Challenges in Optoelectronic Packaging for High Performance WDM Networks

Erik C.M. Pennings

Philips Optoelectronics, WAD-I, Pro$ Holstlaan 4, 5656 AA Eindhoven, The Netherlands Phone: +3I -40-2743037/ Fax: +31-40-2743859 /E-mail: epenning@natlab. research.philips.com

NETWORK EVOLUTION

1995 2000

I. Introduction It is very exciting that, after a long time of research, optical

communication systems based on wavelength division multiplexing (WDM) have now been commercialized and that the market for such systems is growing at a breathtaking pace in order to meet the ever increasing demand for bandwidth. This development presents various challenges to systems and component suppliers, e.g. rapid increase of production capacity, timely development of new components, and addressing specific new requirements that are imposed by WDM technology.

It is the purpose of this paper to investigate the challenges that are posed to optoelectronic packaging by WDM. The main thrust of this paper is to review two key components of WDM systems, i.e. the wavelength (de)multiplexer (Section V) and the WDM laser source (Section VI), and to compare an integrated optoelectronic approach versus a hybrid or modular approach. This allows us to compare different alternatives and provides insight into the relevant packaging issues as well. The paper starts with a review of network and component trends (Section 11), an analysis of the WDM market (Section 111), and a discussion on optoelectronic integration issues (Section IV). Conclusions are presented in Section VII.

11. Network and component trends Figure 1 shows the expected component trend in relation to

the evolution of optical transmission systems. Until the advent of OFDM (Optical Frequency Division Multiplexing techniques, i.e. WDM), transmission capacity could only be increased by employing more fibers or by increasing the bit- rate. Installing new fibers is very costly and increasing the bit- rate beyond 2.5Gbps not only requires new high-speed electronic and optical components, but also runs into dispersion limits. The dispersion of 17ps/nm/km @ 1550nm in standard non-dispersion shifted singe-mode fiber limits the

COMPONENT TRENDS

direct detection 1 2 . 5 G Y lOGbp; 40Gbps

16x2.5Gbps OFDM trunk 4x2.5Gbps 4xlOGbps

increased speed dispersion reduction optical am lifiers couplers - R'@?jmux sets of (tuna le lasers tunable filters 1

transmission distances well below l O O k m @ 10Gbps. This limit has initiated a drive to develop dispersion compensation techniques and very low-chirp laser sources.

These problems can be circumvented by using a bit-rate of 2.5Gbps and by multiplexing a number of different wavelengths in order to increase the transmission capacity as shown in Figure 2. Key components in WDM system are wavelength specific laser sources, Erbium doped fiber amplifiers (EDFA), and dense wavelength multiplexers and demultiplexers. The EDFA significantly increases the cost- effectiveness of a WDM system, since conventional repeaters require separate regeneration of each individual wavelength channel. For these reasons, WDM has first been used to increase the capacity of very long-haul point-to-point transmission systems (typ. 600km).

Several additional stages of WDM deployment can be expected, namely one where the wavelength domain is used to provide routing capability, and one where WDM is used in the access part of the network. There will be a fast introduction of optical add-drop-multiplexers (OADM) (as part of a point-to- point transmission system), with a later extension to optical cross-connects (OXC) since the attractiveness of OXC prerequires a significant installed base of WDM systems. OADM and OXC's require space-switches and wavelength (de)multiplexers. At a later stage, wavelength tunability an$ wavelength conversion may be used to provide increased levels of optical circuit or packet switchtng. This requires (fast) tunable lasers, wavelength converters, and tunable filters. WDM will also be deployed in parts of the network other than the trunk. In a first stage, WDM systems may be used for shorter links or rings, interfacing to the long-haul systems via an add-drop multiplexer for example. At a later stage WDM can also be used for the access network, which will put a price pressure on components and systems.

data in data out

Fig. 1. Expected component trend in relation to optical system evolution. In order to meet the increasing demand for bandwidth, transmission systems using OFDM (optical frequency division multiplexing) are now being deployed in direct competition with single channel direct-detection systems with increased speed such as 10Gbps. Regarding OFDM, several distinct waves of deployment can be expected.

Fig. 2. Generic layout of an optical long-haul point to point transmission system employing wavelength division multiplexing. Key components are wavelength specific light sources, Erbium doped fiber amplifiers, and dense wavelength multiplexers and demultiplexers.

0-7803-3857-Xl97 $4.00 01 997 IEEE 593 1997 Electronic Components and Technology Conference

0.3% Wavelength converters 8.2% WDM filter modules 9.1% Integr. Optoelectronics

20.6% Multichannel receivers

12.18$ 1995 2000 2005

Fig. 3. Global market for WDM components. (Source: ElectroniCast). Very strong WDM market growth is seen between 1995 and 2000. Integrated optoelectronics is forecasted to grow significantly between 2000 and 2005.

111. The WDM component market place Figure 3 shows the global WDM component market

forecast for the 1995 to 2005 time frame. A distinction has been made between multichannel transmitters and receivers, optical amplifiers, integrated optoelectronics, WDM filter modules, and wavelength converters. Very strong growth of more than 100% per year is encountered between 1995 and 2000, leveling off to the overall (non-WDM) market growth rate of 25% per year between 2000 and 2005. Some relative shifts can be observed, such as a replacement of transmitters (repeaters) by EDFA’s between 1995 and 2000, and a strong growth of integrated optoelectronics at the expense of some other segments (such as amplifiers and receivers) between 2000 and 2005.

