P~. D 2 1 0~0
`
`DYNAMIC HOLOGRAMS FOR
`WAVELENGTH DIVISION MULTIPLEXING
`
`A dissertation submitted for the degree of
`Doctor of Philosophy
`aL the University of Cambridge
`
`UNJVERSin
`LIBRARY
`CAMBRIDGE
`
`Michael Charles Parker
`Sidney Sussex College
`
`November 1996
`
`FINISAR 1005
`
`

`

`D eclaration
`
`This dissertation contains the results of research undertaken by the Author between January
`1993 and August 1996 at the Department of Engineering, University of Cambridge. Except
`for the experiment described in §4.3 which is the result of a collaboration between the Author
`and S.T. Warr, no part of this dissertation is the result of work done in collaboration with
`others. The contents have not been submitted, in whole or in part, for any other University
`Degree or Diploma.
`
`/?! CJ~k:v
`
`Michael Charles Parker
`Cambridge, November 1996
`
`

`

`This dissertation consists of six sections, summarised below:
`
`Summary
`
`l ntJ·oduction : This chapter provides an overview of the major concepts and technologies in
`current optical telecommunications research, such as wavelength division multiplexing
`(WDM), optical t ime domain multiplexing (OTDM), and erbium-doped fibre amplifiers
`(EOFAs), the support and development of which has acted as the major motivation behind
`this Ph.D. The current competing technologies for tunable filters and lasers are reviewed,
`followed by an introduction to the use of holography in optical telecommunications.
`
`Ferroclectt·ic L iquid Crystal Spatial L ight Modulators: An introduction to spatial light
`modulators (SLMs) is given, followed by an analysis of how the ferroelectric liquid crystal
`(FLC) material properties, such as the switching angle, affect the diffraction efficiency of a
`FLC SLM. Original analysis of the phase modulation properties of FLC with respect to
`circularly polarised light is given, followed by a method to double the allowable phase
`modulation using a quarter-wave plate and mirror, whi le maintain ing polarisation
`insensitivity. A characterisation of the FLC SLM used in the experiments within this
`dissertation is also presented.
`
`Bina t-y-Pitase Hologram Analysis: Origina l theoretical analysis of the performance of large
`!-dimensional binary-phase holograms is presented. Expressions are derived for the expected
`diffraction efficiency (-36.5%) and minimum signal-to-noise ratios (>N/2p) for binar-y-phase
`holograms with N pixels, fanning out top uniform spots of light. These results can be used
`for system design, since they allow the expected power budget and noise performance of the
`system to be calculated. The results arc equally applicable to binary-phase holograms used in
`beam-steering and spatial fanning out of monochromatic light.
`
`T una ble W avelength F ilter : Results from a polarisation-insensitive, high resolution and
`digital holographic filter are presented. The filter was tunable over 82.4nm in steps of 1.3nm,
`with the filter 3dB passband being 2nm. The filter exhibited an insertion loss of 22.8dB.
`Simultaneous multiple wavelength filtering is also demonstrated. The design of low loss and
`high resolution holographic tunable wavelength filters is discussed, and minimum system
`sizes calculated for a filter resolution of 0.8nm. A filter insertion loss of 6.SdB should also be
`possible. An analysis of the chirp imparted onto a signal passing through a holographic
`wavelength filter is given.
`
`T una ble F ibre Laser : Resu lts are presented of a tunable erbium-doped fibre laser, tuned
`using the holographic wavelength filter. Tuning over 38.Snm in steps of 1.3nm with output
`powers of up to -13dBm has been achieved. The inherent laser linewidth was measured to be
`of the order of 3kHz, and
`the wavelength stability was of the order of 0.1 nm.
`Multiwavelength lasing action has also been demonstrated.
`
`Conclusions: Conclusions are drawn from the results of the experimental holographic
`wavelength filter described in this dissertation, and the requirements for improved tunable
`filters discussed. The use of holographic wavelength filtering for dynamic EDFA gain
`equalisation is proposed (with preliminary results presented in Appendix B.) Future work
`and ideas are presented as to how a compact, low-loss, polarisation-insensitive and high
`speed (- nanosecond switching time) wavelength filter and space-wavelength switch for
`WDM might be realised.
`
`Keywords: Ferroelectric liquid crystal (FLC), spatial light modulator (SLM), WDM,
`computer generated hologram
`(CGll), digital holographic wavelength filter, phase
`modulation, tunable laser, EDFA gain equalisation
`
`ii
`
`

