Photonic Integrated Circuits

Get up to speed on developments that could revolutionize telecom equipment * Huge promise * Huge challenges * Who's winning?

February 7, 2002

45 Min Read
Photonic Integrated Circuits

It's easy to see the benefits of employing optical technologies in telecom networks, but right now there are plenty of obstacles in the way of turning dreams into reality.

One of the big obstacles is cost. Carrier budgets are tightening, and a lot of optical gear is still very expensive. Analysts agree that operators cannot continue to absorb reductions in bandwidth prices unless system vendors and components manufacturers reduce costs drastically.

Another is space. The deployment of DWDM in networks has led to a huge increase in the number of boxes in carrier sites – so much so that the footprint of equipment has become a crucial issue.

Another obstacle is time to market. New developments are happening so fast that life cycles of equipment are shrinking dramatically. Vendors can't afford to dilly-dally, developing from the ground up. They need to buy off-the-shelf subsystems to get their products to market much more quickly.

Guess what? These and many other obstacles could be swept to one side by developments in integrated optics. A whole bunch of vendors, many of them startups, are working hard to put multiple optical functions on a single optical "chip."

An optical chip in this context means a wafer-based component that can be processed using the same automated manufacturing techniques that are used in the electronic chip industry.

Many vendors are targeting passive devices in the first instance – circuits that switch, attenuate, or filter light, but don't amplify it. They plan to incorporate active devices, like lasers and amplifiers, at a later date.

In parallel developments, vendors are putting the electronic control circuits on the same chip as the optical functions. After all, optics cannot exist without electronics: You need the electrons in order to create photons in the first place.

Bringing all three together – passives, actives, and electronics – will be the next obvious step.

Eventually – and this is pretty futuristic – folks hope to develop the optical equivalent of the electronic integrated circuit. In other words, a chip that does its thinking in optics. The bottom line here is processing speed. Optics has the theoretical potential to carry more information per second.

If they succeed (and that's a very big if) the impact on telecom equipment and networks could be just as massive as the impact of electronic chips on pretty much everything. The systems business could become a lot more streamlined and rapid. Telecom equipment that currently occupies whole racks could be shrunk into a single component. Telecom services could end up following Moore's law – doubling in performance and halving in price every 18 months.

Though industry experts believe that this kind of evolution is inevitable in the long run, most see plenty of problems along the way. "We're a long way from developing the optical equivalent of the transistor," says Naser Partovi, managing director of Enterprise Partners Venture Capital.

All the same, photonic integrated circuits promise to have such a profound impact on telecom networks that it's important to understand what's going on, whether you're an investor, equipment manufacturer, or service provider.

As usual, the best way of getting to grips with the issues is to dig into the details. Light Reading has taken the sweat out of the process in this report. Here's a hyperlinked summary:

Page 2: What Is Integration?

The basic aim is to eliminate fibers inside optical systems, but there are degrees of doing this – from using lenses to steer light in free space between devices to the use of waveguides in planar lightwave circuits (PLCs). This report focuses on PLCs.

Page 3: A Slow Start

Some of the obstacles holding back developments include not having the computer power to simulate material behaviour, lack of test equipment, inability to achieve high yields from manufacturing processes, and having to make compromises on the choice of materials to make optical chips.

Page 4: Speeding Up?

A year ago, arrayed waveguide gratings (AWGs) were seen as the obvious starting point for integration initiatives. Other platforms, such as tunable lasers and switches, have now also emerged.

Page 5: Market Forecast

The research group, CIR, reckons it'll be a $2.6 billion market by 2005, predicting the largest application of integrated optics will go into transmitters.

Page 6: Product Primer

How AWGs work, and how they stack up againt Echelle gratings, an alternative. Some examples of subsystems made with AWGs, including Mux-VOAs, optical channel monitors, dynamic gain equalizers, and optical add/drop multiplexers.

Page 7: Materials Primer

A comparision of the strengths and weaknesses of silica, silicon, indium phosphide, gallium arsenide, lithium niobate, silicon oxynitride, sol-gels, and polymers for making different types of device. A market forecast suggests indium phosphide will end up being the dominant material.

Page 8: Vendors: Public

What's cooking at key public companies, including Agere, Agilent, Alcatel, Bookham, Corning, Hitachi Cable, IBM, JDS Uniphase, Lumenon, and NEL.

Page 9: Vendors: Private

A Who's Who of more than three dozen startups working on integrated optics developments.

Next Page: Page 2: What Is Integration?

Let's get some definitions straightened out at the start.

The basic idea behind integration is to eliminate the fibers among individually packaged components in a traditional optics setup. Fibers are undesirable. Each fiber-to-component interface is a source of loss and a potential point of failure. Worse, attaching fibers to components is the hardest part of the manufacturing process to automate, so it's still done manually in a lot of factories. And to cap it all, packaging is expensive.

There is more than one way of achieving this aim, which can lead to some confusion. The best way to clear up this confusion is to consider the different degrees of integration.

  • Co-packaging

    Take any two components, slap them together, stick them in the same package with lenses in between, and you have integration – of a sort. At least, some people think of it that way; to others it's simply co-packaging. (Now that higher levels of integration are possible, components have to work harder to gain the label.)

    To be fair, this approach can be automated, but that's no simple task. It often requires a complete redesign of the optical subassembly so that it can be handled by robots. But the scope for increasing integration levels in this way appears limited, as adding additional components might involve designing yet another entirely new subassembly.

  • Optical benches

    Taking integration to the next level involves miniaturizing and standardizing the cradles and mounts that hold the elements. Making tiny cradles and minuscule lenses is now possible with the latest microfabrication techniques. But making sure that everything's environmentally stable is a challenge. The modules have to be designed to withstand the fact that materials they're made of have different thermal expansion coefficients.

    Axsun Technologies, Cube Optics AG (CUBO), and Digital Optics Corp. are examples of companies doing things this way. Some companies refer to this kind of integration platform as an "optical bench." It's also called "hybrid" integration, which simply means it includes several different materials.

    For more information see Axsun Gets $111 Million for Subsystems and Startup Gets Clever With Cubes.

There's another way to eliminate fibers: using optical waveguides. These guide light by total internal reflection, just as an optical fiber does, but are rigid structures, sitting on top of a semiconductor wafer.

