Emerald Group Publishing Limited
Copyright © 2004, Emerald Group Publishing Limited
3 and 4 March 2004
Keywords: iMAPS, Conferences, Microelectronics
iMAPS UK moved to Cambridge for their annual conference this year, and held it at the modern Möller Centre which overlooks the playing fields of Churchill College (Plate 1). This was both a well-attended and supported affair, run in the customary iMAPS style, a mixture of professionally delivered technical papers, an eclectic mix of exhibitors, and all held in perfect facilities (Plates 2-4).
Plate 1 The Möller Centre
Plate 2 Delegates in the conference hall
Plate 3 Peter Ongley chats with delegates at MICROTECH
Plate 4 Exhibitors at MICROTECH
Steve Muckett (Plate 5), Chairman of iMAPS UK, welcomed the delegates, and got matters under way by inviting each of the 14 exhibiting companies to introduce themselves to the conference, so that everyone could know who they were, what they did and where to find them afterwards.
Plate 5 Steve Muckett and delegates
Day One was devoted to Photonics and Communications
The first speaker was Professor Stephen Ferguson who works with Marconi Communications, the biggest in the UK for fibre optics. He looked at the top ten photonic industries, and the long-term trends. The good news is that traffic is growing, at 30 per cent per annum, and up to 100 per cent in some places, but this is traffic, not revenue; most traffic is on the Internet and not on landline. However, in the next 5-6 years, telecomm will look very different; until 2005 it will be business as usual, with most technologies visible, but after that it will all begin to happen, 40 Gb/s will be deployed. He sees photonics and electronics coming together to offer a huge reduction in the cost of processing. In the meantime, we are building on what we have. He mentioned the fact that 300 R-OADM nodes are being shipped, into Europe. The networks have a lot more capacity spread around than you could measure. Each one carries about one-fifth of the total present traffic but serious money is currently being spent, with PONS being taken more seriously in the USA and Japan. Typically, the EU is stalling.
Digital photonics, or high speed electronics if you like, are multiwave regenerators, clever technology here, and there are photonic integrated devices suitable for manufacture for 40/160 Gb/s which incorporate optical packet switching. Photonic crystals could be amazing, they are smaller, faster and better, and will help to make networks faster and bigger by allowing components to be smaller and cheaper. A 3D photonic crystal is presently being developed in Toronto.
Electronic and optical backplanes were discussed; costs are still high, and connector maintenance is an issue. Boards with embedded optical waveguides are not yet mature, but one day they will be, and he indicated the roadmap for success in this field. In conclusion he said that quantum optics are out there somewhere; he explained what Holey fibres are, how they will take over from optical fibres, and their huge potential in photonics.
Professor David Richardson, Optical Research Centre, Southampton University, came to talk about the rapid growth in fibre lasers. In 1985, it was 1 mW or so, at present we are getting 30 W, and over last 7 years work has progressed to the point where we are approaching the kilowatt power level. This is a truly enabling technology. The attractions of fibre lasers are numerous – they are scalable to high power operation; they have high optical efficiency, typically 85 per cent; diffraction limited beam quality; easy heat removal; compact and robust cavity designs; leverage from telecom grade components. Applications are in industry – welding, printing, defence; in remote sensing; in aerospace – applications in range finding; in medicine – imaging, laser surgery; and in telecoms – Raman lasery, high-power multi-port EYDFA I-watt scale.
Professor Richardson discussed about a technique of cladding pumped fibre devices, where a second waveguide is created in the core, and the unfocusable light can be focused into this second core as a waveguide. This means that kilowatt scale devices can be handled, and larger core dimensions have the necessary fibres; they are at present looking at larger core scales, 50 μm and above, and maybe in the 100-500 μm range when power requires it. These are long thin devices, up to 100 m in length.
Key technologies for the future include the following.