The expectation that integrated optoelectronics will grow in the second phase is caused a) by technological issues that have slowed down commercialization, and b) by a lack of market opportunities so far. Conventional systems require components with a single optical functionality, whereas the advantage of integrated optoelectronics is precisely in integrating complex optical functionality in a single chip. As mentioned in Section 11, the first WDM transmission systems will merely require optical devices with a single functionality, so that the demand for complex optical functionality will only increase later on, e.g. when the wavelength domain is used to realize add-drop multiplexing and cross-connecting functions.

Iv. Technological considerations regarding optoelectronic integration

It has been argued many times that the competitive edge of photonic integrated circuits (PIC’S) comes from increased scale of integration plus the corresponding reduction in packaging costs. This argument is in fact based on an implicit analogy between photonic and electronic IC’s. This analogy, however, has to be treated with care. Firstly, the markets for photonic and electronic IC’s are very different. Secondly, there are several technical reasons why photonic and electronic IC’s are quite different:

Packaging is much more complex for photonic than for electronic IC’s. Packaging related issues are fiber-chip coupling which leads to the use of tapers and lenses, the

influence of reflections and the accompanying use of isolators, and temperature sensitivity which can necessitate Peltier coolers and thermistors. It is important to notice that these packaging issues form, on one hand, the major economic incentive for integration, but that they, at the same time, form a technical obstacle against integration. For example, the very same reflection-sensitivity of lasers and semiconductor optical amplifiers that leads to the use of isolators in the first place, also makes it difficult to actually integrate a laser and a semiconductor booster amplifier on the same chip while maintaining the same performance as the hybrid solution.

The scale of integration seen in electronic IC’s is directly related to intrinsic on-chip amplification and feedback which allow for accurate performance control and thus for the ability to cascade a large number of components. So far, optical feedback is not feasible in photonic IC’s, and integrated optical amplifiers are in their infancy. Even more important is that the characteristic size of the building blocks for photonic and electronic IC’s differ by at least two to three orders of magnitude. Finally, whereas electronic IC’s contain many duplicates of a small set of building blocks, most photonic IC’s integrate a number of very different elements. The accompanying problem of optimizing the performance of each individual sub-element complicates the design and may lead to compromises in performance.

As a result, photonic IC’s will follow their own rule of economy, different from electronic IC’s, and will reach a more limited scale of integration. In order to reach that limit and speed up commercialization, additional effort will be required to reduce packaging costs, incorporate on-chip amplification, reduce component size, and to develop improved processes and fabrication-tolerant components.

V. Wavelength demultiplexers for dense WDM applica- tions: a comparison

In this section, a specific comparison between different technologies is made for the wavelength (de)multiplexer for dense WDM (see also Ref. 1). Although some differences may exist between multiplexers and demultiplexers, they will be referred to as a single group “(de)multiplexers” in this paper. For the comparison of different techniques, the following classification is used:

Micro-optic devices, which rely on diffractive or reflective bulk elements such as lenses or mirrors. Integrated-optic devices, where light is guided in planar waveguides. - Single-component devices, such as lasers, semiconductor optical amplifiers, and phase-modulators. - Photonic integrated circuits, where a number of optical devices are monolithically integrated. Fiber-based devices, such as couplers, that are exclusively made from fiber. Hybrids or modules, which are assembled from any of the above elements. An overview of published wavelength (de)multiplexers and -

their development in time can be seen in Figure 4.

594 1997 Electronic Components and Technology Conference

_ _ _ _ ~ ~ ~-

Jobin-Yvon JDSlFitel Ph.Opt.Corp. OCA

micro-optical: 0 integr. optic (passive): VDA

Ph.Opt.Corp. BTRL / / Hitachi/A~-

amplifier/demux/diodes + Siemens> / / demux/diodes + laserlmux + / kK

. - --.- ~ N T T .~ / phased array --+

. .T&T Bath/BNR - - - - - - - -

- - - - ~ ~ ! x d f i Bellcore

grating (reflective) + STC d

Jobin-Yvon grating (transmissive) - Bellcore Siemens A..-,,- - - uaana - - -

cascaded duplexers + VAT&T V Boeing I I 1 I I I I I I I I I I I

1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996

___) time

Fig. 4. Overview of published wavelength (de)multiplexers, their integration with other functionalities, packaged devices, system experiments and commend availability as a function of time.

Micro-optics

Figure 4 shows that much of the work on micro-optical demultiplexers was already published in the early e i g h t i e ~ ~ - ~ when MM fibers were used with wavelengths of k700- 900nm. The subsequent reduction of interest was due to the shift of focus to coherent detection schemes. In the late eighties, when interest in WDM revived, work on micro- optical demultiplexers continued6'" and several earlier designs were commercialized. Micro-optical wavelength demultiplexers can be divided into the cascaded interference filter type and the grating-based type. Recently, the phased array design, which is typical for integrated optics, has also been realized in micro-optics''.

Integrated optic

The first integrated optic demultiplexers appeared in the late eighties and relied on cascaded d~plexers '~- '~ . Soon afterward, planar versions of the micro-optic grating based demultiplexer were r e p ~ r t e d ' ~ - ' ~ ~ * ~ - ~ ~ . An alternative design is the phased array (PHASAR) or arrayed waveguide grating (AWG)22-29. This design has recently gained in importance due to superior performance and ease of processing. Figure 4 shows how integrated-optic demultiplexers have subsequently been integrated into more complex PIC'S (such as

and an add/drop multiplexer3*) and how they have been used in various system experiment^^^.^^-^. It appears that this development is accelerating which is, for example, demonstrated by the rapid c~mmercialization~~ of the Silica-based demultiplexers.