Each company has coined its own particular term to describe the platform it uses. General terms include optical chips, photonic integrated circuits, and optoelectronic integrated circuits. In our view, the most unambiguous term for waveguide-based components is Planar Lightwave Circuit (PLC).

PLCs can be manufactured using techniques borrowed from the microelectronics industry – a big plus, since those processes are well established and highly automated.

There are two basic types of PLC:

  • HybridThis method uses a wafer as a subassembly. Some components, including the waveguides themselves, can be fabricated directly in the wafer. Others, like lasers, sit in holes etched out of the wafer. Standard semiconductor processing techniques are used to define the waveguides, as well as the pits, grooves, and markers needed to align other components with the waveguides.

  • Monolithic

    Monolithic integrated components are made from one material (or a small subset of compatible materials), so there are no mechanical joins or bonds between different sections of the chip. The component is manufactured from start to finish using a linear sequence of processes.

This report will focus on PLC-based integration, because this technology has the potential to follow the evolutionary path all the way to its conclusion – where we will have the optical analog of the electronic integrated circuit.

Next Page: Page 3: A Slow Start

A recent visitor to the industry might think that optical integration is a new field, since many of the prominent startups first registered on the radar just a few years ago. Dig deeper, and it turns out that many of them have been around for much longer.

Lightwave Microsystems Corp. started out about ten years ago as ROITech. When it couldn't meet its development targets, the company totally reinvented itself – this was mid-1998. Though Bookham Technology PLC (Nasdaq: BKHM; London: BHM) has carried the same marketing message since it was founded in 1988, it only started shipping products in 1995, following an investment by Intel Corp. (Nasdaq: INTC). And Gemfire Corp. spun out of Deacon Research, a contract R&D firm that had been in existence for 12 years. It then recouped all the expertise by doing a reverse merger with its parent.

In other words, very few companies are starting from scratch. The reason that they're only now beginning to make an impact is because there have been a lot of technological obstacles to overcome. And there are many that remain unsolved.

"I suspect everyone's going to go down some more blind alleys," says Drew Lanza, partner at VC firm Morgenthaler and a founder of Lightwave Microsystems. "Boy, we went down some of those blind alleys [at Lightwave Microsystems]. And many of the issues we faced have not changed or gone away."

The issues include:

  • Software simulation
    Algorithms for computing the behaviour of light inside optical components are a relatively recent thing, and the computer muscle that's needed to solve such highly-complex algorithms hasn't been around that long either. Bookham Technology made its mark by being first to show that it is possible to guide light in silicon. There was no intrinsic problem with the silicon material; it was just that people hadn't been able to do the calculations before, says Bookham's president, Andrew Rickman.

  • Test equipment
    Ten years ago, there was no dedicated test equipment for measuring optical components, so companies working in integrated optics had to build their own. That situation has improved. Not only is the kit available, but test and measurement vendors are also starting to automate their gear, so that it can keep up with production quantities (see Testing Gear Breeds Speed and Agilent Automates Component Testing). However, it may not be enough. Most test gear only checks the performance of a finished item, such as its insertion loss. It can tell which chips are bad, but it can't look into the chip at a deeper level to find out where faults originate.

  • Yields
    In an interview with Light Reading, ex-JDS Uniphase Inc. (Nasdaq: JDSU; Toronto: JDU) CEO Kevin Kalkhoven said "The thing about photonics is that you can make one of anything… The real issue is how do you make 10,000 of them a month?" This issue partly comes back to the need for test equipment that can pinpoint the cause of the problem. Yields are also affected by the fact that optical chips are significantly larger than electronic chips, which means: One, it's more difficult to ensure uniformity of the layers across the chip; and, two, every defect, however tiny, kills a whole chip. If that chip occupies a whole wafer, then the yield on that wafer is zero.

  • Materials
    Plenty of materials that are suitable for making optical chips and more are under development. Each has its pros and cons, so achieving high performance in one area often means compromise in another. This topic is covered in more detail on page 7.

Next Page: Page 4: Speeding Up?

A year ago, Light Reading blithely predicted that 2001 could be the year integrated optics finally takes off (see Top 10 Trends).

Well.

2001, of course, turned out to be the year that nothing took off (see The Lost Year ). Instead, several once-promising startups bombed out (see Nanovation Goes Bust and Zenastra Photonics: RIP). And vendors in general failed to deliver on their promise to produce more exciting optical circuits incorporating multiple functions.

That's not to say that things haven't advanced, however. Analysts – among them Lawrence Gasman from Communications Industry Researchers Inc. (CIR) – still believe that integrated optics holds great promise for the future of optical components. The benefits of putting multiple optical functions onto the same piece of semiconductor are still up for grabs. And the obstacles to attaining those benefits are being gradually worn down.

And the integrated optics scene has morphed into something new that is potentially quite interesting.

If Light Reading had published this article a year ago, the emphasis would have been rather different. Back then, the space was dominated by one product family – Arrayed Waveguide Gratings (AWGs). A couple of vendors, notably Bookham Technology PLC (Nasdaq: BKHM; London: BHM) and Lightwave Microsystems Corp., had got round to incorporating the most basic add-ons, such as variable optical attenuator or detector arrays. Even so, it would be no exaggeration to say that optical integration equalled passive demultiplexing technology. (There were a few exceptions: Agilent Technologies Inc. [NYSE: A] created a lot of fizz with the introduction of its bubble switch.)

These days, there is more variety. Vendors of laser-based components are very active. Agility Communications Inc. has developed a type of laser that allows monolithic integration with additional components, and other startups are following suit.

In addition, companies like Bookham, Lynx Photonic Networks, NTT Electronics Corp. (NEL), and Telephotonics Inc. are dabbling in switch-based products. With active, passive, and switching functions all under development, it looks as if optical integration is going to be a more broadminded technology than it has been.

To cap it all, 2001 ended with a pivotal acquisition, which brought together active, passive, and switching functionalities under one roof – Bookham's proposed purchase of Marconi Optical Components (MOC) (see Bookham Gets a Bargain). Bookham has already developed a range of passive components and has demonstrated switching functionality in the past year, while MOC brings high-end active capabilities in the form of what it claims is a nifty tunable laser technology.

Next Page: Page 5: Market Forecast

The big question is: When will the optical networking industry start seeing significant benefits from integrated optics?