Pump diodes. Yb-doped fibres for cladding pumping are attractive, they exhibit small quantum defect and high efficiency, and there followed an explanation about refractive index control. Pump sources – there are different choices here. There are single mode emitters (0.5 W), broad stripe emitters (5 W); bars (50- 100 W) and stacks (500-2 kW). Fibre integrated launch options were covered, and Er-Yb MOPA was described, at present getting powers of 100 W. There is also an all-glass power doped fibre laser (DFL) – no polymer in this system, no concerns about power handling, it can still generate multi-hundreds of watts. They can also get pulsed output, through a Q-switched YDFL giving 10 mJ through parabolic pulse amplification.
In future, he concluded, they would be looking at other fibres, up to the 10 kW regime. The key for the future lies in the ability to make new fibres.
Tim Simmons of Bookham Technology presented a paper entitled "cost-effective monolithic/hybrid fibre optic packaging solutions". Now that the market is turning around, he looked at the effects of the rise of multi-source agreements (MSA) on integration, and made a comparison of monolithic and hybrid integration. There will be a 3-7 per cent growth in fibre optics market this year, with some interesting changes, and many of these changes have already impacted on how the industry is viewed. Cost is now a king, any route to get cost out of the system must be taken; engineering teams are much smaller than they were, but the demands made upon them are bigger, the final product has to pack more ability into less space.
Monolithic means integrating multiple functions on one substrate.
Hybrid means integrating multiple functions within one package.
MSAs are agreements between a number of suppliers to supply parts that are identical in purpose and conform to the Fit Fair and Functional nomenclature. This started in datacom but has now moved over to telecom, and the major differentiation between suppliers now is price. So low cost construction is important, and integration often provides the edge when seeking this cost reduction.
Monolithic integration – the aim is to integrate multiple functions onto the same substrate, and optics will play a part in this in days to come. He gave a GaAs modulator as one example, also a leaded-line travelling wave modulator. A 10 G NRZ GaAs Integrated Modulator was described in functionality,
Hybrid – 10 Gb/s tuneable integrated transmitter – a tuneable laser offering a quarter of C band in one go. It has a GaAs modulator and a lot of optics in it, no less than three lenses; it is very complex and is pure hybrid engineering. It is called the TM10N and Tim painted a very clear picture of how it was designed, how it worked, and its performance.
Dr Steve Methley of Plextek discussed about the Ethernet first mile (EFM), and optical Ethernet. There are three design challenges for EFM. It will drive optoelectronics integration towards both high performance and low cost at the same time. What is EFM? When it first came out at 1 Mb/s, it was a market failure, but now it is in 10 Mb/s coax it has taken off. Twisted pair wiring meant that it became popular in speed, in reach, and gave a decrease in cost per megabit. You get ten times the speed for three times the cost. It was also ubiquitous. This year the EFM standard, IEEE 802, will bring Ethernet everywhere. The objectives for EFM were described, including an optional specification for combined operation on multiple copper pairs. The Ethernet passive optical network (EPON) was illustrated, and Dr Methley looked at the design challenges – the extended temperature range, the reach, and single fibre working for 100 Mb/s and 1 Gb/s.
It is simple, it is robust, economies of scale are there, said Steve. Ethernet is the cheap way to install WAN – copper wire means digging holes!
Adaptive Optical Interconnects was the subject of a paper from Dr Tim Wilkinson from the Department of Engineering at Cambridge University, who described how you put together an optical system. Backplanes are a bottleneck to information switching between boards. He explained the differences between fixed interconnect versus optical interconnect, the use of vertical cavity surface emitting lasers (VCSELs) and CMOS compatible photodetectors. However, they are prone to alignment errors, caused by vibration amongst other things.
He described the way in which the optical interconnect works between two boards. The interconnect is adapted using a hologram or grating, where the hologram is displayed on an FLC microdisplay. But it does have a binary limitation; however, we can compensate for shift, tilt, and focus on both boards, and can compensate for optical aberrations. Holograms offer adaptive alignment.
