fibre array. The extinction ratio between the signals going out at the output and the add ports is 17dB before and after the exposure. This value may be improved by optimising the coupling ratio of the directional couplers and by adjusting the phase between the signals propagating in the two arms of the Mach-Zehnder interferometer. The latter may be achieved by UV-trimming. The power back-reflected to the input port at the Bragg wavelength is also 17dB lower than the one exiting the drop port. This indicates that the Bragg gratings do not create significant phase variation between the signals reflected in the two arms and that W-trimming is therefore not necessary. , 1559.6, , , . , . , . , I60 1 E 1559.4 - c m 50q . 1559.2 - 40.: C .- U) 5 3 1559.0 ” -. 305E KOHNKE, G.E., MILBRODT, M.A., ERDOGAN, T., HENRY, c.H., STRASSER, T.A, WHITE, A.E., and LASKOWSKI, E.J.: ‘Planar waveguide Mach-Zehnder bandpass filter fabricated with single exposure W-induced gratings’. Proc. Optical Fiber Communication, San Jose, CA, 1996, pp. 277 5 KOHNKE, G.E., STRASSER, T.A., HENRY, C.H., LASKOWSKI, E.J., CAPPUZZO, M.A., and WHITE, A.E.: ‘Silica based Mach-Zehnder adddrop filter fabricated with UV-induced gratings’, Electron. Lett., 1996, 32, (17), pp. 1579-1580 6 RUNE, R.J., and JOERGENSEN, B.F.: ‘Impact Of coherent cross-talk on usable bandwidth of a grating-MZI based OADM’, submitted to IEEE Photonics Technol. Lett., 7 STRASSER, T A , CHANDONNET, P J., DEMARCO, J., SOCCOLICH, C.E., PEDRAZZANI, J.R., DIGIOVANNI, D.J., ANDREJCO, M.J., and SHENK, D.S : ‘UV-induced fiber grating OADM devices for efficient bandwidth utilization’. Optical Fiber Communication, San Jose, CA, 1996, Post-deadline paper 8 8 JOUANNO, J.-M , HUBNER, J., and KRISTENSEN, M.: ‘60-dB Bragg gratings in planar waveguides’. Proc. Optical Fiber Communication, Dallas, Texas, 1997, pp. 228-229 L 4- & 1558.8 1558.61 .-Q U 4 I 100 . I . I . I 200 300 400 annealing temperature,deg . I 500 4I 20 Fig. 3 Bragg wavelength and tvansmission dip of UV written Bvagg gratings against annealing tcmperature Spectral method applied to design of spotsize converters A. Vukovic, P. Sewell, T.M. Benson and P.C. Kendall The thermal behaviour of the Bragg gratings written in the planar waveguides was also investigated. As indicated previously, the wafer was first heated to 80°C for 10h. Then the sample was annealed for 2h at temperatures up to 500°C and the changes in Bragg wavelength and in transmission dip were recorded at room temperature after each annealing. Both the Bragg wavelength and the transmission dip are almost linearly decreasing with the annealing temperature as shown in Fig. 3. One of the critical issues for WDM components used in telecommunications is to achieve wavelengths in accordance with I.T.U. requirements. The dependence of the Bragg wavelength on the annealing temperature may be used in order to permanently tune the operating wavelength of the OADM. Conclusion: We present an optical add-drop multiplexer based on a symmetrical Mach-Zehnder interferometer with UV-written Bragg gratings in the arms. The device does not need a subsequent optimisation by UV-trimming. The -30dB bandwidth in transmission is 0.46nm, resulting in a positive BWU parameter. We thus achieved both a low inter-channel crosstalk and a low intra-channe1 crosstalk which is a significant progress in Mach-Zehnder based planar add-drop filters enabling use of such devices in multi-wavelength networks. 0 IEE 1997 Electronics Letters Online No: 19971437 20 August 1997 J.-M. Jouanno, D. Zauner and M. Kristensen (Mikroelektronik Centret, Technical University of Denmark, Bldg. 345 East, DK-2800 Lyngby, Denmark) References 1 VENGHAUS, H., GLADISCH, A., JnRGENSEN, B.F., JOUANNO, I.-M., KRISTENSEN,M., PEDERSEN, R.J., TESTA, F., TROMMER, D., and WEBER, J.