`

`Acknowled gements
`
`I would foremost like to thank my supervisor Robert Mears who has been the main source of
`inspiration and motivation behind my Ph.D. His advice, suggestions, help, open-mindedness
`and light-handed approach to my work, has given me the freedom to explore, experiment
`and discover, so ensuring that my Ph.D. was the enjoyable, fulfilling and stimulating op(cid:173)
`portunity for research in a first-class environment, which was everything I had hoped for.
`
`Many thanks to my friends in Cambridge who have made the past four years so enjoy(cid:173)
`able, exciting, stressful and fun, and who have left me with so many memories, notably
`Brendan, Kent, Barry, Ian, Freddy and Rachel, Charlanne, as well as Irene, Eileen, Fiona
`and Geeta at Addenbrookes! Likewise, thanks to the other ex-'Basement Boys' Steve, Tim
`and Adam for help, useful discussion and argument through the years. I'd also like to thank
`Dave 'Kozza' Kozlowski for mad humour, advice, borrowing of kit and help in the clean
`room, and Bill Crossland for his generous support.
`
`A Ph.D. cannot be undertaken without the technical and practical background support
`of many people, so thanks also go to the workshop: Russell, Mick, Steve and Adrian {the
`Doom Brothers) for their technical assistance and help, to Caryn for her unfailing cheerful(cid:173)
`ness and help, and to Anthea Ansell at Sidney Sussex College, who greatly helped me with
`funding and accommodation when I most needed it.
`
`Finally, and most importantly, I would also like to thank my Parents for the encouragement,
`support and love they have shown me throughout my time at Cambridge.
`
`Ill
`
`

`

`
`@fififififibfififi
`fiy‘/¢/‘../9/4-
`.
`° % W
`fix: \‘3‘
`first ‘b‘
`l.
`
`
`
`

`

`r
`
`Contents
`
`1 Introduction
`1.1 WDM . . . . . . . . . .
`1.2 Technologies for WDM .
`1.2.1 Acousto-optic Tunable Filter
`1.2.2 Fibre Fabry-Perot Etalon
`1.2.3 Semiconductor Filters
`1.2.4 Mechanical Systems
`.
`1.2.5 Emerging Technologies .
`'!Unable Lasers for WDM ..
`1.3.1 Tunable Fibre Lasers .
`1.3.2 External Cavity Tunable Fibre Lasers
`Internal Cavity Tunable Fibre Lasers .
`1.3.3
`
`1.3
`
`1.3.4 Semiconductor Tunable Lasers
`1.4 Holography in Telecommunications
`1.5 Summary of Thesis Chapters
`
`2 FLC SLMs
`2.1
`Introduction to SLMs
`2.1.1 SLM Technologies
`2.2 Experimental FLC SLM Characterisation
`2.3 FLC SLM Properties . . . . . . . . . . . .
`2.3.1 Phase Modulation . . . . . . . . .
`2.3.2 Doubling FLC Phase Modulation .
`2.3.3 Diffraction Efficiency . . . . . . . .
`2.3.4 Polarisation Insensitivity
`. . . . .
`2.3.5 Polarisation-Insensitive Mu lti-Phase FLC Holograms .
`
`3 Binary-Phase Hologr a m Analysis
`3.1
`Introduction .
`3.2 Definitions . .
`
`v
`
`1
`1
`2
`3
`4
`4
`
`5
`5
`5
`6
`6
`7
`8
`8
`9
`
`11
`11
`12
`13
`16
`16
`19
`21
`22
`23
`
`24
`24
`24
`
`