Fiber- based The lack of fiber-based demultiplexers as shown in Figure 4

illustrates the unsuitability of fibers to demultiplex many closely spaced wavelengths simultaneously. Fibers are, on the other hand, very suitable to demultiplex two widely spaced wavelengths (i.e. duplexers), which are typically used for two- channel transmission systems (e.g. 1.31/1.55pm), or in EDFA's (0.98/1.55pm and 1.4811.55pm). It is important to notice that duplexers form more than 90% of the demultiplexer market and are almost exclusively fiber-based.

Modules Fibers can, of course, be used in combination withfilters to

realize wavelength demultiplexers for dense WDM. These filters can be placed in series (cascaded) or in parallel by using a fiber splitter. Popular filters are the (fixed) interference filter46, fiber Bragg gratings47, and the tunable (fiber) Fabry- P6rot filter.

Discussion and conclusion The performance of a variety of commercial demultiplexers

is compared in Table 1. The splitter plus filter configuration provides the advantage of simplicity and tunability, but poses an intrinsic splitter loss of lO.log(N) and may show unwanted back-reflections. Fixed demultiplexers gain in importance, as soon as WDM wavelength channels have been standardized. For cascaded filters, the loss increases with the number of channels, which, so far, has limited their suitability to 4-8 channels. Micro-optic demultiplexers offer proven reliability in addition to excellent performance in terms of number of channels, insertion loss, cross-talk, polarization-independence

595 1997 Electronic Components and Technology Conference

Technique

Module

Ins.Loss Ret.Los X-talk Drift Pol. Dep. Vendor (dB) s(dB) (-dB) (nm/'C) (dB)

<1.5-2.5 ? ? Act. tuning < O S Micron Optics

Cascaded filter 4 . 5 >40-55 >15 0.004 ? OCA

M (nm)

N 5 P e

Splitter + FFP-filter

<4 >22-40 ? Act. tuning ? Queensgate

and thermal stability4*. Integrated optic demultiplexers, especially the phased array type, have seen a rapid development and the recently commercialized Silica-based demultiplexers show a very competitive performance. The performance of these commercial versions can, however, not yet fully match that of micro-optic versions.

Rather than comparing (de)multiplexers at the one- component level, the competitive advantage of photonic IC's is in integrating increased optical functionality and achieving a larger scale of integration, thus offsetting the packaging costs. In order to achieve that, however, the size of integrated (de)multiplexers needs to be decreased.

A specific WDM requirement is that of wavelength (or temperature) stability. Unlike micro-optic (de)multiplexers, silica-based (de)multiplexers experience temperature drift, so that they require either a temperature stabilized package or that the package needs to be equipped with a Peltier cooler and a thermistor.

VI. Laser sources for dense WDM: a comparison. Since the cost-effectiveness of WDM systems is greatly

increased when EDFA's are used, WDM systems have, so far, targeted ultra-long distances (typically up to 600km). The frequency chirping that occurs in directly modulated lasers makes them unsuitable for these distances, so that very low- chirp laser sources are req~ired~'?~', such as a CW laser with an external LiNbO, Mach-Zehnder modulator (MZ) or a DFB laser with an integrated electro-absorption modulator (EA). As can be seen from Figure 5, the comparison is actually not so much between these two alternatives, but much more one between different types of modulators (e.g. MZ versus EA) on one hand and one of packaging options (e.g. modular versus integrated) on the other hand. In this section, we will first address generic issues that are required for any WDM laser source, then review different high-speed modulator types, and conclude by comparing different packaging alternatives.

Generic WDM laser requirements The key characteristic of any WDM laser source is the

(vacuum) wavelength ho. Since it is not possible to

Micro- Optic

1ntegr.- Optic

predetermine the laser emitting wavelength with the desired accuracy (=O.Olnm), the temperature dependence of the DFB laser (=O.lnm/"C) is used to tune and subsequently lock the laser to the required wavelength channel hc (ITU recommends a lOOGHz grid relative to a reference frequency of 193.1THz). This implies the need a) to accurately specify, measure, and monitor the wavelength, b) to define a total temperature tuning range and c) to address wavelength drift.

The tuning requirement of the laser implies that any WDM laser source requires a Peltier cooler for temperature controlling and a thermistor for temperature monitoring. The thermal design (soldering, placement of thermistor relative to the laser chip, etc.) becomes important, which can be expressed by means of the wavelength tracking error, i.e. the total wavelength change due to case temperature variations while keeping the thermistor value constant. A typical value for the tracking error is 0.001 to 0.002nm/"C. Since the wavelength depends on temperature and current, proper specification of the wavelength includes the thermistor value and the operating current. Typical dependencies are 0.2nmIkQ @ 1OkQ for the thermistor and 0.002 to O.OOSnm/mA for the operating current. Since the lasing wavelength is sensitive to reflections, an isolator is required in the package, where the isolator preferably should be cooled since the isolation value is temperature dependent.

The total temperature range which can be used for wavelength tuning needs to be sufficiently large so that the respective desired wavelength channels can be reached within this range. Increasing the range will enlarge the wavelength selection yield. At the same time reliability requirements exert a downward pressure on the upper limit, whereas operating temperature requirements exert an upward pressure on the lower limit. A typical temperature tuning range is 20-40°C.