-P.: ‘Optical addidrop multiplexers for WDM communication systems’. Proc. Optical Fiber Communication, Dallas, Texas, 1997, pp. 280-281 2 KASHYAP, R., MAXWELL, G.D., and AINSLIE, B.J.: ‘Laser-trimmed fourport bandpass filter fabricated in single-mode photosensitive Ge- doped planar waveguide’, IEEE Photonics Technol. Lett., 1993, 5, (2), pp. 191-194 3 HIBINO, Y., ALBERT, J., KITAGAWA, T., HILL, K.O., BILODEAU, F., MALO, B., and JOHNSON, D.c.: ‘Wavelength division multiplexer with photoinduced Bragg gratings fabricated in a planar-lightwavecircuit-type asymmetric Mach-Zehnder interferometer on Si’, IEEE Photonics Technol. Lett., 1996, 8, (l), pp. 8 4 8 6 ELECTRONICS LETTERS 4th December 1997 Vol. 33 Indexing terms: Integrated optics, Optical waveguides, Optical couplers A new spectral method is proposed and applied to the analysis and design of integrated optical spotsize transformers. Results presented are of the same accuracy as those from a finite difference method. The new method is simple, efficient and gives propagation constants and fields over a wide range of structures in seconds, making it well suited to the class of optical waveguide problems involving interacting rib and buried waveguides. Both rib and buried rectangular waveguides are widely used in integrated optical technology. One important application vertically couples these structures to produce spotsize transformers which allow more efficient coupling between semiconductor lasers and optical fibres. These function by tapering down the width of the small spotsize rib waveguide, which adiabatically forces the optical field into a large spotsize mode of the passive buried waveguide [1, 21. Accurate design of these spotsize transformers is of great importance as they offer improved coupling efficiency and reduced alignment sensitivity. Although numerical solvers such as finite differences, (FD) [3], and finite elements, (FE) [4], can be used for analysing these devices, they are often too computationally cumbersome for use within an interactive design environment. A new, computationally efficient and accurate spectral method for the analysis of coupled rib and buried waveguides is reported in this Letter. The method combines two previously developed methods, namely the spectral index (SI) method [5],used for the analysis of single and coupled rib waveguides, and the free space radiation mode (FSRM) method [6] used for buried waveguide analysis. Although both of these methods have proved highly successful in their own right, neither is individually suitable for the analysis of the composite structure described above. The new method takes advantage of the simplicity and speed of both SI and FSRM, to provide the propagation constants and the field profiles of the supermodes of the structure in a matter of seconds. These can then be used to optimise the design of devices such as spotsize transformers, making the new method of calculation timely. The accuracy of the new approach is confirmed by comparison with results from large scale FD calculations. Consider the coupled rib and buried waveguide structure illustrated in Fig. 1, for convenience divided into the separate regions shown. The air-semiconductor part of the rib structure may be simplified by use of an effective width and depth, W and K , respectively, which form the basis of the SI approximation and allow the polarised field in the rib to be expressed as [5] No.25 2121 (1) The conventional spectral index variational principle is applied at the base of the effective rib (j = 0), linking the solutions in the two regions above and below eqns. 1 and 2, giving a transcendental equation for p. The variational condition of the form: where where h is the operating wavelength, p is the waveguide propagation constant, D,are the amplitudes of the rib terms (to be determined) and N, is the number of terms necessary for convergence. IV Results: The new method is applied to a structure with dimensions H = 0.5pn, a = 1pn, d = 0.1pn, 2c = 4 ~with, refractive indices n, = 3.4382, n, = 3.1665, n3 = 3.2819 and h = 1 . 3 ~The . modal indices (pik,,) for the two quasi-TE supermodes supported by the structure are shown in Table 1, obtained from the present method and from a semi-vectorial FD method [3]. The structure supports two modes, the rib dominant mode (pr) and the slab dominant mode (p,). The results from the present method are obtained in a matter of seconds on a Pentium PC, whereas the FD method required significantly longer calculation time on a work station, particularly for the slab dominant mode which, by design, is close to cutoff and weakly confiied, thus requiring that many FD mesh parts are used. Example field profdes for the case with W = 0 . 