It turns out that there are two answers, depending on how you segment the market, according to Communications Industry Researchers Inc. (CIR).In terms of AWGs, the benefits are already here. AWGs have stolen significant market share away from the incumbent technology of thin-film filters and are becoming more economic at lower channel counts, opening up more of the DWDM components market to them.

CIR estimates that shipments of AWGs totalled $27.5 million in 2001, according to analyst Lawrence Gasman. He forecasts that AWG shipments will top $192.7 million in 2005.

In terms of more advanced optical circuits, things have been pushed out, says Gasman. Some of this is due to the industry slowdown, and some is due to the fact that technological problems are taking longer to solve than expected. Gasman expects things to chug along slowly for the next year or so, start to expand rapidly in 2003, and reach a grand total of more than $2.6 billion in 2005 (see Integrated Optics: $2.6 Billion by 2005?).

That figure comes from a recent CIR forecast for the integrated optics market, which focuses on emerging technologies. AWGs aren't included, because Gasman considers them an established product. The idea behind the report, he says, was to look at the progress towards more exciting, highly-integrated technologies.

10593_2.gifIn the above chart, CIR breaks out the integrated optics market by application type, which is no mean task (we've also included AWGs). At the moment, there are a range of standard products, based on a limited number of integrated functions. As the number of functions on a chip goes up, the components will become increasingly customized, and they will no longer fall neatly into categories. Hence the "Other" category is quite large.

The "amplifier" category includes waveguide amplifiers from the likes of Teem Photonics, as well as Semiconductor Optical Amplifiers (SOAs) from companies such as Kamelian Ltd. "Monitors" means optical circuits that integrate taps or splitters with AWGs for demultiplexing, and a bunch of detectors. "DWDM passives" includes the VOA-Muxes from Lightwave Microsystems Corp., for example, which integrate an AWG with an array of variable optical attenuators. In "Switches," the type of switch fabric under consideration is waveguide-based and is built with thermo-optics (Lynx Photonic Networks, for example) or electrical carrier injection (Bookham Technology PLC's approach).

It's easy to see how it would be possible to devise a component that could span several groupings. For example, a loss-less mux, which incorporates an AWG with SOAs to amplify each channel, isn't a perfect fit with "DWDM passives," nor "Amplifiers." So, it gets booted into "Other."

Next Page: Page 6: Product Primer

Developments in photonic integrated circuits kicked off in earnest about five years ago with the invention of a new component, the arrayed waveguide grating (AWG). It's the first product for many startups in this field. For that reason, it's worth going into more detail about how an AWG works.

  • Arrayed Waveguide Gratings (AWGs)

    • An AWG – also called a phased array (phasar) demultiplexer or waveguide grating router – is a planar device that takes a bunch of wavelengths arriving at the same input, breaks them out, and sends them to different outputs. It can also be used in reverse to aggregate wavelengths.

      In multiplexing and demultiplexing applications, the AWG replaces 40 or more discrete thin-film filters (TFFs) with a single planar component. As a result, it's smaller and cheaper to make. For high channel counts, AWGs cut the costs by nearly half, from $750 down to $400 per channel, according to Vivek Tandon, a principal of Viventures Partners and a former VP of marketing and business development at Kymata, an AWG startup acquired by Alcatel Optronics (Nasdaq: ALAO; Paris: CGO.PA).

      Other benefits of AWGs include the fact that the insertion loss is uniform across the channels. On the downside, the performance, in terms of crosstalk, is worse.

      But the economics of deploying AWGs is constantly changing. "We thought we could just make an AWG and people would love them," Tandon says, "But the development took longer than we expected, and in that time, thin-film filters became very cost effective at 8 and 16 channels. That forced AWGs to 40 channels." He figures that AWGs will become more cost effective than TFFs at 16 channels in the coming year.

      Tandon calculates the takeup of AWGs (combined annual growth rate) at about 50 to 60 per cent per year.

      An AWG operates as a diffraction grating that splits light into its constituent colors. The diffraction grating is formed by an array of waveguides that have incrementally increasing lengths. That's why the structure is a sweeping curve.

      inputwavesmall.gifAt either end of the AWG is a free-propagation region, called a "star coupler" in the above figure, from Alcatel Optronics. On the input side, this gives the beam space to spread so that light falls on all the waveguides in the array. When the device is used in reverse (as a multiplexer), it focuses light from the waveguides back onto a single spot.

      On the output side, the star coupler focuses each wavelength of light onto a different output. It does this because of interference effects among the beams emerging from the multiple waveguides in the array. Changing the properties of the waveguides in the array – their length or temperature, for example – changes the way the beams interfere, and what's seen at the output.

      A simplified equivalent arrangement of lenses and prisms is shown in the following figure, also from Alcatel Optronics. In this case, the size is likely to be much larger than that of a typical AWG, which measures about 5 cm by 3.5 cm.

      outputwavesmall.gifAWGs come in two flavors: Gaussian and flat top. The names describe the passband characteristics of the device. A gaussian curve has an optimum operating point. Either side of this point, the transmitted optical power falls off sharply, so the device isn't very tolerant of changes in wavelength. A flat-top AWG is designed to have more tolerance, but this is usually achieved at the expense of insertion loss.

      Gaussian AWGs are typically used for multiplexing. Flat-top devices are best for demultiplexing, because the signal level is much weaker on the far end of a system.

    Just recently, an alternative to AWGs has emerged. In one sense, it's new, because the technology never made it to commercial application before. But, in fact, it's based on a technology that preceded the AWG. It's called an echelle grating.

  • Echelle Gratings

    • Two companies are known to be working on echelle gratings. Optenia Corp. – a spinoff from Mitel Corp. (NYSE/Toronto/London: MLT) – and MetroPhotonics Inc. Both companies take their technology from the National Research Council of Canada, but they are pursuing different materials. Optenia is working on silica components, while MetroPhotonics is targeting indium phosphide.

      An echelle grating operates in pretty much the same way as an AWG. But instead of using a waveguide array to create interference, the Echelle grating uses a stepped mirror. The steps in the mirror are responsible for the interference effects that break out the light into its constituent wavelengths.

      In some ways, the echelle grating looks like an AWG that's been folded back on itself. There is a free-propagation region, which allows the beam to spread out sideways before hitting the grating – in this case the mirror. Light is then reflected back towards the input waveguide, actually coming to focus on a series of output waveguides located adjacent to the input.