Another WDM laser parameter is the wavelength drift. So far, only a few investigations on wavelength drift have been p u b l i ~ h e d ~ ~ j ~ ~ . Wavelength drift appears to be temperature dependent52 with activation energies of 0.4 to 0.6eV, allowing accelerated drift experiments to be performed. At Philips Optoelectronics, wavelength drift experiments have been

Cascaded filter 4-8 1.6 <3-4 AO-55 >30 0.004 <o. 1 OCA 1.6-8.0 <2-4 >55 0.005 0.1 typ. DiCon Cascaded filter 4-8

8 ? <IO >25 >20 ? < I Hitachi Phased Array 8 1.61 <10 >30 >20 0.01 (tuning) ? Lucent (Silica-based) 4-32 0.8-2 <6-7 >40 >20-22 PeltierlNTC <0.3 NTT(NEL)

Grating 4-4 1 1-16 <3-5 ? 230-55 0.02-0.004 ? Jobin-Y von

4-32 0.8-2 <7-9 >3 5 >22 PeltierNC <0.3 PIRI

596 1997 Electronic Components and Technology Conference

1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 - time

20 Gbps

10 Gbps

5 Gbps

2.5 Gbps

40 Gbps 20 Gbps 10Gbps 2 2.5Gbps .g 40 Gbps 20 Gbps 10 Gbps 2.5 Gbps DC

DC 9

f

Fig. 5. Overview of reported high-speed modulators, their integration with DFB lasers, their integration into more advanced PIC’S, systems experiments, and commercial availability as a funtion of time. The numbers indicated next to the published system experiments give the total length of standard (non-dispersion- shifted) single-mode fiber that was covered without using dispersion-compensation techniques.

conducted to warrant wavelength stability. Wavelength drift can be positive as well as negative, which contradicts a single aging-based explanation and rather indicates that there are two opposing drift mechanisms. Rather than by the chip itself, wavelength drift seems to be caused by the package (such as thermal issues or reflection). Wavelength drift is not linear with time, and stabilization has been observed after several 1000 hours of accelerated operation. This indicates that wavelength drift should be treated as a reliability issue, e.g. by specifying a maximum end-of-life drift value and to perform mean-time-to- failure (MTTF) calculations. A target value is half an ITU channel spacing (0.4nm) for a 20 year laser life resulting in a maximum temperature tuning drift of E 4°C over the laser life.

High-speed modulators As shown in Figure 5, high-speed modulators can be

divided into two main categories: the Mach-Zehnder interferometer and the electro-absorption modulator.

Mach-Zehnder modulators were pioneered by L ~ c e n t ~ ~ - ~ ~ , B e l l ~ o r e ~ ~ , F u j i t ~ u ~ ~ - ~ ’ , NTT6l, and Sumitomo6*. Using electro-optic LiNb03, very high 3dB bandwidths of up to 50 and 75GHz have been r e p ~ r t e d ~ ~ . ~ ~ . The interest in polymer Mach-Zehnder modulator^^^^^' results from their potenbal to achieve even higher bandwidths. Publications on LiNbO modulators were soon followed by systems experiment^^^^^^,^' and by wide-spread commercialization by various suppliers. The sensitivity of the electrode voltages of LiNb03 modulators to temperature and drift has led to various reliability investigation^^^,^^. LiNb03 modulators have been integrated into more complex PIC’S, such as an optical time-domain

multiplexer7’ and with an Er-doped LiNb03 laser’’. The advantage of InP-based Mach-Zehnder modulators (e.g. by

Lu~ent’~-’~~’~, Hita~hi”,’~, N~rtel’~~~’, and Alcatel”) is their potential for monolithic integration with DFB lasers (or with other components such as DBR-la~ers’~ or amplifier^^^). Using InP-based Mach-Zehnder modulators various lOGbps transmission experiments have been r e p ~ r t e d ~ ~ , ~ ’ , ~ ~ covering distances up to 133kmS1. The advantage of Mach-Zehnder modulators in general is that they are virtually chirp-less and even can be designed to have a fixed negative chirp82 leading to improved transmission performance. Disadvantages are the high driving voltages (4 to 15V range) that are incompatible with standard low-voltage (<2V) driving conditions.

Electro-absorption modulators are based on a voltage induced shift of the bandgap so that the modulator becomes absorbing for the lasing wavelength. Three different effects can be utilized, i.e. the Franz-Keldysh effect in bulk semiconductor material, the quantum-confined Stark effect (QCSE) in MQW’s, and the Wannier-Stark effect in superlattices. The strong Wannier-Stark effect leads to very low driving v ~ l t a g e s ~ ~ ~ ~ ~ , ~ ~ , but is unsuitable for high output powers due to saturation effects. The trade-offs between the QCSE effect using MQW and the Franz-Keldysh effect are that QCSE leads to lower drive voltages, but may show more pronounced saturation effects and may be more critical in terms of bias and driving conditions. Important advantages of electro-absorption modulators in general are their low driving voltages, high- speed operation, and suitability for integration with InP -based lasers. First electro-absorption modulators and their integration

597 1997 Electronic Components and Technology Conference

with DFB lasers were reported by L u ~ e n t ~ ~ - ~ ~ , KDD90-92, F u j i t ~ u ~ ~ ~ ~ ~ and NTT98-'04, subsequently followed by Hitachi 105- 109 , Nortell lo, Alcate111'-"5, Ericsson116,117,

Electro-absorption modulators have been integrated with amplifier^'^^, DBR lasers131, and into optical time domain multiplexer^'^^. Integrated laser-modulators have been integrated monitor photo diode^'^^, and a second m o d ~ l a t o r ' ~ ~ , ' ~ ~ for short pulse generation. A substantial number of system experiments have been conducted at 2.5Gbps with distances up to 1291km114 and at lOGbps with distances up to 150km' l9 using standard non-dispersion shifted single-mode fiber. Transmission experiments using packaged devices have demonstrated suitability for bit-rates up to 20Gbps102,'27 and 40Gbpslo3.