`

`. . . . . . . . . . . . . . . . . ...
`3.3 1-D Analysis
`3.3.1 Definition of Hologram and Output Field
`3.3.2 Steering Light to a Single Spot . . ... .
`3.3.3 Fanning Out to Multiple Spots . . ... .
`3.3.4 Calculation of Diffraction Efficiency and SNR .
`3.3.5 Sine Envelope . . . . . . . . . . . . .
`3.3.6 Experimental Validation of Theory .
`3.3.7 Conclusions and Summary.
`3.4 Algorithm . . . ...
`3.4.1
`Introduction
`3.4.2
`T he Hologram Generation Algorithm .
`3.4.3
`Final Word about Stability of Holograms
`
`4 H olographic >.-Filter
`4.1
`Introduction ....
`4.2 Principle of Operation
`4.3 Proof-of-Principle ...
`4.3.1 Experimental ArchiLectme.
`4.3.2 Results ..
`4.3.3 Discussion .
`4.4 Experiment II . . .
`4.4.1 Design . . .
`4.4.2 Theoretical Performance .
`4.4.3 Results ..
`4.4.4 Discussion .
`4.4.5 Temporal Modulation of Filtered Light
`4.4.6 MulLiple Wavelength Filtering .
`4.5 Conclusions . . . . . . . .
`4.6 Wavelength Filter Design
`4.6.1 Design Specifications .
`4.6.2 Linear Architecture
`4.6.3 Folded Architecture
`4.6.4 Architectures Without a Fixed Grating
`4.6.5 Summary . . . ..
`4. 7 Temporal Dispersion . . .
`4.7.1 Simplified Analysis
`4.7.2 Rigorous Analysis
`
`Vl
`
`26
`26
`29
`32
`34
`38
`40
`42
`43
`43
`44
`45
`
`47
`47
`47
`49
`49
`51
`51
`53
`53
`56
`58
`58
`61
`62
`62
`64
`64
`65
`68
`71
`73
`74
`74
`75
`
`

`

`5 Digitally Tunable Fibre Laser
`
`5.1
`
`Introduction .
`
`5.2 Experiment
`5.2.1 EDFL Architecture .
`5.2.2 Roundtrip Losses ..
`
`5.2.3 EDFA Specification
`5.3 Results . . . . . . . . . . . .
`
`5.3.1
`5.3.2
`
`5.3.3
`
`Tuning Range . . .
`Wavelength Stability .
`
`Fine Tuning of EDFL
`Laser Linewidth
`. . .
`
`5.3.4
`5.3.5
`Characteristic LI-Curve
`5.3.6 Simultaneous Multiple-Wavelength Lasing .
`5.3.7 Q-switching .
`
`5.4 Conclusions
`
`6 Conclusions
`6.1 Further Work
`6.1.1
`Low-loss Compact Space->. Switch
`FLC Technology ..
`6.1.2
`
`6.1.3
`6.1.4
`6.1.5
`
`6.1.6
`6.1.7
`
`Apodisation . . . .
`
`Continuous 'lUning .
`Dynamic Spectral Equalisation
`
`2-Dimensional Hologram Analysis
`
`Semicond uctor SLM . . . . .. . .
`
`Bibliography
`
`A Appendix
`A.l Justification of Basic Assumptions
`A.2 Geometric (Binomial) Series Results
`. ..
`
`A.3 Non-Symmetric Holograms
`A.4 Sterling Approximation for F{p)
`
`B Associated Publications
`B.1 Journals ..
`B.2 Conferences
`
`VII
`
`79
`79
`80
`80
`81
`81
`
`83
`83
`84
`86
`
`88
`91
`92
`93
`93
`
`95
`96
`97
`98
`98
`98
`98
`99
`99
`
`101
`
`112
`112
`
`115
`
`116
`119
`
`121
`121
`121
`
`

`

`List of Figures
`
`2.1 Generic eledrically addressed spatial light modulator {EASLM) .
`2.2 Birefringent axes of FLC in 2 states separated by Oo ..... . .
`2.3 Desired hologram and actual hologram displayed on SLM
`. . . .
`2.4 Unpolarised light incident on a FLC cell at arbitrary orientation
`2.5 FLC cell and mirror together give no phase modulation
`. . . . .
`2.6 FLC cell, quarter-wave plate and mirror doubles phase modulation
`2. 7 Arbitrary binary-phase hologram H,p(X) with ¢ phase modulation
`
`Intensity profile of arbitrary holographic image . . . . . . . .
`3.1
`3.2 Arbitrary 1-D binary-phase hologram consisting of 12 pixels .
`3.3 A pair of delta functions fomier-tra.nsforms to a sinusoid ...
`3.4 Arbitrary 1-D binary-phase hologram and its fourier transform
`3.5 Expected values of I cos(kY) I . . . . . . . . . ...... .
`3.6 Normalised intensity profile of a spot of light . . . . . . .
`3.7 FWIIM of holographic peak, normalised so that b.x = 21r
`3.8 The sine envelope due to the finite pixel width W ....
`3.9 Comparison of measured and predicted 1-D hologram performance
`3.10 Increase of spot intensity non-uniformity with fanout p ...
`3.11 Arbitrary holographic image I(x) with a fanout of p spots .
`
`4.1 Diffraction of light due to a regular grating
`. . . . . . . ..
`4.2 High resolution tunable wavelength filter using SLM and a fixed grating
`4.3 Polarisation-sensitive 2.5nm resolution wavelength filter
`4.4 Cross-section of reflective binary-phase fixed grating
`4.5 Pump laser-diode power spectrum
`. . . . . . . . . .
`4.6 Filtered power spectrum using 3 different holograms
`4.7 Polarisation-insensitive 1.3nm resolution wavelength filter
`4.8 Photograph of polru·isation-insensitive wavelength filter .
`4.9 Cross-section of transmissive binary-phase fixed grating
`4.10 Off-axis coupling of light into a glass fibre . .. .
`.
`4.11 Linear plot of filter passband with FWHM=2.0nm
`
`viii
`
`12
`13
`15
`16
`19
`20
`21
`
`25
`27
`28
`29
`
`31
`35
`36
`39
`41
`42
`44
`
`48
`49
`50
`51
`52
`
`52
`54
`55
`56
`57
`
`59
`
`