8 ~ are shown in Fig. 2a and b, respectively. y 2c m n2 becomes a system of N, x N, equations, the non-trivial solutions of which yield the propagation constants of the composite structure. These procedures are straightforward to implement as a consequence of the spectral nature of both SI and FSRM methods. Fig. I Coupled rib and buried waveguide geometry In region 111, referred to for convenience as the 'buried slab region', the field is expressed as a superposition of the N, guided slab modes (supported by the symmetric waveguide of core width 2c) and a radiation field (unguided), which is assumed to exist in a uniform medium of a suitable refractive index, typically n2.This FSRM approach has proved valid for situations where the indices differ by no more than 10% [7],which is the case for many practical structures. The fields in regions I1 and N are simply expressed as superpositions of plane waves. Let E,(s),As), g(s) and C,(s) denote the Fourier transform, with respect to x,of the ith buried slab guided mode, the radiation spectrum on the planes y = -a' and y = <U' + d) and each rib term, respectively. Let A, and B, be the amplitudes of the ith slab mode on the planes y = -d and y = <a' + d).The propagation constant in the y direction for the plane waves in regions I1 and IV and also for the radiation field in region I11 is: y = (p2 - sg + s 2 ) 1 / 2 The propagation constants in the y direction for the slab waveguide modes are 7 2 = (P2 - P y2 where pzis the z th slab mode propagation constant. Therefore, the transformed field in region I1 can be written as 32612/ Fig. 2 Field profile (10% contours) of rib dominant mode and slab dominant mode for W = 0 . 8 (quasi-TE ~ mode) a Rib dominant mode b Slab dominant mode Table 1: Modal indices of two quasi-TE supermodes obtained with present method and with finite difference (FD) method where 6'= (a' + 4. The transformed field in region I11 is and finally, the field in region IV is [email protected](S,Y) [ g(s) N, 1 + CBzEt(') exp(y(y + b')) a=l (4) Enforcing continuity of tangential electric and magnetic fields at the planes y = -a', and y = -(U' + d),and requiring that the radiation field in region 111 is orthogonal to each guided slab mode, gives a set of 2N, equations from which A, and B, can be expressed in terms of the unknown constants D,. 2122 In conclusion, a new spectral method has been formulated for the analysis and design of mode spot tapers. The method is simple, fast and gives excellent agreement with benchmark FD solutions. Acknowledgment: The authors are grateful to BT Laboratories and EPSRC (Grant ref. GR/L28753) for financial support and to J. Heaton, DERA for helpful suggestions. ELECTRONICS LETTERS 4th December 1997 Vol. 33 No. 25 0 IEE 1997 16 September I997 Electronics Letters Online No: 19971405 A. Vukovic, P. Sewell, T.M. Benson and P.C. Kendall (Department of Electrical and Electronic Engineering, University of Nottingham, Nottingham NG7 2RD, United Kingdom) References SELTZER, c.P., RIVERS, L.J., HARLOW, M.J., and ‘Low threshold current 1.6 mm InGaAsPiInP tapered active layer multiquantum well laser with improved coupling to cleaved singlemode fibre’, Electron. Lett., 1994, 30, pp. 973-974 2 LEALMAN, I.F., SELTZER, c.P., RIVERS, L.J., HARLOW, M.J., and PERRIN, s.D.: ‘InGdAsP/InP tapered active layer multiquantum well laser with 1.8 dB coupling loss to cleaved singlemode fibre’, Electron. Lett., 1994, 30, pp. 1685-1687 1 LEALMAN, I.F., PERRIN, s.D.: 3 STERN, M.s.: ‘Semivectorial polarised finite difference method for optical waveguides with arbitrary index profile’, IEE Proc. J., 1988, 135, pp. 56&63 4 RAHMAN, B.M.A., and DAVIES, J.B.: ‘Finite-element solution in integrated optical waveguides’, IEEE J. Lightwave Technol., 1984, 2, pp. 682-688 5 MCILLROY P.W.A : ‘Spectral index method: Single rib waveguide’, in ROBSON, P.N., and KENDALL, P.C. (Eds.): ‘Rib waveguide theory by the spectral index method’ (John Wiley & Sons Inc., 1990),Chap. 5 6 REED, M., BENSON, T.M., SEWELL, P., KENDALL, P.c., BERRY, G.M., and DEWAR, s.v.: ‘Free space radiation mode analysis of rectangular dielectric waveguides’, Opt. Qia“mm Electron., 1996, 28, pp. 11751179 M.J.:‘Semiconductor laser facet reflectivity using free space radiation modes’, ZEE Proc. J., 1993, 140, pp. 49-55 ROBERTSON, cations. To demonstrate these improvements we report wavelength conversion of a 40Gbitis signal over a maximum of 24.6nm. MQ W SOA: The operating conditions of an SOA when used as a linear amplifier, and when used for nonlinear signal processing (such as FWM), are very different. In the linear regime, the carrier density is very high and, due to band filling effects, the wavelength of highest gain can be as much as 5Onm from the