      Echelle grating technology suffers from a couple of serious difficulties. First, making the mirror requires a deep etch that has to be absolutely vertical and smooth. The second problem is the fact that the device is polarization sensitive. (The AWG was invented in 1996 by Meint Smit at the Technical University of Denmark to overcome these problems.)

      Recent advances in fabrication have brought echelle gratings back into contention, says Emil Koteles, CTO of MetroPhotonics. Although etched grating demultiplexers have not reached the same level of maturity as AWGs, they will catch up, he feels, and could prove to be a superior technology.

      The main difference between the two devices is the size. Echelle gratings are about four or five times smaller than an equivalent AWG. This affects not only the number of chips per wafer, but also helps boost yields. Non-uniformities across the chip become more critical the larger it is.

      Other potential echelle advantages are:

      • Easier to scale to small channel spacings
        When channel spacing decreases, more resolution is needed in the grating element. It's a lot easier to put more facets in a stepped mirror than it is to add extra waveguides to an array.

      • Lower crosstalk
        Crosstalk is generated by fabrication errors. Machines that make photolithographic masks write the patterns as a series of dots, and the smallest step between dots depends on the overall size of the area to be exposed. Since echelle gratings are smaller than AWGs, the resolution of the photolithographic mask can, in principle, be superior.

      • Flatter passband
        In AWGs a flatter passband can be achieved at the expense of a slightly higher insertion loss. Flat passbands can be created in echelle gratings without this tradeoff, says Koteles.

    As noted, the AWG is just a starting point, albeit a popular one. It's possible to build modules with more complicated functionality around the basic function of an AWG. For example:

    • MUX-VOA
      An AWG separates the channels, which can then be attenuated individually. Variable optical attenuation can be implemented in several ways, including thermo-optically and electronically.

    • Optical channel monitor (OCM)
      A tap splits off a fraction of the optical input from the main fiber, and sends it to an AWG. After the AWG, an array of detectors measures the optical power of each channel.

    • Dynamic gain equalizer
      This incorporates both the MUX-VOA and the OCM. Information from the OCM is used to control the degree of attenuation applied to the VOAs. After the array of VOAs, another AWG – this time in a multiplexing configuration – is used to aggregate the channels, which now all have equal power.

    • Optical Add/drop multiplexer (OADM)
      One way to make an ADM is shown at http://www.bbv-software.nl/support/faqnew/00001164.html. An AWG breaks out the channels, which then pass through a series of switches, where a channel can be dropped or another one added. The pass-through channels are aggregated by a second AWG, while the dropped channels are combined by a third. This isn't the only way of making an OADM: Semiconductor Optical Amplifiers (SOAs) can also be used as basic on/off switching elements. A third design, using an optical circulator (a non-reciprocal element), can avoid waveguide crossings.

    Next Page: Page 7: Materials Primer

    A major challenge facing developers of integrated optical subsystems is that different devices are best made using different materials. As a result, folk can be faced with some awkward compromises if they want to make a subsystem out of a single material – the key to making automation easier.

    Some of the key properties of materials used to make optical subsystems are listed in the table below, which comes from Louay Eldada, CTO and cofounder of Telephotonics Inc.

    Table 1: Properties of Materials Used in Integrated Optics

    Material

    Emits light?

    Losses in dB/cm, dB/chip

    Refractive index

    TO coeff. (dn/dT) * 100k

    Modu-lation

    silica

    no

    0.1, 0.5

    1.44-1.47

    1

    slow (TO)

    silicon

    no

    0.1, 1.0

    3.46

    18

    can be fast (EO)

    indium phosphide

    yes

    3, 10

    3.1

    8

    fast (EO)

    gallium arsenide

    yes

    0.5, 2.0

    3.37

    25

    fast (EO)

    lithium niobate

    no

    0.5, 2.0

    2.21, 2.14 (along x, z)

    1

    fast (EO)

    silicon oxynitride

    no

    0.1,1.0

    1.44-1.99

    1

    slow (TO)

    sol-gels

    no

    0.1, 0.5

    1.2-1.47

    1

    slow (TO)

    polymers

    no

    0.1, 0.5

    1.3-1.7

    -10 to -40

    fast (EO)



    Here's a column-by-column explanation of what it means:

    Emits light

    This shows that only two materials, indium phosphide and gallium arsenide, emit light when an electric current is applied to them. In other words, light sources such as lasers have to be made out of these materials.

    That's why folk got excited last year when some U.K. researchers managed to get silicon to emit light (see Light From Silicon).

    Losses in decibels per centimeter and decibels per chip

    These figures are the total propagation loss, which includes absorption, scattering of light caused by the roughness of waveguide surfaces, and radiation loss as light goes around bends, together with other types of loss resulting from differences in refractive indices and mode mismatches, according to Telephotonics' Eldada.

    The key point here is that indium phosphide's losses are high, compared to other materials. The upshot: It's tough to make a complicated subsystem completely out of indium phosphide without having the losses get unacceptably large.

    Refractive index

    The refractive index is important from two points of view. First, if it's very different from the refractive index of the silica core found in fibers – 1.44 to 1.47 – then the surface between the fiber and the chip acts like a mirror, reflecting back the light and resulting in significant losses. The way around it is to coat the chip with an anti-reflective film, but that's a complicated business. Bottom line: This isn't a problem for silica, sol-gel, and polymers, but it is a problem for silicon, indium phosphide, and gallium arsenide.

    The second issue works in the opposite way. If there's a big index contrast, then the light can whistle its way around tighter corners. Bottom line: Waveguides can be shorter and chips can be smaller with materials like indium phosphide.

    Thermo-optic coefficient

    This is a measure of how the refractive index changes with temperature – something that might be required when tuning a device, or might be viewed as a big problem for athermal widgets, ones where the performance doesn't alter if the temperature changes.

    Modulation

    The "fast EO" materials can be used to make modulators. The "slow TO" ones indicate speeds that might be too slow for some switching and tuning requirements.

    Market Forecast

    Here are a couple of charts showing how the use of different materials for integrated optics might change over the next few years, according to Communications Industry Researchers Inc.:

    10593_3.gifCIR is forecasting a big upsurge in the use of indium phospide, largely at the expense of lithium niobate and gallium arsenide:

    10593_1.gifNext Page: Page 8: Vendors: Public

    Agere Systems Inc. (NYSE: AGR)As the former components division of Lucent Technologies Inc. (NYSE: LU), Agere developed silica-on-silicon AWGs for internal consumption.