`

`List of Figures
`
`2.1 Generic electrically addressed spatial light modulator (EASLM) .
`2.2 Birefringent axes of FLC in 2 states separated by 00 • . . • . . .
`2.3 Desired hologram and actual hologram displayed on SLM
`. . . .
`2.4 Unpolarised light. incident on a FLC cell at arbitrary orientation
`2.5 FLC cell and mirror together give no phase modulation . . . . .
`2.6 FLC cell, quarter-wave plate and mirror doubles phase modulation
`2.7 Arbitrary binary-phase hologram H<t>(X) with 4> phase modulation
`
`Intensity profile of arbitrary holographic image . . . . . . . .
`3.1
`3.2 Arbitrary 1-D binary-phase hologram consisting of 12 pixels .
`3.3 A pair of delLa functions fourier-transforms to a sinusoid . . .
`3.4 Arbitrary 1-D binary-phase hologram and its fourier transform
`3.5 Expected values of I cos(kY)I . . . . . . . . . . . . . . . .
`3.6 Normalised intensity profile of a spot of Hght
`. . . . . . .
`3.7 FWHM of holographic peak, normalised so that D.x = 21r
`. . . .
`3.8 The sine envelope due to Lhe finite pixel width W
`3.9 Comparison of measured and predicted 1-D hologram performance
`3.10 Increase of spot intensity non-uniformity with fanout p . . .
`3.11 Arbitrary holographic image I(x) with a fanout of p spots .
`
`. . . . . . . . . . . . . . . .
`4.1 Diffraction of light due to a regular grating
`4.2 High resolution tunable wavelength filter using SLM and a fixed grating
`4.3 Polarisation-sensitive 2.5nm resolution wavelength filter
`4.4 Cross-section of reflective binary-phase fixed grating
`4.5 Pump laser-diode power spectrum
`. . . . . . . . . .
`4.6 Filtered power spectrum using 3 different holograms
`4. 7 Polarisation-insensitive 1.3nrn resolution wavelength filter
`4.8 Photograph of polarisation-insensitive wavelength filter .
`4.9 Cross-section of transmissive binary-phase fixed grating
`4.10 Off-axis coupling of light. into a glass fibre . . . . .
`4.11 Linear plot of filter passband with FWHM=2.0nm
`
`viii
`
`12
`13
`15
`16
`19
`20
`21
`
`25
`27
`28
`29
`31
`35
`36
`39
`41
`42
`44
`
`48
`49
`50
`51
`52
`52
`54
`55
`56
`57
`59
`
`

