Electronic, photonic integration question shifts from 'how' to 'why'

Just a few years ago, a tight marriage between electronic control and wavelength translation was seen as a driver for bringing electronic and photonic components to a single substrate. While III-V materials had their proponents, silicon was seen by most researchers as the great equalizer, making popular the phrase "silicon optical bench." Advocates of such integration anticipated seeing the melding of electronic control plane and optical transport plane converging into a small form-factor transceiver package in the near term, followed by traditional BGA semiconductor packages before the first decade of the 21st century closed.

But the telecommunication recession, that began in late 2000, changed the dynamics for achieving silicon optical bench integration. Instead of asking how such melding of components might be accomplished, equipment developers started wondering why such integration was necessary in the first place. If high-density all-photonic switching was a long-term play, and optical cross-connect mesh topologies in long-haul networks were finding slow acceptance, then why did it make sense to meld electronic control and photonic switching elements in a single component, or even a single line card? Maybe the segmentation between control and transport planes represented the most economical option for broadband carriers.

"Any time you retain an electronics layer, the requirement in transponders and in conversion equipment becomes a nightmare for carriers, in terms of both capex and opex," said Hanan Anis, founder of optical ADM startup Ceyba, Inc. "But as much as a pure optical pass-through concept makes sense, the big 3D MEMS components did not price very well. We believe that some middle ground in optical and electronic is necessary." Anis' colleague at Ceyba, John Gruber, director of networking architecture, addresses the need to move beyond classical MEMS capabilities in his article in this week's In Focus.

But shrinking an optimal O-E-O architecture down to sub-line-card size is not as easy as it appeared on paper. A little-noticed victim of the 2001-2002 recession was the fully-automated optical subassembly plant. As semiconductors and optical components shifted from Lucent Technologies to Agere Systems, Lucent's Manufacturing Realization Center (Brienigsville, Penn.) was declared superfluous and shut down, with many developers shifting back to New Jersey fab sites. Similarly, Nortel Networks Inc.'s impressive Paignton, U.K. optical components facility was shuttered, with engineers and manufacturing experts dispersing to several area startups in England.

To be sure, not all substrate-integration efforts are failures. Bookham Technology, for example, gained a U.S. manufacturing foothold when it opened a Maryland manufacturing plant, located (not coincidentally) near the headquarters of the U.S. National Security Agency. Earlier this year, Bookham acquired the optical components business of Marconi Communications, giving the company the chance to combine Marconi's passive optical devices with Bookham's "ASOC" (Application Specific Optical Circuits) technology. But optocomponent expansion stories since 2000 have been the exception rather than the rule.

Sometimes, this has changed the economic dynamics of when and why a company decides to build its own fab facilities, for either active photonic elements or passive components. In the optical heyday of the mid-1990s, it made sense for even system-level OEMs to build their own facilities to maintain tight control of intellectual property. Ciena Corp. (Linthicum, Maryland), for example, developed its own fab for manufacturing waveguide gratings used in the manufacture of its dense wave-division multiplexer equipment.

Now, even the component manufacturers are looking to outsource certain steps in the manufacturing process, electing to turn over either silicon and III-V wafer manufacturing, or back-end packaging and assembly operations, to general-purpose contractors like Flextronics, Inc. Arlon Miller, vice president of marketing at tunable-laser specialist Agility Communications, Inc. (Santa Barbara, Calif.), said that "it's a careful balance - the photonics industry is almost at a point of using the fabless semiconductor model, but there are still crucial points in the manufacturing chain that you don't want to outsource if you're concerned about quality assurance."

Silicon tends to be the preferred substrate by those moving from passive device integration to inclusion of active electronic control elements. Bookham, for example, has worked with various multiplexer/VOA and modulator/VOA combinations in its ASOC lines.

When laser vendors such as Agility or ADC/Altitun look at substrate integration, however, the choice is dictated by the material best for the semiconductor laser, often indium phosphide or a derivative. Agility starts with the assumption of using InP for laser and grating quality in its Sampled Grating Distributed Bragg Reflector (SGDBR) laser, then adds additional functions to the substrate to create a Photonic Integrated Circuit. The company's first year of product shipments was dominated by the 3040 continuous-wave tunable laser, but at the recent Optical Fibers in Communications conference, Agility moved up to laser assembly level by debuting the 3105, a laser that integrates a semiconductor optical amplifier.

At first blush, the role of MEMS, does not seem to be central to optoelectronic device integration on single substrates. Particularly in the case of micromirrors, where structures are often gimbaled or driven through gear-based systems, it makes more sense to scale from one-dimensional to 2D and 3D micromirror arrays, than to risk influencing reliability by combining micromirrors with other devices. In general, micromirror subsystems from vendors such as Texas Instruments Inc., Nortel Networks Inc., Agilent Technologies Inc., Optical Switch Corp., and OMM, Inc. are broadened by offering micromirror arrays of varying densities, rather than integrating them with other devices.

But MEMS are not always synonymous with micromirrors. There are many good ideas for integrating other types of MEMS structures with electronic logic and passive photonic devices.

Contributor Yves LeMaitre, vice president of marketing at LightConnect Inc. (Newark, Calif.), and Robert J. Monteverde, product marketing manager at Silicon Light Machines (Sunnyvale, Calif.) — acquired by Cypress Semiconductor — provide us with examples in their articles, such as the role of diffractive MEMS, created by structuring "ribbons" of silicon nitride over a substrate.

Nanovation, Inc. was a leader in designing structured waveguides on a substrate where they could be combined with multiplexers and modulators. Unfortunately, Nanovation went out of business last year, but waveguide MEMS research continues at institutions like Northwestern University, UCLA, and University of California - Santa Barbara.

An alternative tactic to building machined structures on top of a substrate, is to use photolithography to create substructures within a substrate. At the March OFC, NanoOpto Corp. (Somerset, N.J.) unveiled the first in a line of what the company calls "subwavelength elements," using a lithography-created mold to stamp a an etched image into a silicon wafer. NanoOpto president Barry Weinbaum called the nano-imprint lithography process "pre-built for integration." In the first step, NanoOpto is introducing polarizers, polarization beam-splitters, and combiners in products to be released this month. But by late summer, NanoOpto will be talking to passive and active device manufacturers on the possibility of integrating other elements on board the nano-imprinted substrate.

Hubert Kostal, vice president of marketing and sales at NanoOpto, said that the integration of passive and active optical elements either must rely on working around MEMS substructures, or must use layering of elements applied through semiconductor methods such as lithography. The layering used in NanoOpto's lithography methods is more conformant with semiconductor processes than methods such as the use of on-chip waveguides, he said. While the telecom recession has hit all optical communication markets over the last two years, Kostal said that cost constraints experienced by carriers actually may hasten, rather than push out, some adoption of on-chip electronic and photonic integration.

"Obviously, this will be dependent on the application, but in metropolitan and access markets, the trends may be to hasten integration," Kostal said. "You need to have passive and active optical device integration via silicon optical bench or similar methods, in order to get costs down to the level where carriers can afford advanced optical deployment."