Systems in a package case study by Piers Tremlett of Zarlink. Zarlink normally sell ICs, but their customers prefer module solutions for lower engineering design input; effectively a system in a box. The module had to be smaller than competitor modules; it had to be surface mountable on a PCB and should preferably look like a semiconductor component!
Inside the module they use a bare die and wire bond; needless to say, with every new design comes new problems. The PCB (not ceramic) design was chosen because of customer acceptance; it had to look like a BGA, but this involves finer geometries than thick film. Purchasing the PCB was difficult, the fabricator had to learn what was needed, and consideration had to be given to aspects such as layer build and materials, wire bondable finishes, micro vias, so finding a suitable supplier had been tricky, one company specified the wrong prepregs for a start, which did not help. Some of the production problems included corner BGA ball delaminating off the PCB after reflow, also delamination due to lid stress relief during reflow. They currently use flexible glue to attach the lid.
Piers concluded that the target specifications had been met, and the first unit had passed the Telcordia GR-253 specification, and their customers thought that it met what they wanted. It had a similar footprint to a true SMT component and had high performance. The potential here is for a product that does what it should and makes a profit.
Paul Misselbrook came down from Celestica at Kidsgrove to address the conference that Opto- Electrical Circuit Boards (OECBs) can replace copper circuit boards where vertical cavity surface emitting lasers (VCSELs) are used. Computational modelling work had been done at Celestica on a novel method of producing integrated OECB mirrors, the key to it all. As a background, Paul said that optical networks form the basis of the modern telecommunications infrastructure, all around the world, in long haul and long transatlantic link lengths. Long haul networks used the DWDM systems, and DFB (distributed feedback) laser diodes that were good for long haul. For Metro networks DFB lasers were uneconomical, whereas, for Access networks VCSELs were good.
So, DFB laser diodes are being developed to allow, as are SELs also being developed for high power, single mode and long wavelength, replacing conventional circuit boards with very short reach (VSR) discrete optical fibre interconnects using third generation embedded waveguides. VCSELs are available as flip-chips and are low cost due to in wafer processing. The coupling between the waveguide and the VCSEL is done by assembled mirrors, or integrated mirrors, and project work on this is at present nearing completion. Greenwich University are looking at some of the manufacturing questions. Paul rounded up his paper by adding that models are nearly complete, and they hope to be able to manufacture an OECB board by the end of the year, to achieve optical interconnect availability by 2005.
Nick Parsons did not have far to travel; he comes from Polartis Limited of Cambridge. Advances in optical switching was the title of his talk, which covered application trends, optical switch requirements, current technologies, direct beam steering technology and performance tests.
Mesh networks give greatest routing flexibility, he said, easing traffic planning. OEO conversion currently represents 50 per cent of equipment costs. An optical switching layer can significantly reduce equipment costs, as 70-80 per cent of traffic is just passing through. But low optical loss is essential if the system is to be viable.
The ideal optical switch has low optical loss, has no interference between paths, can handle high power, low noise is induced on signals, and it has scaleable technology. But it is not simple to set-up. At Polartis they have a product in which low optical loss is now assured, it has minimal PDL and outstanding repeatability, and the instrumentation world is very interested in this switch, which will significantly reduce the cost of test automation. It has clean switching characteristics, with no overshoot, and control- induced amplitude of <0.02 per cent. It is also insensitive to vibration – but how to package it? They had studied low cost reliable sealing of large optical assemblies – low loss multifibre feedthrough and joining technologies; and window sealing of sub-assemblies to enable pay as you grow architectures.
Metro and storage area network trends are moving towards unamplified mesh architectures and low loss switching fabrics are a critical enabler in this regard.