`

`4.12 Logarithmic plot of filter passband . . . . . .
`4.13 Amplified spontaneous emission from EDFA .
`4.14 Filtered ASE using 11 different holograms . .
`4.15 Linear plot of dual-wavelength filter passbands
`4.16 Logarithmic plot of multiple wavelength filtering
`4.17 Wavelength filter with a linear architecture
`. . .
`4.18 Minimum dimensions for a 0.8nm resolution filter with linear architectme
`. . . . . . . . . . . .
`4.19 Reflective holographic filter in Littrow configuration
`4.20 Wavelength filter with a folded architecture . . . . . . . . . . . . . . . . .
`4.21 Minimum dimensions for a 0.8nm resolution filter with folded architecture
`4.22 Folded architecture wavelength filter without a fixed grating . . .
`4.23 Uniform intensity lighL of width L incident on a grating . . . . .
`4.24 Light with a gaussian intensity distribution incident on a grating
`
`5.1 Holographic digitally tunable erbium doped fibre laser . . . . . . . . . . . .
`5.2 Photograph of EDFL experimental setup, with tunable filter, EDFA, 3dB
`coupler, SLM power supply and controlling PC in view ....
`5.3 ASE spectrum of HP EFA2002 at a pump current of 350mA .
`5.4 Eleven successively tuned lasing wavelengths
`. . . . . . . . .
`5.5 Histogram and temporal plot of wavelength stability of EDFL .
`5.6 Histogram and time plot of wavelength stability of EDFL with no SLM in
`cavity
`. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
`5.7 Three holograms with a similar fundamental spatial frequency.
`5.8 Lasing wavelengths of EDFL using the three similar holograms
`. . . . . . . . . . . . . . .
`5.9 Digital oscilloscope trace of EDFL output power
`5.10 Heterodyne measurements of the EDFL linewidth with 5km delay line . . .
`5.11 Heterodyne measurement of the EDFL linewidth without SLM in cavity, 5km
`delay line
`. . . . . . . . . . . . . . . . . . . . . . . . . .
`5.12 Characteristic curve for the EDFL . . . . . . . . . . . .
`5.13 Two competing lasing modes at 1556nm and 1562.5nm .
`
`6.1 Exploded 2f compact 3 x 3 space-wavelength switch
`6.2 Packaged 3 x 3 space-wavelength switch
`6.3 Compact space->. switch using transmissive semiconductor-SLM with refrac-
`tive prism . . . . . . . . . . . .
`
`59
`60
`60
`63
`63
`65
`67
`68
`69
`70
`71
`74
`76
`
`81
`
`82
`83
`84
`85
`
`86
`87
`87
`88
`89
`
`91
`92
`92
`
`96
`97
`
`99
`
`A.1 PDF of the function y = cos(x) . . . . . . . . . . . . . . . . . . . . .
`A.2 Frequency distribution of cos{mX) for X = 0.5 and m = 1 ~ 10,000
`A.3 PDFs over the range 0 to 1r radians . . . . . . . . . . . . . . . . . . .
`
`114
`114
`115
`
`IX
`
`

`

`Axl Plots of |¢-.ns{(1’)| and JSEHM’H for —.’«E < A” g g ................
`
`A5 Plot of Fm) against its approximation .....................
`
`i:lc-1
`VI
`~
`v
`
`X
`
`

`

`Chapter 1
`
`Introduction
`
`W e are presently experiencing the dawn of a new age in our society - t,lle Age of Infor(cid:173)
`mation. The ability of people to have easy access to huge quantities of information and
`data has never been greater. The advent of the Information Superhighway mostly in the
`form of the World Wide Web and the Internet has even occurred during the relatively short
`period of this Ph.D. Digitalisation, a global fibre-optical network incorporating all optical
`amplifiers, ever faster and more powerful personal computers have underpinned this trans(cid:173)
`formation in the way we live. A whole host of technologies have been brought together to
`make this revolution possible. But human nature dictates that the technology must always
`be improved. The Internet is all ready too slow; people want to receive on-line photographic
`quality pictmes, CD quality sound and real-time high definition video into their PCs, others
`want to perform intensive distributed computing over the Net and still others arc using it for
`real-time virtual reality applications. If the revolution is not to falter, then ever higher data
`bandwidths must be made available to individual users. New technologies, concepts and
`ideas need to be discovered and developed so as to support these demands. This dissertation
`describes a new technology, holographic wavelength filtering, which may be a contributing
`factor to the continued advance and success of optical telecommunications.
`
`1.1 Wavelength Division Multiplexing
`
`Telecommunications is currently undergoing yet another revolution. From the middle of
`the last centw·y, when the digital electrical telegraph could send transatlantic messages
`at a data rate of about 1b/s (10° b/s), to the analogue Mb/s {106 b/s) Jinks of a. cen(cid:173)
`tmy later, through to the digital optical links, capable of transmitting Tb/s {1012 b/s)
`by the turn of the millennium, telecommunications has developed at an exponential rate.
`The major telecommunications technological advance in recent years, enabling Tb/s data.
`rates, has been the erbium-doped fibre amplifier (EDFA) [1) for passive broadband op(cid:173)
`tical amplification. This development has opened up the possibility of almost limitless
`
`1
`
`

