`
`851
`
`100-GHz-Resolution Dynamic Holographic
`Channel Management for WDM
`
`A. D. Cohen, M. C. Parker, Associate Member, IEEE, and R. J. Mears, Associate Member, IEEE
`
`Abstract— We present results that demonstrate the active
`management of eight 100-GHz-spaced wavelength-division-
`multiplexed (WDM) channels using a polarization-insensitive
`spatial-light-modulator-based holographic filter. The filter
`has a fundamental stepping resolution close to 25 GHz and
`3-dB width of each individual passband equal to 42 GHz.
`Arbitrary permutations of dropped and passed channels are
`possible. Additionally, passed channels can be transmitted
`according to an equal-amplitude comb filter function or be
`transmitted with weighted-amplitude passbands to effect power
`equalization where necessary. The holographic technique is
`further extendable to passband spectral engineering, yielding
`near-rectangular “top hat” passbands. Suppression of
`the
`dropped channels, which is consistently >15 dB in these
`experiments, can be straightforwardly improved by deploying
`a spatial
`light modulator of greater resolution and higher
`pixel number. The technology has potential application as the
`key element in both an optical add–drop multiplexer and a
`dynamic multichannel equalizer. The passband 3-dB width can
`be straightforwardly reduced to allow processing of multiple
`channels at the 50-GHz spacing of future WDM systems.
`
`Index Terms—Equalizers, holographic optical components, liq-
`uid crystals, optical amplifiers, optical communication, optical
`filters, spatial light modulators, wavelength-divison multiplexing.
`
`I. INTRODUCTION
`
`THE APPLICATION of liquid-crystal (LC) based etalon
`
`devices to wavelength filtering is well established [1].
`The combination of a pixellated LC element with a bulk
`diffraction grating, allowing on device assembly the separate
`processing of each member of a fixed set of wavelengths,
`has also been extensively studied [2]. In the technique of
`[2] each LC pixel operates to rotate the polarization of
`light at a single given wavelength. By contrast, the work
`reported in this letter concerns recent developments of a
`technique in which a diffraction grating and an LC spatial
`light modulator (SLM) combine to form a wavelength filter by
`virtue of the holographic Fourier replay of the SLM’s binary
`phase array [3]. In this technique, an arbitrary set of WDM
`channels—arbitrary in both number and individual channel
`wavelengths—are processed collectively.
`
`Manuscript received November 17, 1998; revised March 2, 1999.
`A. D. Cohen was with the Department of Engineering, Cambridge Univer-
`sity, Cambridge CB2 1PZ, U.K. He is now with JDS FITEL Inc., Nepean,
`ON, K2G 5W8, Canada.
`M. C. Parker was with the Department of Engineering, Cambridge Univer-
`sity, Cambridge CB2 1PZ, U.K. He is now with Fujitsu Telecommunications
`Europe Ltd., Northgate House, Colchester CO1 1HH, U.K.
`R. J. Mears is with the Department of Engineering, Cambridge University,
`Cambridge CB2 1PZ, U.K.
`Publisher Item Identifier S 1041-1135(99)05134-4.
`
`Fig. 1. Schematic diagram of experimental architecture.
`
`equalizing
`reconfigurable multichannel
`Dynamically
`add–drop wavelength-division-multiplexed
`(WDM) filter
`elements have been identified as key components in future
`generation wavelength-switched networks [4]. Prototyping of
`such devices is typically based on the acoustooptic tunable
`filter
`(AOTF)
`[5]; although the AOTF-based technology
`is relatively well developed, the achievement of 100-GHz
`resolution requires a design of increased complexity [6]. By
`contrast,
`the holographic technique attains such resolution
`by means of straightforward modifications to filter design
`parameters, as described in this letter. The facility to provide
`dynamic holographic spectral equalization of WDM channels
`(for instance, to compensate for erbium-doped fiber amplifier
`(EDFA) gain tilt and fluctuations in individual channel
`powers) has already been demonstrated in proof-of-principle
`experiments [7]. The in-line holographic filter is based on a
`pixellated programmable binary phase diffractive element—a
`ferroelectric LC SLM (FLC–SLM) of
`relatively coarse
`spatial resolution—which provides wavelength dispersion in
`conjunction with a higher resolution fixed diffraction grating.