    Agere is also developing silicon waveguide technology, putting it in a position to compete with Bookham Technology PLC (Nasdaq: BKHM; London: BHM). According to the company's 2001 annual report, it recently started sampling its first product based on this material platform: optical dynamic gain equalizers.

    Agilent Technologies Inc. (NYSE: A)Agilent's contribution to integrated optics is its bubble switch, which was introduced in March 2000. Alcatel SA (NYSE: ALA; Paris: CGEP:PA) says that it still might use it for the core of its CrossLight photonic crossconnect. Right now, however, a MEMS (micro-electro-mechanical system) core appears to have the edge.

    Agilent's bubble is based on a silica chip containing two sets of parallel fluid-filled channels that cross at 120 degrees. Above each intersect there is a miniature heater that creates a bubble, using a technology that's similar to ink-jet printers. If there is no bubble at the intersect, the light travels straight on. When a bubble is introduced, light is reflected down the alternative route. Light can be added and dropped as well as switched. The chip comes as a 32x32 array, or two 16x32 arrays on the same chip.

    See: Alcatel Backs the Bubble
    Agilent Unveils Optical Switching Breakthrough

    Alcatel Optronics (Nasdaq: ALAO; Paris: CGO.PA)In September 2001 Alcatel rose to prominence in the integrated optics world when it bought Scottish startup Kymata Ltd. for $119 million. Alcatel Optronics CEO Jean-Christophe Giroux said in a press release that the acquisition "will allow us to gain at least two years in time-to-market for planar products."

    Alcatel internal developments include products made using both flame hydrolysis deposition and chemical vapor deposition. In Q2 2001 it began full-scale production of silica-on-silicon AWGs for internal use.

    Following the acquisition, Alcatel Optronics made cutbacks, which reduced the total headcount at the former Kymata by 90, or approximately half the remaining workforce at its Livingston, U.K., site (see Alcatel Optronics Slims Down).For more information on Kymata, see its entry under Private Companies.



    Its also worth noting that Alcatel makes a wavelength converter optical circuit based on indium phosphide, which it sells to Optovation Inc., and possibly others. The chip comprises two SOAs in the arms of a Mach-Zehnder interferometer (see Interest Grows in Wavelength Conversion).

    Bookham Technology PLC (Nasdaq: BKHM; London: BHM)

    Bookham is one of just two public companies wholly devoted to integrated optics (the other being Lumenon). Founded in 1988, its first goal, according to its founder and chairman Andrew Rickman, was to prove, using mathematical modeling, that it was feasible to make waveguides with acceptable optical properties out of silicon. This it did, but things moved rather slowly until 1998, when Cisco Systems Inc. (Nasdaq: CSCO), and Intel Corp. (Nasdaq: INTC) invested about $10 million in the then-startup.

    The technology platform, which Bookham calls ASOC (not an acronym), had been launched back in 1994. The big advantage, the company claims, is that it can borrow process techniques used in the semiconductor electronics industry to pattern and alter the wafers. "Silicon has shown itself to be one of the best materials in the world for manufacturing," according to company literature. It is also simple to integrate silicon electronics with optics on the same chip.

    Bookham's standard products include:

    • AWGs with or without integrated temperature control

    • VOA-mux and VOA-demux

    • electronic VOA arrays for 1, 4, or 8 channels

    • optical channel monitor

    • single-fiber transceiver modules

    • 2.5- and 10-Gbit/s receiver modules



    It has also pushed the envelope with more "exciting integration" that couples different types of functions, such as active, passive, and switching, on the same chip. At NFOEC 2001, it demonstrated an add/drop multiplexer subsystem. And in November 2001, it shipped prototypes of a switched-input 40-Gbit/s channel monitor to Nortel Networks Corp. (NYSE/Toronto: NT).

    Financially, Bookham has yet to break even. But it has made two acquisitions: Measurement MicroSystems A-Z for its electronic control systems, in January 2001, and Marconi Optical Components for its active components, in January 2002.

    See: Bookham Gets a Bargain
    Bookham Claims Integration Milestone
    Bookham: Is the Tide Turning?
    Bookham's Not Bookin'
    Bookham Readies New Round
    Bookham's Share Price Soars After IPO

    Corning Inc. (NYSE: GLW)Integrated optics falls under the Corning Photonic Technologies division, which makes a range of other components, including Erbium Doped-Fiber Amplifiers (EDFAs), thin-film filters, modulators, and pump lasers.

    Corning doesn't manufacture its own PLCs; instead it designs them and outsources the manufacturing. It uses either silica-on-silicon or silica-on-silica technology.

    According to product line manager Jaymin Amin, Corning is in pilot production of AWGs. These come in C- and L-band varieties with up to 40 channels at 100 GHz spacing, and a choice of flat-top or gaussian response. Amin claims that Corning is developing an ultra-low-loss version of its AWG technology.

    Amin says the big advantage of Corning's AWGs is that they feature passive athermal packaging, meaning they do not require electrical power, software, or temperature control – a passive component that actually lives up to its name. This is achieved by pigtailing the fibers directly to the "slab coupler" portion of the AWG rather than using output waveguides, and adding a block of material that compensates for the effect of temperature. Amin claims that doing temperature compensation in the packaging is simpler than doing it on-chip because it gives volume advantages: The same chip can work at different wavelengths simply by aligning the fibers in a slightly different position inside the package. Corning acquired the passive temperature compensation technology with the purchase of Siemens Communication Cables in February 2000.

    At OFC 2001, Corning demonstrated a second PLC product, a dynamic gain flattening filter, which can be used to equalize the power of the different wavelengths coming out of an EDFA, for example. The chip comprises a bunch of splitters, and waveguides that act as optical delay lines. The delay in each segment is controlled with thermo-optic switches in a Mach-Zehnder interferometer arrangement, and the result is a device that has a complicated filter response in much the same way that Fiber Bragg Gratings (FBGs) do.

    Corning says it hasn't cut back on its PLC efforts. It is also developing other PLC components, such as switches, VOAs, and dispersion compensators, according to Amin.