`

`CHAPTER 1. INTRODUCTION
`
`1.2. TECHNOLOGIES FOR WDM
`
`bandwidth data pipes, and enabled the use of wavelength division multiplexing (WDM)
`to supply that bandwidth(2), or to allow new optically transparent network architectures
`such as wavelength-routed networks(3). The huge impact of this new technology on all
`levels of optical telecommunications was immediately apparent(4), and has been rapidly
`developed, so that within 10 years EDFAs are already to be found in significant com(cid:173)
`mercial telecommunications systems, such as the 2.5Gb/s optically amplified submarine
`transatlantic network TAT-12/13 [5]. A single-mode (SM) optical fibre, due to the mate(cid:173)
`rial properties of glass, has a number of wavelength 'windows' where low at,tenuation and
`dispersion at those wavelengths make them suitable for telecommunications usc. 'rhc win(cid:173)
`dow at the l.3Jtm wavelength is currently widely used, since it has zero dispersion and an
`attenuation of 0.35dB/km. However, another window at a central wavelength of 1.55p.m is
`becoming increasingly popular, since although it has a dispersion of about 12ps nm- 1km- 1
`with an attenuation of only 0.2dB/km, it has a useful bandwidth of about 40nm at which
`the optical gain of t.hc EDFA can compensate the fibre attenuation. This corresponds in
`the frequency domain to an available bandwidth of about 5THz. No single source can be
`currently, or is expected to be data-modulated much faster than lOOGb/s [6], so that most
`of the bandwidth is currently unused. At present there are two competing technologies
`being developed to enable that available bandwidth to be used most effectively. The first
`one, as mentioned before, is WDM which uses multiple carrier wavelengths on which to
`modulate the data in paraJlel. The second technology is optical time division multiplexing
`(OTDM) where very short (often less than lps) soliton pulses are produced by a source
`with a relatively 'low' repetition frequency (typically 40GHz). The outputs from multiple
`soliton sources, each temporally slightly out. of phase with respect to the other, can then be
`interleaved onto a single fibre to give a very high bit rate. The OTDM technology is less
`well developed than WDM, and tho highest data rate using OTDM is current.ly still only
`40Gb/s over 560km of SM fibre(7). The first WDM Terabit/second data transmissions down
`a single SM optical fibre have recently been reported(8](9) where 1.1 Tb/s (55 wavelengths
`x 20Gb/s) was transmit.ted over 160km of standard single-mode fibre. It can be expected
`that hybrid WDM/OTDM systems wiJI form the basis of future long-haul high-capacity
`optical telecommunications networks.
`
`1.2 Technologies for WDM
`
`The main thrust of this Ph.D. has been the development of a new generic technology, which
`can serve to underpin many of the diverse functional components required in a WDM
`telecommunications network. 'These WDM components, all of whose functionality requires
`a tunable wavelength filter in some form, include tunable sources and receivers, recon(cid:173)
`figurable optical amplifiers, space-wavelength switches and routers, as well as wavelength
`
`2
`
`