`Both the dispersive elements are located in the collimated
`beam of a 4-
`relay extending between input and output
`single mode fibers. This technique is polarization-insensitive
`[8], robust (owing to the inherent redundancy of displayed
`holograms) and scalable to many more channels. With an
`optimized electronic drive scheme the fast FLC response can
`be exploited to yield device reconfiguration as fast as 20 s [9].
`
`II. EXPERIMENT
`The experimental configuration is similar to that described
`in detail in a previous publication [7]. The key modification
`1041–1135/99$10.00 ª
`
`1999 IEEE
`
`FNC 1013
`
`
`
`852
`
`IEEE PHOTONICS TECHNOLOGY LETTERS, VOL. 11, NO. 7, JULY 1999
`
`is the adoption of a folded architecture to accommodate
`the reflective blazed grating (see Fig. 1). In the experiments
`reported in this letter, the fixed diffractive element was a
`300 line-pairs/mm reflective-blazed grating. The resultant filter
`27.5 GHz lies well within the current
`stepping resolution of
`100-GHz ITU channel spacing standard. Since the blazed
`grating used was optimized for a wavelength of 1 m, its
`deployment reduced the basic filter insertion loss by only 1.1
`2-dB improvement should be achieved by using
`dB; a further
`a grating optimized for high efficiency 1.55- m operation at
`the design angle of incidence. The SLM consists of a two-
`128 pixels on a 165- m pitch with
`dimensional array of 128
`interpixel dead space of 15 m. However, in this application
`the device is simply used to display one-dimensional (1-D)
`binary phase holograms.
`
`Hologram Design
`Design calculations were based on numerical solutions to
`the following equation, which is derived by consideration of
`summed diffraction angles:
`
`(1)
`
`is the displacement of
`is the filter wavelength,
`In (1),
`mm is
`the output fiber from the optical axis,
`is the number of
`the focal length of the lens,
`m is the SLM pixel
`pixels in the 1-D hologram,
`m is the period of the fixed grating.
`pitch and
`is the angle made by the normal of the fixed grating
`to the input optical axis. Within the physical constraints of
`the arrangement the value of this angle was optimized for
`maximum blazed grating diffraction efficiency and minimum
`(to reduce output spot aberration). Use of a standard simulated
`annealing (SA) algorithm generally yields optimum results
`is an integer in
`when the design spatial frequency parameter
`, i.e., 0–64 in this particular case. However,
`the range 0–
`can be input to a modified algorithm that,
`fractional values of
`10%–20% lower
`despite a tendency to produce holograms of
`diffraction efficiency, allows quasicontinuous filter tuning [10].
`With the SLM displaying a hologram designed to filter a
`single channel, the individual filter passband is near-Gaussian
`in form having a 3-dB width, determined by the overlap
`integral of the dispersed input spectrum over the output fiber
`core, of 42 GHz. This filter function results in adjacent channel
`isolation (defined at 30 GHz from adjacent channel center) of
`32 dB. The filter stepping resolution is given approximately
`to yield
`by differentiating (1) with respect to
`
`(2)
`
`which, for the parameters in question, implies
`nm, i.e., a frequency resolution
`GHz, for integer
`. This agrees well with the experimentally
`increments in
`GHz. To achieve
`GHz,
`observed mean
`which would yield conformance to the ITU grid by setting
`, the fixed grating pitch must simply be reduced to
`3.14 m.
`
`(a)
`
`(b)
`
`(a) Normalized output of equalizer with channel #3 dropped. (b)
`Fig. 2.
`Normalized output of equalizer with channel #4 dropped.
`
`Holograms have been designed by a version of the SA algo-
`rithm adapted to give multiple, arbitrary-amplitude passbands
`[7], [10] to demonstrate various pass/drop and equalization
`between
`permutations of eight channels. By setting
`110 GHz is achieved. The
`channels, a mean spacing of
`modified SA algorithm allows holograms to be designed
`for equal transmission of passed channels (i.e., a flat filter
`function) or to obtain equalized outputs, i.e.,
`in this case
`compensating for the spectral variation of the EDFA ASE. The
`equalization technique can also be straightforwardly extended
`to compensate for dynamic input channel power variation
`[7]. A further possibility presented by the modified design
`algorithm is the control of passband shape; near-rectangular
`flat-topped composite passbands can be obtained by ensuring
`channels
`overlap of three individual passbands. Control of
`, or
`for
`incurs an excess filter loss of
`flattened passbands, but at the high-channel counts anticipated
`in future systems the incremental penalty will be small.