bandedge. However, for good FWM performance, the amplifier is very strongly saturated i.e. low carrier density. Thus, an MQW SOA with an active region optimised for linear applications will not be optimised for FWM, and vice versa. We have designed and fabricated an MQW SOA specifically for FWM. We use strained ternary MQW material for the active layer as it has a lower density of states than bulk or quaternary material. Thus, the inversion factor close to the bandedge will be higher, which results in a lower noise figure. For a given carrier density fluctuation, larger index fluctuations (which are beneficial to FWM efficiency) are produced when the linewidth enhancement factor (aH)is large. We have positioned the bandgap wavelength close to the operating wavelength (-1 550 nm) to obtain a large a,. This also ensures that the saturated gain slope at the operating point with respect to wavelength is either flat or negative, thus avoiding a strong reduction of the pump-signal power ratio as the beams propagate through the SOA. This could lead to data pattern dependent effects through gain inodulation. 25 c 20 -15 =/-’ -- mym i o -- 40Gbit/s wavelength conversion over 24.6nm using FWM in a semiconductor optical amplifier with an optimised MOW active region D -15 m 0 5- < a A.E. Kelly, D.D. Marcenac and D. Nesset Indexing terms: Semiconductor optical amplifiers, Opticalfrequency Introduction: Semiconductor optical amplifiers (SOAs) are likely to be key components in future optical fibre networks. This is because, as well as performing linear amplification, SOAs exhibit nonlinearities which have proven useful in numerous applications. An important nonlinear process is four wave mixing (FWM) in which a signal is mixed in the SOA with a ‘pump’ beam. This results in the generation of optical sidebands which can be isolated by filtering and used for demultiplexing [l], wavelength conversion [2] and dispersion compensation [3]. To avoid waveform distortion caused by gain modulation in the SOA, the pump power must be high in comparison to the signal. Furthermore, to reduce ASE noise, the amplifier should be operated in saturation. With these constraints, the FWM efficiency (and signal-noise ratio) can be too low for high bit rate systems. Hence, there has been a great deal of interest recently in improving the eficiency of the FWM process in SOAs. Optimisation of the SOA length has recently led to CW performance improvements [4] and 40 Gbitis transmission over 400km of standard fibre using mid-span spectral inversion [5]. Since the efficiency of the FWM process decreases rapidly with increasing pump probe detunings, the maximum possible wavelength translation is a good indication of device performance. Recently a record wavelength conversion range of 18nm has been demonstrated at lOGbit/s [6]. In this Letter, we report the improvements obtained when the SOA active region has been designed specifically for FWM appli- ELECTRONICS LETTERS 4th December 1997 Vol. 33 -I?:/ -20 .= i i ~ The MQW active region co;sists of seven periods of 40A unstrained ternary wells and 60A (-1 YOstrain) ternary barriers, resulting in a photoluminescence wavelength of 1.553pn. SOA devices were fabricated using buried heterostructure technology. A 2mm long device with multilayer antireflection coated facets was packaged with lensed fibres. Fig. 1 shows the fibre-to-fibre gain and the noise figure against current at 1.55pn. The device exhibited 20dB maximum fibre-to-fibre gain and a noise figure (2n,/c,) of 7.6dB which includes at least a 3dB coupling loss. The increase in the noise figure at currents 90mA is caused by residual reflections. 4OGbit/s OTDM system performance: To demonstrate the enhanced performance of the FWM optimised SOA, it was used to wavelength convert a 4OGbit/s, 27-1 PRBS, optical time division multiplexed (OTDM) data stream. Full details of the OTDM system can be found in [SI. The power penalty against wavelength shift at 10-9 bit error ratio (BER) is shown in Fig. 2. The circles denote wavelength downshifts with an input data signal wavelength of 1557nm and the squares denote wavelength upshifts with the data signal wavelength at 1543.5nm. Downshifts of up to 24.6nm and upshifts of up to 8.6nm are possible with BERs of < lVY, indeed a negative penalty is observed for wavelength downshifts of -9nm. The negative penalty is thought to be due to the No. 25 2123

1/--страниц