Day Two was given over to sensors, actuators and optical MEMS
Dr J. Malcolm Wilkinson is the Managing Director of TFI Ltd and he steered us through the roadmaps in sensors, actuators and optical MEMS. There is a nice spread of application areas for sensors. But why should we integrate sensors and actuators? Well, the cost advantage for one, space saving for another, an improvement in reliability and a better performance were attractive benefits. BUT against that you have to bear in mind that a hybrid solution is faster to market. The product roadmap for optical MEMS in communications made for an interesting topic, as bandwidth costs increase, so the pressure on wavelength switching increases, with MEMS with integrated optics proving invaluable. He came to the subject of plastic electronics – why bother? Here the answer is that plastic is good at large areas, has low cost, is flexible, is not high performance and can be used in photonics. Might be a good subject for the next iMAPS conference, he suggested.
Two gentlemen from BAE Systems, Roger Hill and Duggie Bremner came to address the conference about Automated Optical Assembly. They discussed calmly about a project that must have been quite daunting in conception and in execution, and that it took 2 years to create indicates the complexity of creation. The biggest problem is getting the alignment, a variation of 1 μm throws coupling out by 15 per cent, for instance. When it comes to fixing things, the word welding comes into the conversation quite a lot, and post-weld shift due has to be allowed for so that the random displacement of the lenses due to post-weld error is allowed for. Describing the system, Duggie said that the 40Gb project modulator unit is pretty conventional, but the lens, which needs to be aspheric with a waveform error of 0.2 lander, was made for them in Holland. This is mounted in a stainless steel chassis with a cobar clip to give stability. The entire unit is now operational and is for the production of high-yield optoelectronic packages for military end use.
Mark McNie is the Senior Scientist for Microsystems at QinetiQ, Malvern. In his paper he commented on the process used at QinetiQ to produce hollow waveguides in silicon substrates using MEMS process technology including photolithography and deep dry etching. The process is still at an early stage of development, but hollow waveguides are held to be an enabling structure for integrated optical circuit technology for the low cost automated manufacture of optical modules both discrete and monolithic functions. But it all looks very promising, and trials with monolithic integration are still going on.
The University of Durham has a Microsystems Technology Group, where Tennyson Nguty has been working on the packaging of fibre-optic pressure sensors. The sensor itself is based on the microbending in an optical fibre that causes attenuation of the guided light. In a project which began 7 months ago, Tennyson described how microbending in the fibre is induced by a set of corrugated surfaces, or teeth. These teeth structures are formed by a standard wet etch process, on silicon with a v-groove structure added to the bottom to ensure teeth alignment. In a detailed paper, Tennyson was able to demonstrate that the objectives – the packaging should be very small, noise free, and inexpensive – had been met, whilst temperature sensitivity was much reduced. The applications – in the automotive and medical fields – were considerable.
Someone else from the NorthEast is Andrew Gallant from the School of Engineering at the University of Durham; he discussed about widely tuneable capacitors and the challenges of integration. One of the benefits of being in the Northeast was that one of the leading gallium arsenide plants was just down the road. Andrew discussed and described the build-up technology employed to produce RF MEMS switches and, through a complex but logical micromachining technique, described how MEMS are formed. Tuneable capacitors with a tuneable ratio of 7.3:1 have been produced in a clean room facility at Durham.
Dr Eric Nowak works for VEECO, and has done so for some time. The three-dimensional characterisation of MEMS devices is an affair close to his heart, and coming all the way to Cambridge from Tucson in Arizona he looked less pale than some of the delegates in the room. The MEMS industry has $9 million as revenues, diverse types in manufacture, micromirrors, and gas control systems, etc. Metrology is valuable only with complete accurate and timely feedback into the process. Reliability is the key factor, so where do they fail – contamination, electrostatic clamping; environmental attack, fatigue to name but a few. Eric ran through the white-light optical interferometry for static MEMS, which measures MEMS surfaces both statically and dynamically. This is a non-contact technique which operates through the use of an interferometric microscope which employs a specialised objective that combines light reflected from a high-quality reference surface with that from the test object. All the dynamic properties are covered in this exercise, and MEMS can be fully assessed accurately and rapidly.