`

`CHAPTERl. INTRODUCTION
`
`1.2. TECHNOLOGIES FOR WDM
`
`converters. An ideal tunable wavelength filter would have a rectangular passband of width
`0.8nm with greater than 35dB sideband suppression(lO], a continuous and large tunable
`range across the erbium window, multiple wavelength tuning capability, fast tuning speed
`(measured in nanoseconds), low loss, optical transparency with polarisation insensitivity,
`low power consumption, compactness and reliability. Many competing technologies have
`been developed to provide tunable filters fulfilling some, but not all, of these requirements.
`It is hoped that dynamic holographic filtering will ultimately satisfy these demands. The
`most significant of the current technologies are outlined below.
`
`1.2.1 Acousto-optic Tunable Filt er
`
`The technology which has aroused the greatest interest to date has been the acousto(cid:173)
`optic tunable filter. Its distinguishing feature (like that of holographic filtering) is mul(cid:173)
`tiple wavelength operation(ll]. The first polarisation sensitive wavelength filter using
`polarisation-mode-coupling to enable tuning was an electro-optic device(12], quickly fol(cid:173)
`lowed by the acousto-optic equivalent[13]. The acousto-optic version was developed in
`favour of the electro-optic filter, since only it had the capability of multiple wavelength
`operation. Polarisation-insensitive wavelength filters soon appeared[14}[15}[16}[17]. Typ(cid:173)
`ical filter 3dB passbands of between lnm to 3nm have been achieved over a couple of
`hundred nanometres tuning range(18}[19]. Insertion losses of the order of 3dB have also
`been achieved(20]. Acousto-optic filters with near rectangular passbands have also been
`demonstrated(21}[22]. Acousto-optic filters have been used to flatten the gain spectrum of
`an EDFA [23], while a combined wavelength filter with gain using Er-doped LiNb03 has
`also been reported[24]. Actual multiple wavelength operation of an acousto-optic filter has
`also been demonstratcd(25] using 4 wavelengths spaced by 4nm, as has multiwavclength
`add-drop multiplexing(26], but this is where the limits of the technology begin to become
`apparent. Acousto-optic filters suft'er from high levels of crosstalk(27] and noise from a large
`variety of inherent sources:
`
`• imperfect polarisation splitting
`
`• interaction between the different RF surface acoustic waves (SAWs)
`
`• frequency shift of signal and time-dependent signal modulation due to RF wave
`
`• high sidelobes outside the passband
`
`The typical inter channel crosstalk is about -15dB, which is increased to -9dB when multiple
`( e.g. four) wavelengths arc being filtered[28]. The high sidelobes of the AOTF arc due to
`the sinc2 shape of the passband. A technique using tapered acoustic directional couplers
`has been developed to reduce the sidelobes, but the suppression is still only at best 15.5dB
`(20]. Other problems arc the minimum lldBm of power required for each RF SAW to
`
`3
`
`

`

`r
`
`CHAPTER 1. INTRODUCTION
`
`1.2. TECHNOLOGIES FOR WDM
`
`switch a channel(25], which limits the number of wavelengths which can be simultaneously
`tuned, before the device overheats. The LiNb03 substrate, which must. be of a high purity,
`also requires very highly controlled fabrication, so that the birefringence tolerance is of the
`order of w- 5 over its ent.ire length(27]. The passband width is inversely proportional to
`the device length, and achieving sufficient birefringence control over a length greater than
`20mm is difficult, so limit.ing the resolution to between l-2nm. Indeed very recent work
`suggests that AOTFs for use in wavelength routing switches will only work with channel
`separations 2-4nm (29](30]. This could be reduced down to 1.6nm by space and wavelength
`dilation(31](32], but at t.he expense of greater component count and multi-stage switch
`architectures. Dilat.ed switching architectures have achieved crosstalks of about -30dD [28].
`The switching speed of t.hcse devices is limited by the velocity of the acoustic wave across
`the surface of the LiNb03 substrate to about lOp.s [33].
`
`1.2.2 Fibre Fabr y-P erot Etalon
`
`Fibre Fabry-Perot (FFP) etalons are attractive since they offer very high resolution wave(cid:173)
`length filtering over relatively large ranges, in a low loss, low power, compact a.nd cheap
`device. They can only filter one wavelength at a time. Early devices used the piezoelectric
`effect to mechanically alter the cavity width and achieve tuning[34]. However, piezoelec(cid:173)
`trically controlled FFPs tend to have a slow switching time, measured in milliseconds.
`Rather than physically move apart the reflecting faces of the ctalon to achieve tuning, the
`preferred option now is to usc birefringent liquid crystal (LC) to electro-optically change
`the optical cavity width[35)[36]. Unfortunately LC in the cavity tends to make the filter
`polarisation sensitive, although various schemes have been adopted to attain polarisation
`diversity(37][38]. These polarisation insensitive LC-FFP filters have a 3dB passband of
`0.4-0.6nm and a Luning range of 50-70nm, so completely covering the erbium window, and
`insertion losses of a.bout 3dB. Tuning speeds of LC-FFPs are generally also in t.he millisec(cid:173)
`ond range, but less than 10JLS has been achieved[39], although at Lhc expense of Luning
`range and with polarisation sensitivity. A compound wavelength filter consisting of a FFP
`in series with an AOTF has also been reported with a tuning range of 130nm, and a 3dB
`passband of only 0.5A [40].
`
`1.2.3 Semiconductor F ilters
`
`Semiconductor wavelength fllt.crs can be compactly integrated with a photodiodc or used to
`couple the filtered signal directly into a SM fibre. They have the advantages of small size,
`low power and compatibility with other optoelectronic devices, but tend t.o be polarisat.ion
`sensitive and will only filter one wavelength at a time. A grating assisted vertical coupling
`filter fabricated ouL of InGaAsP /InP tunable over 37nm and with a 3dB passband of 5.5nm
`has been report.ed[41]. A filter with reduced polarisation sensitivity has also been reported,
`
`4
`
`