Packaging issues

Few direct comparisons have been published between the system performance of LiNb03 based MZ modulators and InP- based integrated laser -mod~la tors~~~. Since there do not appear to be major differences in performance, the issue may not necessarily be one of MZ versus EA modulators, but much more one of packaging, i.e. one of monolithic integration versus a hybrid or modular approach. Actually, three options exist: hybrid (separate), monolithic integration, and a single dual-chip package.

Since commercial LiNb03 modulators are polarization sensitive, polarization maintaining fiber (PMF) is required for the hybrid solution. Achieving and maintaining a high polarization extinction ratio thus become critical. The PMF fiber is more expensive, time-consuming to align and splice, and different incompatible PMF fiber types exist. The hybrid approach thus seems bound to be less cost-effective than the integrated approach, unless the packaging costs can be considerably reduced. Although the modulator presents an additional insertion loss of 4 to 5dB, in practice this can be compensated for by using high-power CW lasers.

Monolithic integration is commonly expected to be the most cost-effective approach. The modulator coupling loss is eliminated, the source is more compact, the pinning is compatible with directly-modulated lasers, and the use of PMF is avoided. The trade-off, however, is that the potential to reduce packaging costs on one hand may be offset by the fact that integration of different functionalities in a single PIC may lead to a compromise in performance or yield. For example, several forms of cross-talk can arise. Electrical cross-talk can occur between the modulator and the laser (on the chip or in the package). Optical cross-talk takes the form of modulated reflection^'^^ back into the laser, that can lead to chirp and degrade performance. Thermal cross-talk can result, e.g. when the absorption of light in the modulator heats the chip138 thereby leading to wavelength changes. Another form of thermal cross-talk is caused by the fact that the temperature tuning used to lock the laser wavelength to a specific channel also affects the modulator and that the temperature dependencies differ for the laser and the modulator. The change in lasing wavelength is determined by variation of the refractive index with the temperature 6n/6T (about 0. lnm/"C),

philips 1 18- 120 NEC 121,122 ~ m ~ l 2 3 - I 2 7 and 0 ~ 1 2 8 , 1 2 9 9 9

598

whereas the variation of the modulator is determined by the dependence of the bandgap on temperature 6Eg/6T (about 0.4nd"C). With proper design this does not need to be a problem: in one experiment, a total temperature tuning range has been reported of 3.2nm for modulator integrated DFB

139 laser . The disadvantages of the hybrid and the monolithic

approach may be avoided altogether by including a separate modulator and a laser chip in a single dual-chip BTF package. This may present a packaging challenge both technically and in terms of the ability to do so cost effectively.

VII. Conclusions The commercialization of WDM technology provides an

exciting opportunity for the optoelectronic community, just as well as a challenge in terms of speed of production capacity increase, development of new components, and meeting specific new requirements set by WDM.

In this paper, different fabrication techniques have been compared for two key components in WDM networks, namely the wavelength (de)multiplexer and the laser source. The two comparisons reveal several similarities.

Because WDM provides an increased need for optical functionality, there is a clear opportunity for optoelectronic integration, but this has to be compared with a hybrid approach, where the trade-offs are compact size and a potential reduction of the packaging costs on one hand versus the difficulty of simultaneous optimization, the associated cross- talks, and yield, on the other hand.

One of the key issues for WDM in general is that of wavelength stability and tuning, which translates directly into temperature stability and tuning. Eliminating reflections and improving thermal behavior and control may therefore prove crucial.

VIII. References E.C.M. Pennings et al., J. Sel. Top. Quantum Electron., vol. 2, no. 2,

M. Seki et al., Electron. Lett., vol. 18, no. 6, pp. 257-258, 1982. J.P. Laude et al., Proc. qh ECOC '83, (Oct. 23-26, 1983, Geneva, Switzerland), pp. 417-420. J. Hegarty et al., Electron. Lett., vol. 20, no. 17, pp. 685-686, 1984. J. Lipson et al., J. Lightw. Technol., vol. 3, no. 5, pp. 1159-1 162, 1985. Y. Fujii et al., Appl. Opt., vol. 28, no. 7. pp. 1305-1308, 1989. B. Moslehi etul., Opt. Lett., vol. 14, no. 19, pp. 1088-1090, 1989. G.J. Cannel1 et al., IEEE J. Sel. Areus. in Cr~mm., vol. 8, no. 6, pp.