    Hitachi Cable Ltd. – majority owned by parent Hitachi Ltd. (NYSE: HIT; Paris: PHA)

    Hitachi Cable is one of the market leaders in PLCs, having them under development since at least 1992. It started AWG production in Feb 1997 with an eight-channel device, according to press releases issued at the time.

    Hitachi's process technology is silica-on-silica rather than the more common silica-on-silicon. Hitachi uses physical vapor deposition (PVD) to deposit the waveguide material onto the substrate, a process that some companies claim gives better results because it does not require a high temperature during deposition.

    Hitachi Cables' product range includes:

    • AWG mux/demux (channel spacings of 50, 100, 200 GHz; maximum number of channels, 48)

    • 1xN and 2xN Optical Splitter (N = 2, 4, 8, 16, 32)

    • optical tap (3dB, 6dB, 10dB, 20dB, etc.)

    • VOA mux (introduced July 2001)

    • "wavelength splitter" (Hitachi Cable's term for an interleaver) (introduced Jan 2001)

    • customized PLCs



    IBM Corp. (NYSE: IBM)Most of IBM's efforts in integrated optics are joint with Kymata (now Alcatel Optronics) although it does have a small amount of independent work, according to Cary Ziter, a spokesman.

    The joint venture involves IBM Zurich licensing its silicon oxynitride (SiON) technology to Alcatel Optronics. Few places have access to this particular material, IBM being one of them, integrated optics foundry Lion Photonix Technologies BV another. SiON has a higher refractive index than plain silica and can be used to create tighter waveguide bends and, hence, smaller components.

    IBM Microelectronics claims it has already demonstrated a seven-channel gain equalizer. It also has its sights on thermo-optic switch arrays.

    See: IBM Moves Into Integrated Optics

    JDS Uniphase Inc. (Nasdaq: JDSU; Toronto: JDU)JDSU's entry here is by virtue of the fact that it bought SDL Inc., which in turn had acquired Photonic Integration Research Inc. (PIRI), one of the early developers of PLCs (see SDL to Buy AWG Vendor for $1.8B). PIRI's expertise is in high-temperature flame hydrolysis of silica-on-silicon, the same technique as Kymata. Thus, in a single stoke JDSU became one of the leading suppliers of AWGs.Matters were slightly complicated by the fact that in June 2000 JDSU had also purchased E-Tek Dynamics, which had an arrangement to buy AWGs from Bookham and then package them (see E-TEK Launches Components).

    JDSU says a "typical" AWG configuration is 16, 32, or 40 channels with 100 GHz spacings, although 50 and 25 GHz spacings are available. The package requires temperature control to keep the output wavelengths locked on the ITU grid.

    Lumenon Innovative Lightwave Technology Inc. (Nasdaq: LUMM)Lumenon makes AWGs for DWDM and CWDM applications, as well as couplers and splitters. At the end of 2001 it started shipping samples of a 16-channel, 100-GHz AWG to PolyScientific, a division of Northrop Grumman, and "a global tier 1 network equipment manufacturer."

    Lumenon's technology is unusual in that is based on "hybrid sol-gel glass" (HSGG), which is based around an organically-modified glass that can be patterned under ultraviolet light. The term sol-gel refers to the process of making glass from liquid precursors. When the solution is deposited as a thin-film, its molecules start linking together to form a gel, and from this state it can be consolidated into a solid.

    The sol-gel can contain both polymers and glass, which provides access to a broad range of materials properties, from plastics at one end of the scale to glass at the other. Lumenon claims this gives it a key advantage over its competitors.

    Founded in 1998, Lumenon went public very early on by doing a reverse merger with a small public company, but in many ways it is still a startup.

    Marconi Optical Components (MOC) – recently acquired by Bookham (see entry for Bookham Technology

    Marconi launched its components division in December 2000 to commercialize technologies coming out of its R&D laboratories in the U.K. – Caswell Technology in Northampton and Applied Technologies in Chelmsford. A year later, it showed a demonstrator of its first product, a widely tunable laser.

    MOC has planned to include active components, like the tunable laser, on the same optical chip as passive components (a plan that is probably unchanged by the Bookham acquisition).

    However, after sinking a significant amount of cash into the venture, its parent ran into serious financial trouble and turned off the funding tap (see Marconi Stock Tanks and Heads Roll at Marconi). Bookham snapped up the division for a mere £19.7 million ($28.7 million).

    See: Bookham Gets a Bargain
    Marconi Claims Tunable Laser Advance
    Marconi Components: Up For Sale
    Marconi Joins Optical Components Field

    NTT Electronics Corp. (NEL) – majority owned by parent NTT Group (NYSE: NTT)

    NEL is considered to be one of the leaders in AWG components in terms of revenues and product range, according to Communications Industry Researchers Inc. (CIR). Indeed the company now describes its AWGs and DWDM components as "a mainstay of the new business". Its passive components are based on silica-on-silicon, while its active ones use a silicon submount.

    NEL's product range includes:

    • all kinds of silica-based AWGs, including C- and L-band; ultra-low-loss; up to 80 channels, down to 50 GHz; AWGs with built-in temperature controllers, and athermal AWGs, which don't require temperature control; and a 40-channel polarization maintaining AWG

    • eight-VOA array

    • optical channel level controller, comprising AWG and VOA arrays in the same package, connected by fiber ribbon, up to 40 channels

    • thermo-optic switches, in 8x8 and an 8 array of 2x2 configurations

    • 1310 and 1550 nm Wavelength Division Multiplexing (WDM) transceivers, comprising active components integrated onto a silicon platform.



    Actually, a list like this is probably underplaying NEL's achievements. For example, its ultra-low-loss AWG has an optical loss of 2 decibels, whereas other components vendors consider ultra-low-loss to mean 4 to 5 dB – in real terms, that's nearly twice as much light loss. And it has reported some significant firsts at conferences, such as a 16x16 thermo-optic switch, and a 200-channel AWG.


    Next Page: Page 9: Vendors: Private

    Agility Communications Inc.

    Founded by University of California, Santa Barbara (UCSB) professor Larry Coldren, Agility has raised more than $186 million so far, $168 million in equity and $18 million in leasing agreements. Perhaps more significantly, the most recent round of funding included two strategic investors, Ciena Corp. (Nasdaq: CIEN) and Tellabs Inc. (Nasdaq: TLAB; Frankfurt: BTLA). This underscores the lead Agility has developed in making widely tunable lasers. It has gone farther than most in monolithically integrating other components – namely, a modulator and amplifier – on-chip with its laser.