`
`III. RESULTS
`All results were recorded using an optical spectrum ana-
`lyzer. Fig. 2(a) and (b) shows a single channel dropped in two
`different spectral locations, channels #3 and #4, respectively,
`with the other seven passed and equalized. The dynamic
`7-nm wavelength range is
`range of equalization over the
`about 3 dB, corresponding to the range of EDFA ASE levels
`in this portion of the spectrum. Previous experiments have
`demonstrated that a dynamic range of the order of 10 dB
`is within the capabilities of this system [7] and simulations
`show that the use of a higher resolution SLM would yield a
`commensurately greater dynamic range of operation. In addi-
`tion to providing attenuation at the dropped channel frequency,
`the filter also provides suppression of the accumulated EDFA
`ASE between passed channels; this additional functionality
`is particularly beneficial in systems containing concatenated
`optical amplifiers. For hologram 1, interchannel ASE sup-
`pression is consistently greater than 15 dB and the dropped
`channel #3 is suppressed by approximately 19 dB. Hologram
`19-dB suppression of
`2, however, while similarly yielding
`the dropped channel #4, has binary phase quantization noise
`peaks that cause the poorest interchannel suppression to fall
`6 dB. While this result demonstrates the flexibility of the
`to
`channel management technique, the hologram performance is
`suboptimal. Further hologram design iterations would reduce
`local replay noise peaks in number and magnitude and force
`
`
`
`COHEN et al.: 100-GHz-RESOLUTION DYNAMIC HOLOGRAPHIC CHANNEL MANAGEMENT FOR WDM
`
`853
`
`further physical redesign of the filter, including further scale
`reductions in pixel size and fixed grating period, would be
`necessary for the composite “top hat” passband to conform to
`the ITU grid with acceptable isolation performance.
`
`IV. CONCLUSION
`This letter has reported 27.5-GHz resolution active holo-
`100
`graphic management of eight WDM channels spaced by
`GHz. The flexibility of the computer-based hologram design
`algorithm allows arbitrary add/drop permutations in which the
`filter passbands have arbitrary relative transmission. Dropped
`15 dB and the filter
`channels are consistently suppressed by
`also provides commensurate ASE suppression between passed
`channels. The suppression and adjacent channel crosstalk can
`be straightforwardly improved by deploying a spatial light
`modulator of greater resolution and higher pixel number. The
`holographic technique can be extended to facilitate spectral
`engineering of the passband, yielding a near-rectangular “top
`hat” shape. The FLC-SLM-based holographic filter is robust
`and can incorporate an SLM capable of reconfiguration as fast
`20
`s. The technology is based upon elements which
`as
`are low cost in volume production; it could form the key
`component
`in optical add–drop and equalization nodes of
`future WDM networks.
`
`REFERENCES
`
`[1] A. Sneh and K. M. Johnson, “High-speed continuously tunable liquid
`crystal filter for WDM networks,” J. Lightwave Technol., vol. 14, pp.
`1067–1080, June 1996.
`[2] J. S. Patel and Y. Silberberg, “Liquid crystal and grating-based multiple-
`wavelength cross-connect switch,” IEEE Photon. Technol. Lett., vol. 7,
`pp. 514–516, May 1995.
`[3] S. T. Warr, M. C. Parker, and R. J. Mears, “Optically transparent
`digitally tunable wavelength filter,” Electron. Lett., vol. 31, no. 2, pp.
`129–130, 1995.
`[4] A. E. Willner, “Systems issues for WDM components,” in IEEE LEOS
`Summer Topical Meeting WDM Components Technology Montr´eal,
`Qu´ebec, Canada, 1997, pp. 5–6, paper WB1.