D.R. Wisely etal . , Electron. Lett., vol. 27, no. 6, pp. 520-521, 1991. K.J. Hood et al., J. Lightw. Technol., vol. 11, no. 516, pp. 680-687, 1993. M . Shirasaki, Proc. MOC'95, (Oct. 18-20, 1995, Hiroshima, Japan), paper PD3. B.H. Verbeek et al., J. Lightw. Technol., vol. 6, no. 6, pp. 1011-1015, 1988. J.P. Lin et al., Opt. Lett., vol. 16, no. 7, pp. 473-475, 1991. M. Gibbon etal . , Electron. Lett., vol. 25, no. 21, pp. 1441-1442, 1989. S. Uraetal., Appl. Opt., vol. 29, no. 9, pp. 1369-1373, 1990. C.H. Henry et al., J. Lightw Technol., vol. 8, no. 5, pp. 748-755, 1990. J.B.D. Soole etal . , Electron. Lett., vol. 27, no. 2, pp. 132-134, 1991. C. Cremer et aL, Appl. Phys. Lett., pp. 627-629, 1991 P.C. Clemens et al., IEEE Photon. Technol. Left., vol. 4, no. 8, pp. 886- 887,1992. J.G. Bauer et al., Proc. ECOC'94 (Sept. 25-29, 1994, Florence, Italy),

pp. 151-164, 1996.

1141-1145,1990.

pp.751-754.

1997 Electronic Components and Technology Conference

[231 [241

[441

E. Gini et al., Proc. ECOC’95, (Sept. 17-21, 1995, Brussels, Belgium)

A.R. Vellekoop et al., Proc. ECOISA’89, (Sept. 25-28, 1989, Amster- dam, The Netherlands), paper D3. H. Takahashi et al., Electron. Lett., vol. 26, no. 2, pp. 87-88, 1990. C. Dragone, IEEE Photon. Technol. Lett., vol. 3, no. 10, pp. 896-899, 1991. M. Zirngibl et al., Electron. Left., vol. 29, no. 2, pp, 201-202, 1993. B.H. Verbeek et al., Proc. OFC’94, (Febr. 20-25, 1994, San Jose, USA), post-deadline paper PDP13. H. Bissessur et al., Electron. Lett., vol. 30, no. 4, pp. 336-337, 1994. M.R. Amersfoort et af., Proc. IPR’95, (Febr. 23-25, 1995, Dana Point, USA), post-deadline paper PD3. P.C. Clemens et al., Proc. 7‘* ECI0’95, (April 3-6, 1995, Delft, The Netherlands), pp. 505-508. J.B.D. Soole et al., Electron. Lett., vol. 28, no. 19, pp. 1805-1807, 1992. M. Zirngibl et al., Electron. Lett., vol. 30, no. 9, pp. 701-102, 1994. M. Asghari et ul., Electron. Lett., vol. 30, no. 20, pp. 1674-1675, 1994. C. Cremer et al., IEEE Photon. Technol. Lett., vol. 4, no. 1, pp. 108- 1 IO, 1992. J.B.D. Soole et al., Electron. Lett., vol. 29, no. 6, pp. 558-560, 1993. C. Cremer et ul., Proc. ECIO’93, (April 18-22, 1993, Neuchiitel, Swit- zerland), p. 2-10. M.R. Amersfoort et al., Proc. ECOC’93, (Sept. 12-16, 1993, Mon- treux, Switzerland), post-deadline, pp. 49-52. C.A.M. Steenbergen et al., Priic. ECOC’95, (Sept. 17-21, 1995, Brus- sels, Belgium) pp. 211-214. K. Okamoto et al., Pmc. OFC’95, (Febr. 26 - Mar. 3, 1995, San Diego, USA), paper PDlO. Y. Tachikawa et al., Electron. Lett., vol. 29, no. 24, pp. 2133-2134, 1993. S . Suzuki et al., Electron. Lett., vol. 30, no. 13, pp. 1091-1092, 1994. 0. Ishida et al., Proc. OFC’95, (Febr. 26 - Mar. 3, 1995, San Diego, USA), paper PD9. B.R. Hemenway et al., Proc. OFC’95, (Febr. 26 - Mar. 3, 1995, San Diego, USA), paper PD8. L.H. Spiekman et al., Priic. ECOC’95, (Sept. 17-21, 1995, Brussels, Belgium) pp. 1055-1058. K. Oda et al., Proc. ECOC’95, (Sept. 17-21, 1995, Brussels, Belgium)

H. Uetsuka et al., €“e. OFC’95, (Febr. 26 - Mar. 3, 1995, San Diego, USA), paper Tu07. L. Eskildsen et al., IEEE Photon. Technol. Lett., vol. 6, no. 11, pp.

G.P. Agrawal, and S. Radic, IEEE Photon. Technol. Lett., vol. 6, no. 8,

M. Miyachi et al., Proc. ECOC’95, (Sept. 17-21, 1995, Brussels, Bel- gium) pp. 67-70. F. Koyama and K. Iga, J. Lightw. Technol., vol. 6, no. 1, pp. 87-93, 1988. J.A.J. Fells et al., IEEE Photon. Techncil. Lett., vol. 7, no. 11, pp. 1279- 1281, 1995. R.S. Vodhanel et al., Proc. OFC’94, (Febr. 20-25, 1994, San Jose, USA), paper WG5. Y.C. Chung et al., Proc. OFC’94, (Febr. 20-25, 1994, San Jose, USA), paper WG6. S.K. Korotky et al., J. Lightw. Technol., vol. LT-3, pp. 1-6, 1985. S.K. Korotky et al. Appl. Phys. Lett., vol. 50, no. 23, pp. 1631-1633, 1987. S.K. Korotky et al., J. Lightw. Technol., vol. LT-3, no. 5, pp. 1027- 1031, 1985. A.H. Gnauck et al., Proc. OFC’86, (Feb. 24-26, Atlanta, USA, 1986), post-deadline paper PD9. J.L. Jackel et al., J. Lighhv. Technol., vol. 7, no. 6, pp. 937-940, 1989. M. Seino et al., Priic. 15th ECOC’89, (Sept. 10-14, 1989, Gothenburg, Sweden), paper ThB22-5. M. Seino et al., Priic. 16th ECOC, (Sept. 16-20, 1990, Amsterdam, Netherlands), post-deadline paper, pp. 999-1002. T. Okiyama er al. Proc. IOOC’89, (July 18-21, 1989, Kobe, Japan),