Noel Cherowbrier hails from the Channel Islands, but is the familiar face and Sales Director of TECAN in Weymouth and knows a lot about nickel microstructures. There are no products from TECAN, only processes – photo electroforming, metal finishing and photochemical milling. Ultra precision makes micro and macro metal parts, but also micro and macro tooling and replication, and Tecan are always looking at how the techniques used in one industry can be used in another. PEF and micro-PEF are the onward paths; the latter used for discrete micro-components, and MEMS. With MOEMS and FLEMS they call for small. Small is what Tecan does, with micro tools which can yield features down to 2 μm in size, and Noel described their HDI imprinting technique. They also offer wafer replication and hybrids, multi-level deposition masks. Challenges which lie ahead include In Chip Cooling – Fuel Cell Technology – Mini Reactors – and mini heat exchangers.
Dirk Enderlain came over from Germany where he works for HL – Planartechnik GmbH. He discussed MEMS Foundry Services vs In-House Production. The initial investors in MEMS lost a lot of money, but the end of last year and the beginning of this year has seen the market recover, for 2004 it will, he says, be a good year. The theme of his presentation was based on – if you have a new product idea, how do you come to the market? In this consideration you have to decide if you can do everything yourself, do you have the key knowledge. Do you have the experience? How can you protect your intellectual property? What are the costs? Is external production flexible enough? Do you have the right skills to manage this process? What are your customer's expectations?
IC Foundries are well known, have a high degree of standardisation, and a second source is always available. With a MEMS Foundry, before you invest in your own production there are a number of things you have to look at – the provision of a clean room facility, raising the capital for the equipment which alone can cost around 20 million; you have to set-up the process, the logistics, the quality system, SPC, this take a long time. On the other hand, there are some advantages you have control over manufacturing; cost transparency; higher added value; better IP protection, and the flexibility to modify processes.
On the other hand, the Foundry production industry is a big community already producing MEMS. In Europe they are all members of the Europractice Consortium and they can manage everything from the design of your concept down to full production. Bosch is a typical example of this service. They have had experience of 80 projects, over 200 designs.
Amongst some interesting case studies was a gyroscope used in navigation systems that gave camera stabilisation, and in robotics. A mass air flow meter module that carries a thin membrane, and provides a flow response below 10 ms, is now used in high volume production; and a fuel cell for a car, using hydrogen. Now that was interesting!
Hayden Taylor from University of Cambridge discussed to us enthusiastically about his work on MEMS and actively manipulating optic components, this included making use of surface tension in beads of solder when they melt to lift lenses out. There was a strong emphasis on simplification, and he described the work done on how to machine laser well. He had been studying the laser ablation of thin film, using UV laser for ablation of silicon nitride. Extracting Young's modulus of thin films was also studied. We know how to machine thin films, but what are we going to achieve? Three concepts they are trying out are deep reactive ion etching, out-of- plane bistable clamps and thin film microclips.
It came as no surprise to hear that Hayden will be going over to the States to study at MIT for his doctorate once he has taken his BSc this year. We wish him well.
One of the unique aspects of an iMAPS Conference is the opportunity given to the many suppliers to the microelectronics industry to meet the people who matter. The exhibitors also matter greatly, and the following distinguished companies were present at Cambridge.
Cookson Electronics/CPSM – email@example.com
CoorsTek – www.coorstek.com
Chip Supply Inc. – firstname.lastname@example.org
Dyconex – www.dyconex.com
Hybrid Laser Tech – www.hit.co.uk
Inseto (UK) Ltd – www.inseto.co.uk
Lambda Photometrics Ltd – email@example.com
Laser S.O.S. Ltd – firstname.lastname@example.org
Microponents Ltd – www.microponents.com
Microstencil Limited – www.micro-stencil.com
Namics Corporation – email@example.com
NTK Technical Ceramics – www.nktech.com
Olin (UK) Ltd – www.olinaegis.com
PandA Europe – www.pandaeurope.com
Tecan Ltd – www.tecan.co.uk
Williams Advanced Materials – www.william-adv.com