`[5] S. H. Huang, A. E. Willner, Z. Bao, and D. A. Smith, “Experimental
`demonstration of active equalization and ASE suppression of three
`2.5-Gb/s WDM-network channels over 2500 km using AOTF as trans-
`mission filters,” IEEE Photon. Technol. Lett., vol. 9, pp. 389–391, Mar.
`1997.
`[6] T. Nakazawa, M. Doi, S. Taniguchi, Y. Takasu, and M. Seino,
`“Ti:LiNbO3 AOTF for 0.8 nm channel-spaced WDM,” in OSA Conf.
`Optical Fiber Communication, San Jose, CA, Feb. 22–27, 1998,
`postdeadline paper.
`[7] M. C. Parker, A. D. Cohen, and R. J. Mears, “Dynamic holographic
`spectral equalization for WDM,” IEEE Photon. Technol. Lett., vol. 9,
`pp. 529–531, Apr. 1997.
`[8] S. T. Warr and R. J. Mears, “Polarization-insensitive operation of ferro-
`electric liquid crystal devices,” Electron. Lett., vol. 31, no. 9, pp.
`714–716, 1995.
`[9] H. J. White, G. M. Proudley, C. Stace, N. A. Brownjohn, R. Dawkins,
`A. C. Walker, M. R. Taghizadeh, C. P. Barrett, D. T. Neilson, W. A.
`Crossland, J. R. Brocklehurst, and M. J. Birch, “The OCPM demonstra-
`tor system,” in OSA Topical Meeting Photonics in Switching, Salt Lake
`City, Utah, 1995, page PPd1.
`[10] A. D. Cohen, “Spatial light modulator technologies for WDM,” Ph.D.
`dissertation, Univ. of Cambridge, Cambridge, U.K., 1998.
`[11] A. D. Cohen and R. J. Mears, “Dynamic holographic eight-channel
`spectral equlalizer for WDM,” in IEEE LEOS Summer Topical Meeting
`WDM Components Technology, Montr´eal, PQ, Canada, 1997, pp. 46–47,
`paper ThC2.
`
`(a)
`
`(b)
`
`(a) Normalized output of equalizer, channels #5 and #7 dropped. (b)
`Fig. 3.
`Normalized output of equalizer, channels #2 and #3 dropped.
`
`(a)
`
`(b)
`
`(a) Normalized filter function of eight-channel equalizer operating
`Fig. 4.
`at 100-GHz channel spacing. (b) Spectral shaping of the holographic fil-
`ter passband. Single Gaussian passband shown overlaid (broken line) for
`comparison.
`
`the majority of the residual noise to peripheral regions of
`the spectrum, such that the performance would match that
`of hologram 1. The hologram design process would also be
`improved by the use of an in situ feedback loop to compensate
`for system aberrations.
`Fig. 3(a) and (b) demonstrates the facility to drop two
`channels: In Fig. 3(a) channels #5 and #7 are dropped. In this
`9 dB and
`case, the interchannel ASE suppression is at least
`10 dB, that of
`while the suppression of channel #5 is only
`channel #7 approaches 20 dB. Again, further design iterations
`would yield improved performance. Fig. 3(b) shows channels
`#2 and #3 dropped, with suppression greater than 12–15 dB
`in both cases and passed channels uniform to within 1.8 dB
`following equalization. Fig. 4(a) illustrates a filter function
`with transmission equalized from 3 dB to within 1.5 dB
`for all eight channels passed; interchannel ASE suppression
`ranges from 12 to 23 dB.
`Holograms designed to produce ‘top hat’ passbands yielded,
`in the previous filter configuration, responses having sig-
`20 dB : 3 dB-width merit functions
`nificantly improved
`[11]. The reduced effective Gaussian width was achieved by
`designing a hologram to yield three overlapping passbands. In
`the new, higher resolution reflective configuration, a passband
`0.7 nm ( 88 GHz)
`top that is flat to within 3 dB over
`20-
`has been demonstrated [see Fig. 4(b)] together with a
`0.9 nm ( 112 GHz). The merit function
`dB width of
`derived in the new configuration is approximately 1.35, a
`20-dB: 3-dB-width ratio
`considerable improvement on the
`of 2.25 typically demonstrated in previous experiments. Some