    Agility claims to have about 35 paying customers, of which about half a dozen are big equipment vendors.

    See: Agility Gets $83M Third Round
    Agility Packs Three Into One
    Agility Unveils Long-Haul Laser
    Agility Launches First Product
    Agility Communications, Inc

    Axon Photonics Inc.

    Axon is developing silica-on-silicon for passive components and indium phosphide for active components, which it plans to combine to produce hybrid subsystems later on. Its first products are a 40-channel 100 GHz AWG, an optical supervisory channel filter, and Semiconductor Optical Amplifiers (SOAs) . Vertex Management Inc. is an investor.

    Cirrex Corp.

    Cirrex is developing integrated optical subsystems using automated assembly techniques borrowed from the microelectronics industry. It's combining ceramic substrates, electronic integrated circuits, silicon optical benches, and silica waveguides. The company's first product is a mini optical add/drop multiplexer.

    See: Cirrex Tidies Up
    Startup Automates Module Manufacture

    Cisilias A/S

    Denmark-based Cisilias is one of a number of companies developing low-power, low-cost amplifiers that could be the starting point for making integrated subsystems for metro equipment. They work on the same principle as Erbium Doped-Fiber Amplifiers (EDFAs) but use erbium-doped waveguides rather than erbium-doped fiber to boost the power of light going through them.

    See: Which Amp is the Champ?

    Clarendon Photonics Inc.

    Clarendon is a stealth-mode company funded by New Enterprise Associates (NEA), Sevin Rosen Funds, and Vortex Partners. It's planning to make "high-density monolithic optical components." It's thought to be working with photonic crystals – viewed as a promising technology for avoiding the problems of having to make passive devices in one material and active devices in another.

    See: A Gem of an Idea
    Clarendon Raises Cash, Opens HQ

    DenseLight Semiconductors Pte. Ltd.

    This Singapore startup appears to be having some success in making integrated components using indium phosphide, one of the few materials that can be used for making both passive and active devices.

    The startup plans to launch its first truly integrated products, a tunable laser and an optical channel monitor, at the OFC trade show in March.

    See: DenseLight Goes to the Quantum Well

    GalayOr Networks

    This Israeli startup touts "micro-opto-mechnical systems" or MOMS, as opposed to MEMS (micro-electro-mechnical systems). The technology is based on "suspended waveguides fabricated on chips" that can be combined with other devices to make optical switches, tunable DWDM components, and dispersion compensators.

    See: Israeli Startup Gets $8M Seed Funding

    Galian Photonics Inc. – formerly Bandgap Semiconductor Inc.

    Another startup working on photonic crystals, thought to be at an early stage of development.

    See: 2001 Top Ten: Components

    Gemfire Corp.

    Gemfire's philosophy is to use whatever material is the best for particular devices, using hybrid integration technologies. It says it can work in ceramics, silica, gallium arsenide, silicon, and polymer – and points out that its founder, Dr. David Deacon, holds five key patents relating to indium phosphide.

    Gemfire's products include a pump laser array and a VOA array.

    The startup has an impressive lineup of investors, including Cisco Systems Inc. (Nasdaq: CSCO), Corning Inc. (NYSE: GLW), and Intel Capital.

    See: WaveSplitter Gets Dynamic Over DWDM
    Gemfire Comes Out Blazing

    InPlane Photonics Inc. – formerly Key Optics Inc.

    This stealth-mode company has been financed by Morgenthaler

    Intense Photonics Ltd.

    Scotland's Intense Photonics is using a technology called quantum well intermixing to make monolithically integrated optical components from III–V semiconductor materials (gallium arsenide and indium phosphide). Quantum well intermixing enables it to grow a crystal and then modify its optical and electrical characteristics afterwards, using a diffusion process.

    Intense Photonics' first product is an array of ten 980nm pump lasers on a single gallium arsenide chip.

    See: Intel Invests in Scottish Foundry
    Intense Photonics Puts on a Show
    Intensive Care
    Scotland Spawns Component Startups

    Ionas A/S

    This Danish startup offers foundry services for silica-on-silicon devices.

    See: Micro Managed Photons A/S

    Kamelian Ltd.

    The U.K.’s Kamelian is making SOAs capable of boosting the light power of multiple wavelengths at the same time. In this respect, it’s following in the footsteps of Genoa Corp., which already has a similar product.

    Kamelian views its SOAs as building blocks for hybrid integration projects, based on its indium phosphide developments and other vendors’ passive devices, such as AWGs. It’s already demonstrated an optical add/drop multiplexer based on three of its SOAs and has talked about developing all-optical wavelength converters and signal regenerators.

    See: Kamelian to Upstage Genoa?
    OFC's Hot Products
    Kamelian Gets Green Light

    Kymata Ltd. – acquired by Alcatel Optronics (Nasdaq: ALAO; Paris: CGO.PA)

    See: Kymata Sold for $119 Million
    What's Cooking at Kymata ?
    Kymata: For Sale?
    Kymata CEO to Step Down
    Kymata Goes West

    Lambda Crossing Ltd.An early-stage Israeli startup developing silicon-based optical components including optical add/drop multiplexers, optical channel monitors, and wavelength equalizers.

    Investors include Agilent and Lucent Venture Partners Inc.See: Agilent Invests in Components Startup

    Lion Photonix Technologies BV

    Spun out of University of Twente's MESA+ Institute in the Netherlands, Lion specializes in silicon oxynitride and silica developments.

    Lightwave Microsystems Corp.As noted earlier, Lightwave Micro has been around for a while. It started out developing polymer products and then refocused on silica-on-silicon AWGs. It's gone further than most in adding other devices to create an optical add/drop multiplexer and a VOA.

    It filed for an IPO last year but didn't manage to float before the market went sour.

    The startup currently ranks eighth in Light Reading's Top 10 Private Companies.

    See Lightwave Microsystems
    Lightwave Micro Launches VOA Mux

    Lynx Photonic Networks

    Makes solid state optical switches in silica-on-silicon and lithium niobate.

    See: Look Ma, No Moving Parts!
    Infineon Caught Cheating
    Lynx Claims Optical Switch Advance

    Mesophotonics Ltd.