`

`CHAPTER 1. INTRODUCTION
`
`1.3. TUNABLE LASERS FOR WDM
`
`with a theoretical tuning range of llOnm and 3dB passband of 1.25nm [42].
`
`1.2.4 Mechanical Systems
`
`High resolution wavelength filtering by means of a diffraction grating and a spatial filter
`'I\ming is achieved by mechanically rotating the grating, but
`is a standard technique.
`this tends to be relatively slow and is also more prone to failure. However, high quality
`filtering is attained (one wavelength at a time), with low crosstalk. A space and wavelength
`switch has been demonstrated using a rotatable grating with 2 axes of rotation[43]. It
`has a 3dB passband of 2.5nm, crosstalk of better than -30dB, less than 3dD insertion loss
`and a tuning time of 4ms. A tunable wavelength wa·•ele);J;~Ht filter employing a rotating
`interference filter has been demonstrated[44]. It has a bandwidth of 0.5nm, a tuning range
`of 30nm, switching time of 2.5ms and less than -30dB crosstalk. It has an insertion loss of
`2.5dB, but is polarisation sensitive.
`
`1.2.5 Emerging Technologies
`
`Lithium niobate is best known in AOTFs to achieve wavelength switching, but another
`technique has been developed using it to achieve switching. Switcl1ing between the cross
`and bar branches of a LiNb03 directional coupler can be made to depend on the input
`intensity of the incoming light. A switching ratio of 1:5 and throughput of 80% has been
`reported[45]. Wavelength filtering using passive antiresonant reflecting optical waveguide
`(ARROW) couplers has been reported with a 3dB bandpass of 4nm [46] and an arrayed
`waveguide grating has achieved a 3dB passband of 0.3nm [47]. A comb ination of the two
`techniques in the future may allow passive self-routing or self-wavelength filtering.
`
`1.3 Tunable Lasers for WDM
`
`Tunable lasers find uses in many areas, apart from telecommunications, such as sensing,
`spectroscopy and general scientific research. A good tunable laser will have a large tuning
`range, fine tunability, stable lasing wavelength and perhaps has the flexibility of simultane(cid:173)
`ous multiple wavelength lasing. Other important attributes of a tunable laser are its output
`power, linewidth, sidemode suppression, modulation capability, speed of tuning, reliability,
`size and cost. All of these characteristics when taken together ultimately decide the type of
`tunable laser best suited to a particular task. As in the case of the tunable wavelength filters,
`there are many competing tunable laser technologies, all offering an optimum combination
`of these features. Semiconductor lasers currently form the main technology offering tunable
`somces for telecommunications, but rare-earth-doped fibre lasers are now also emerging as
`strong competitors .in this important area.
`
`5
`
`

`

`CHAPTER 1.
`
`INTRODUCTION
`
`1.3. TUNABLE LASERS FOR WDM
`
`1.3.1 Tunable F ibre Lasers
`
`The main features of fibre lasers are their very narrow linewidths and very closely-spaced
`longitudinal cavity modes, enabling almost continuous tuning, which is due to their rela(cid:173)
`tively long cavity lengths. High output powers and a broad tuning range arc also possible
`with fibre lasers. Thjs makes them ideal for many applications in WDM systems, such
`as stable tunable sources, modelocked sources for solitons and local oscillators in coherent
`detection schemes. Fibre lasers have the additional advantages that they can be pumped
`with efficient and compact laser diodes, and are compatible with fibre optic components,
`leading to Low coupling losses within telecommunications systems.
`
`1.3.2 External Cavity Tunable F ibre Lasers
`
`There arc a many different schemes to make tunable fibre lasers, but they can be basically
`divided up into external cavity free-space architectures and intra-cavity mechanisms. The
`intra-cavity fibre lasers have the slight advantage of a greater intrinsic mechanical stabil(cid:173)
`ity. The first tunable fibre lasers and erbium-doped fibre lasers {EDFLs) employed external
`rotatable gratings in a LiUrow configuration[48)[49). Early EDFLs had a bidirectional archi(cid:173)
`tecture, with the light passing up and down the erbium-doped fibre, but this caused standing
`waves and spatial hole burning with resulting multi longitudinal-moded behaviour. Most
`EDFLs now adopt a unidirectional ring structure which