive first order (the upper half), 64
`negative first order (the lower half), and 1 zeroth
`order (the center). Therefore, in practice, the num(cid:173)
`ber of selectable ports is 64. The positive and nega(cid:173)
`tive orders are generated because holograms on the
`LCD are simple phase-only gratings. The center
`spot (zeroth order) appears owing to two factors.
`One is that the original light reaches the screen
`when the hologram is erased for switching. The
`other is that the phase-modulation depth is less than
`7T rad.
`This experiment used simple holograms (64 types
`of one-dimensional grating) because they do not re(cid:173)
`quire much calculation time to generate or much
`memory to store. The 64 gratings are obtained by
`rotating three one-dimensional gratings having line
`pitches of 2, 3, and 5 pixels. The three sets of grat(cid:173)
`ings with different pitches diffract the lights and
`generate three square arrays of the light spots on
`the output plane as shown in Fig. 3. The largest,
`middle, and smallest squares correspond to the grat(cid:173)
`ings with the line pitches of 2, 3, and 5 pixels, re(cid:173)
`spectively. These line pitches change slightly during
`the rotation of the gratings owing to the algorithm
`for generating the gratings. This pitch variation
`gives the square arrangement of the light spots in(cid:173)
`stead of the circular arrangement.
`The pixel size mainly restricts the number of out-
`
`Insertion loss
`Fig. 4. Loss distribution of the 1 X 64 holographic
`switch .
`
`

`

`1230
`
`OPTICS LETTERS I Vol. 17, No. 17 I September 1, 1992
`
`•••••••••
`• •• • •• •
`• •
`• •
`• ••
`• ••
`• ••
`• •
`•••••••••
`
`It . . . . . . . . . . ..
`
`(a)
`
`. ~ ........ ~
`. ... . . . . ..
`..........
`. ' ..... .
`•••••••••
`..
`. ,.,
`• .• :t····
`•. ....... .
`••••••••••
`
`an output port was changed from 1/70 to 1/35.
`These changes approximately halve the diffraction
`angle of the output lights compared with that of the
`1 X 64 switching. Therefore the number of output
`ports decreases to 32, and the spot diameters are
`larger than in Fig. 3. Figure 5 shows the experi(cid:173)
`mental result for 2 X 32 switching, which is taken
`by multiple exposure. Figure 5(a) shows that the
`output light is switched to any one of 32 output
`ports when one input beam is incident, and Fig. 5(b)
`shows the output spots when two input beams are
`incident. These figures prove that the two inputs
`from different input ports are directed to the same
`output ports by the optical setup.
`In this Letter we have proposed a new configura(cid:173)
`tion of holographic switches that facilitates multi(cid:173)
`channel switching that uses an LCD. Whereas the
`previous configuration needed N 2 types of holo(cid:173)
`grams for strictly nonblocking N X N switching,
`the proposed configuration requires only N types of
`holograms, which is equal to the number of output
`ports. Furthermore the proposed configuration
`makes it easy to remove unwanted diffracted light.
`The 1 X 64 switching showed that the main factor
`restricting the number of the output ports is the
`pixel size. The 2 X 32 switching verified that the
`configuration can handle multichannel switching.
`The aberration of the expansion and reduction opti(cid:173)
`cal system restricted the number of input ports to
`one or two in the experiment. But if an LCD had
`640 X 400 pixels (pixel size 5 JLm X 5 JLm) and
`80 X 80 pixels were assigned to each input, the ex(cid:173)
`pansion and reduction system would not be neces(cid:173)
`In this
`sary, and the number of inputs would be 40.
`case, the number of pixels on the LCD would deter(cid:173)
`mine the number of inputs. If the twisted nematic
`LCD were changed to a homogeneous LCD and the
`thickness of the liquid-crystal layer were optimized
`to obtain the phase-modulation depth of 1r, a single(cid:173)
`path optical system through the LCD would be suffi(cid:173)
`cient to switch the light beam, and the switch
`configuration could be simple, as shown in Fig. 1.
`
`References
`
`1. M. Yamaguchi, T. Matsunaga, and M. Okuno, in Confer(cid:173)
`ence Record of GLOBCOM '90 (Institute of Electrical
`and Electronics Engineers, New York, 1990), p. 1301.
`2. M. Yamaguchi and I. Kobayashi, in Digest of European
`Conference on Optical Communication (Convention of
`National Societies of Electrical Engineers of Western
`Europe, Gothenberg, Sweden, 1989), p. 272.
`3. J. P. Herriau, A. Delboulbe, J. P. Huignard, G. Roosen,
`and G. Pauliat, in Digest of European Conference on
`Optical Communication (Convention of National Soci(cid:173)
`eties of Electrical Engineers of Western Europe,
`Venice, 1985), p. 419.
`4. D. 0. Harris, Appl. Opt. 30, 4245 (1991).
`5. P. Gunter, Phys. Rep. 93, 199 (1982).
`6. H. Yamazaki and M. Yamaguchi, Opt. Lett. 16, 1415
`(1991).
`7. T. H. Barnes, T. Eiji, K. Matsuda, and N. Ooyama,
`Appl. Opt. 28, 4845 (1989).
`8. J. L. Horner and J. R. Leger, Appl. Opt. 24, 609 (1985).
`9. D. Casasent and S. F. Xia, Opt. Lett. 11, 398 (1986).
`
`•
`
`tit
`
`.. .
`
`•
`
`If • • . . . .
`
`(b)
`Fig. 5. Output light spots of the 2 X 32 switching. The
`light sequentially switched to any one of 32 output posi(cid:173)
`tions is shown by multiple exposure.
`(a) Output light
`when one input beam is incident upon the optical setup.
`(b) Output light when two input beams are incident upon
`the optical setup.
`
`It determines and
`put ports in the experiment.
`quantizes the square size of the output spot arrays
`in Fig. 3. The number of output ports on the sides
`of the squares is determined by the square size and
`the spot size.
`If the pixel size were smaller, the
`square size and the number of the available spot
`squares could be increased, and then the number of
`output ports also could be increased.
`Figure 4 shows the loss distribution of the 1 X 64
`switch. The average, maximum, and minimum of
`the loss are 19.3, 20.5, and 17.7 dB, respectively.
`The loss dispersion is within 3 dB. Most of the loss
`is caused by reflections at the LCD, which has no
`antireflection coating, and at the many lenses. If
`we used an antireflection-coated LCD with smaller
`pixels and eliminated the beam-expansion and
`beam-reduction system, the loss would decrease
`remarkably.
`The 2 X 32 switching was demonstrated to verify
`that a hologram can direct the light beams from dif(cid:173)
`ferent input ports to the same output port. The op(cid:173)
`tical setup shown in Fig. 2 was used, and two parallel
`light beams are incident upon the LCD as inputs.
`In this setup, the aberration increases if the distance
`from the optical axis becomes longer, which results
`mainly from the optical system for expansion and
`reduction of the light beam. Therefore, to suppress
`the aberration, the diameter of the input beams was
`reduced to 1 mm, and the input positions were
`moved closer to the optical axis. Furthermore the
`beam-reduction ratio on the way from the LCD to
`
`