    London-based startup developing photonic crystal technology.

    See: Crystal Startup Gets $4M
    NanoOpto Thinks Small

    MetroPhotonics Inc.

    Ottawa startup developing eschelle gratings.

    See: Echelle Gratings Make a Comeback
    OFC's Hot Products
    MetroPhotonics Discloses Product Plans

    Micro Managed Photons A/S (no Website)Danish startup working with exotic optical effects of thin gold films.

    See: Startup Strikes Gold
    Danes Break Through in Optics

    NanoOpto Corp.

    A startup that may be leading the race in developing photonic crystal technology, its investors include some blue chip VCs.

    See: NanoOpto Thinks Small

    OpsiTech SA

    OpsiTech's work originated at the Laboratory for Electronics,Technology, and Instrumentation (LETI), operated by Commissariat à l'Énergie Atomique in Grenoble, France.Founded in July 2000, OpsiTech raised €6.6 million (US$5.75M) in Sept. 2000 and another €6.6 million in Sept. 2001, from venture capitalists Banexi Ventures, Innovacom Ventures, Sudinnova Partners, Emertec development fund, and CEA-Valorisation funds.

    OpsiTech started out to commercialize "mobile" waveguides, which are cantilevered beams in miniature that can be moved with electrostatic forces in order to align an input with different outputs. This offers better performance than, say, thermo-optic switches, the company claims.

    However, there may have been a change in direction, as OpsiTech's Website now offers evaluation samples of optical fan-out; 50-, 100-, and 200-GHz frequency spacing mux/demux; interleavers; and cyclic interleavers. There is no mention of mobile waveguides.

    Optenia Inc.

    A spinoff of Mitel Networks, developing echelle gratings, Optenia's first product is an optical channel monitor.

    See: Optenia Offers Monitor
    Echelle Gratings Make a Comeback

    Optical Micro Devices Ltd. (OMD)

    Swindon, U.K., startup offering foundry services specialized in silica-on-silicon PLCs and silicon microbenches.

    Photodigm Inc.Founded in mid 2000, Photodigm pulled in some interesting investors for its seed round ($2.5 million in January 2001), namely Compass Technology Partners, Arkoma Venture Partners, Corning Inc. (NYSE: GLW), and TriQuint Semiconductor Inc. (Nasdaq: TQNT).

    It is developing what it calls "grating outcoupled laser technology" or GO laser. The idea originated with Gary Evans, a professor of electrical engineering at Southern Methodist University in Dallas.

    Qusion Technologies Inc.Qusion's long-term aim is to make integrated optical circuits in indium phosphide. But it obviously has a long way to go. It has announced one product, a 40-Gbit/s electroabsorption modulator (EAM).

    Funding-wise, Qusion scored seed capital of $1 million and a $9.25 million A round in April 2001. The lead investor was VantagePoint Venture Partners. Other investors were Cathy Lego, Wasserman Ventures, and W.R. Hambrecht & Co.

    See: Qusion Touts Wideband Modulator
    Qusion Thinks Big on Optical Chips

    Redfern Integrated Optics

    Funded in March 2001, Redfern is a spinoff from Redfern Photonics Pty Ltd., which exists to commercialize photonic technologies. Most of the expertise in its portfolio comes out of the Australian Photonics Co-operative Research Centre in Sydney.

    See: Redfern Breaks Through in Optics
    Redfern Attacks Chromatic Dispersion
    Redfern Raises $28 Million

    Sparkolor Corp.Sparkolor comes from the same stable as Gemfire. David Deacon founded both companies and both companies believe in hybrid integration – using whatever materials are best for different parts of a subsystem. Its first product is an indium phosphide-based tunable laser, but that, it promises, is just for starters.

    See: Sparkolor Secrets Surface
    Sparkolor Plays Catch Up

    Symmorphix Inc.Another outfit like Denmark's Cisilias, making low-power, low-cost amplifiers that are like miniature versions of EDFAs.

    See: Symmorphix Joins the Amp Camp

    Teem Photonics

    Teem was first out of the gate with "erbium-doped waveguide amplifiers" (EDWAs) and has a lead in integrating these devices with other passive components, such as splitters.

    See: Which Amp is the Champ?
    Teem's Tiny Metro Amp Makes Waves
    Teem Raises Component Steam

    Telephotonics Inc.Telephotonics isn't just another AWG developer. It's aiming to make all sorts of components by bringing together three ways of controlling light – thermo-optics (heat), magneto-optics (magnetism), and electro-optics (applied voltage).

    All the same, the first products announced by the company are pretty much what AWG makers are also focusing on: VOA arrays and optical channel monitor arrays.

    See: Telephotonics Starts a Price War
    Telephotonics Unveils Metro Package
    Startup Creates Component Cocktail

    Terahertz Photonics Ltd.This Scottish startup is offering to design and manufacture silica-on-silicon PLCs for others, using polymer technology that eliminates some of the normal steps in making such devices. The bottom line is significantly lower costs and faster time to market, according to the company.

    See: Scottish Startup Shrinks
    Tunable Filters Go Solid State

    ThreeFive Photonics BV

    This Dutch startup is into monolithic integration using indium phosphide. It's already made a 2x2 switch, although the first version wasn't suitable for commercial use; losses were too high.

    See: Chipmaker ThreeFive Secures Funding
    Dutch Startup Seeks Passage to Indium

    Translume Inc.

    Translume is aiming to make optical components by burning patterns into glass blocks with a high-power pulsed laser – a method known as "direct-writing."

    See: Can Translume Direct-Write Success? (PS: The "Article Talk" messages after the story are especially worth reading.)

    WaveSplitter Technologies Inc.

    Wavesplitter is another silica-on-silicon AWG maker that filed for an IPO last year and didn't make it before the market went sour. Since then, it's formed some interesting alliances, with Gemfire and with Infineon Technologies AG (NYSE/Frankfurt: IFX), to jointly develop some integrated optical subsystems.

    See: WaveSplitter Gets Dynamic Over DWDM
    IPO Window Shuts Tighter
    Wavesplitter Files for $155 Million IPO

    Other companies that may be working on PLCs include:

    Avicron Technologies Inc.
    Scion Photonics Inc.

Subscribe and receive the latest news from the industry.
Join 62,000+ members. Yes it's completely free.

You May Also Like