pp. 207-210.

pp. 59-62

1321-1323, 1994.

pp. 995-997, 1994.

59

paper 2OD4-5. K. Kawano et al., Electron. Lett., vol. 25, no. 20, pp. 1382-1383, 1989. T. Fujiwara et al., IEEE Photon. Technol. Lett., vol. 2, no. 4, pp. 260- 261,1990. G.K. Gopalakrishnan et a!., Proc. IPR’92, (Apr. 13-16, 1992, New Orleans, USA), paper TuG3. D.W. Dolfi and T.R. Ranganath, Electron. Lett., vol. 28, no. 13, pp.

K. Noguchi et al., J. Lightw. Technol., vol. 13, no. 6, pp. 1164-1168, 1989. D.G. Girton et al., Appl. Phys. Lett., vol. 58, no. 16, pp. 1730-1732, 1991. C.C.Teng et al., Proc. OFC’42 (Febr. 2-7, 1992, San Jose, USA), post- deadline paper PD5. M. Seino etal., Proc. OFC’92, (Febr. 2-7, 1992, San Jose, USA), post- deadline paper PD3. H. Nagata and J. Ichikawa, Opt. engineering, vol. 34, no. 11, pp. 3284- 3293,1995. G. Ishikawa et al., Proc. 2Znd ECOC’96, (Sept. 15-19, 1996, Oslo, Norway), paper ThC.3.3. J. Sochtig et al., Electron. Lett., vol. 32, no. 10, pp. 899-900, 1996. H. Takeuchi et al., IEEE Photon. Technol. Lett., vol. 1, no. 8, pp. 227- 229,1989. J.E. Zucker et al., IEEE Photon. Technol. Lett., vol. 2, no. 1, pp. 32-34, 1990. J.E. Zucker et al., Appl. Phys. Lett., vol. 60, no. 3, pp. 277-279, 1992. J.E. Zucker et al., Electron. Lett., vol. 28, no. 20, pp. 1888-1889, 1992. T. Tanbun-Ek et al., IEEE Photon. Technol. Lett., vol. 7, no. 9, pp.

H. San0 et aZ., Proc. OFC’9.2 (Febr. 2-7, 1992, San Jose, USA), paper ThG4. H. San0 et al., Proc. OFC/IOOC’93 (Febr. 21-26, 1993, San Jose, USA), paper ThK5. C. Rolland et al., Proc. OFC/IOOC’93 (Febr. 21-26, 1993, San Jose, USA), post-deadline paper PD27. D.M. Adams et al., Electron. Lett., vol. 32, no. 5, pp. 485-486, 1996. Ph. Delansay et al., Proc. OFC’97 (Feb. 18-20, 1997, Dallas, USA), paper TuDl . P. Jiang and A.C. O’Donnell, IEEE Photon. Technol. Lett., vol. 8 , no.

F. Devaux et al., Appl. Phys. Lett., vol. 61, no. 23, pp. 2773-2775, 1992. M. Allovon et al., IEEE Photon. Technol. Lett., vol. 7, no. 2, pp. 185- 187,1995. D. Lesterlin et al., Proc. OFC/IOOC’93 (Febr. 21-26, 1993, San Jose, USA), paper ThK4. T.H. Wood et al., Electron. Lett., vol. 21, no. 16, pp. 693-694, 1985. U. Korenetal., Appl. Phys. Lett., vol. 51, no. 15,pp. 1132-1134,1987. J.E. Johnson et al., Proc. I4th IEEE Semic. Laser. Cmf., (1994), paper M4.7. Y.K. Park et al., IEEE Photon. Technol. Lett., vol. 8, no. 9, pp. 1255- 1257, 1996. M. Suzuki et al., Electron. Lett., vol. 22, no. 6, pp. 312-313, 1986. M. Suzuki et al., J. Lightw. Technol., vol. 5, no. 9, pp. 1277-1284, 1987. H. Tanaka et al., J. Lightw. Technol., vol. 8, no. 9, pp. 1357-1362, 1990. H. Soda et al., Electron. Lett., vol. 23, no. 23, pp. 1232-1234, 1987. H. Soda et al., Proc. OFC’89 (Feb. 6-9, 1989, Houston, USA), post- deadline paper PDl . H. Soda et al., Proc. OFC’90 (Jan. 22-26,1990. San Francisco, USA), paper THI2. H. Morito et al., Electron. Lett., vol. 31, no. 12, pp. 975,-976, 1995. K. Morito et al., IEEE Photon. Technul. Lett.., vol. 8, no. 3, pp. 431- 433,1996. Y. Kawamura et al., IEEE J. Quantum Electron., vol. 23, no. 6, pp.

K. Wakita etul., Electron. Lett., vol. 24, no. 21, pp. 1324-1325, 1988. K. Wakita et al., J. Lightw. Technol., vol. 8, no. 7, pp. 1027-1032, 1990.

1197-1198, 1992.

1019-1021, 1995.

10, pp. 1319-1321, 1996.

915-918, 1987.

19 1997 Electronic Components and Technology Conference

0. Mitomi et ul., Appl. Opt., vol. 31, no. 12, pp. 2030-2035, 1992. T. Kataoka et al., Electrt,n. Lett., vol. 30, no. 11, pp. 872-873, 1994. H. Takeuchi et al., Proc. 22nd ECOC’96, (Sept. 15-19, 1996, Oslo, Norway), paper ThD3.1. K. Yoshino et ul., Proc. 22nd ECOC’96, (Sept. 15-19, 1996, Oslo, Norway), paper ThD3.5. H. San0 et ul., Proc. OFC’90 (Jan. 22-26, 1990, San Francisco, USA), paper WM15. M. Aoki etal., Electron. Lett., vol. 27, no. 23, pp. 2138-2140, 1991. M . Aoki etul., Electron. Lett., vol. 28, no. 12, pp. 1157-1158, 1992. M. Aoki et al., Electron. Lett., vol. 29, no. 22, pp. 1983-1984, 1993. T. Ido etul., Electron. Lett., vol. 31, no. 24, pp. 2124-2125, 1995. G. Mak, et ul,, IEEE Photon. Technol. Lett., vol. 2, no. 10, pp. 730- 733, 1990. J. Langanay et al., Appl. Phys. Lett., vol. 62, no. 17, pp. 2066-2068, 1993. H . Haisch et al., Proc. ECOC’94 (Sept. 25-29, 1994, Florence, Italy), pp. 801-804. K. Satze et ul., Electron. Lett., vol. 31, no. 23, pp. 2030-2032, 1995. B. Clesca et u L , Proc. ECOC’95, (Sept. 17-21, 1995, Brussels, Bel- gium) post-deadline paper ThA.3.8. D. Lesterlin et ul., Proc. 22nd ECOC’96, (Sept. 15-19, 1996, Oslo, Norway), paper WeP.20. P. Ojala et ul., Electron. Lett., vol. 29, no. 10, pp. 859-860, 1993. 0. SahlBn et ul., Electron. Lett., vol. 32, no. 2, pp. 120-122, 1996. P.I. Kuindersma et al., Electron. Lett., vol. 29, no. 21, pp. 1876-1878, 1993. P.I. Kuindersma et ul., Proc. CLEO/Europe, (Aug. 28 - Sept. 2, 1994, Amsterdam, Netherlands), paper CThH3. J.J.M. Binsma et ul., Opt. & Quuntum Electron., vol. 28, pp. 455-462, 1996.

[121] K. Komatsu et al., Proc. OFC’94, (Febr. 20-25, 1994, San Jose, USA), paper TuC3.

[I221 M. Ushirozawa et al., Proc. 22nd ECOC’96, (Sept. IS-19, 1996, Oslo, Norway), paper WeD.3.3.

[123] E Devaux etul., Electron. Lett., vol. 27, no. 21, pp. 1926-1927, 1991. [124] E Devaux et al., IEEE Photon. Technol. Lett., vol. 4, no. 7, pp. 720-

722, 1992. [125] E Devaux etul., Electron. Lett., vol. 30, no. 16, pp. 1347-1348, 1994. [126] A. Ramdane et al., IEEE Photon. Technol. Lett., vol. 7, no. 9, pp.

[127] A. Ramdane et ul., Proc. 22nd ECOC’96, (Sept. 15-19, 1996, Oslo, Norway), paper WeD.3.2.

[128] S. Oshiba etab, Electron. Lett., vol. 29, no. 17, pp. 1528-1530, 1993. [129] K. Yamada et al., Proc. OFC’95, (Febr. 26 - Mar. 3, 1995, San Diego,

USA), paper TuF4. [130] R.E. Behringer et al., Proc. OFC’97 (Feb. 18-20, 1997, Dallas, USA),

paper WGI. [131] U. Koren etal., Proc. OFC’92 (Febr. 2-7, 1992, San Jose, USA), paper

WG5. I1321 F. Zamkotsian et al., Electron. Lett., vol. 31, no. 7, pp. 578-579, 1995. [133] M. Gibbon et al., Proc. ECOC’93, (Sept. 12-16, 1993, Montreux,

Switzerland), paper, pp. WeC7.5. [134] K Sat0 etal., Electron. Lett., vol. 30, no. 14, pp. 114-1145, 1994. [I351 N. Soulli etal., Proc. OFC’97 (Feb. 18-20, 1997, Dallas, USA), paper

WG6. [136] B. Velshow eial., Proc. OFC’95, (Febr. 26 - Mar. 3, 1995, San Diego,

USA), paper wL5. [I371 D.A. Ackerman etul., Proc. OFC’97(Feb. 18-20, 1997, Dallas, USA),

paper WG7. [I381 H. Tanakaetal., Electron. Lett., vol. 26, no. 25, pp. 2124-2125, 1990. [139] B Clesca etal., Electron. Lett., vol. 32, no. 10, pp. 927-929, 1996.

1016-1018, 1995.

600 1997 Electronic Components and Technology Conference