Optical innovations go old school

The Optical Fiber Communications/National Fiber Optics Engineers Conference show in Anaheim, Calif., this month contained a mix of optimism and dread along with a strong dose of déjà vu. The oft-repeated sentiment that consolidation is needed to remedy the cutthroat optical component industry could have reminded show-goers of any of the shows of the last several years, in which the same thing was said.

And even the new, gee-whiz technologies weren't so new. Passive optical networking, no spring chicken to be sure, was arguably the star of this year's show by dint of the attention focused on deployments by Verizon Communications and Japan's NTT. In fact, some of the optical technologies showing particular advancement today toward the future of commercial networks have been around for a while already — and to some extent, already counted out.

CDMA technology, commonly used in wireless networks, is nothing new, nor is applying it to fiber optic networks. A few vendors, including APN and CodeStream Technologies, offered optical CDMA (OCDMA) products during the telecom bubble. But the spectral efficiency inherent in those attempts was low, and signals would interfere with one another when they got too crowded. After those start-ups failed (neither APN nor CodeStream exist today), OCDMA dropped out of the typical telecom carrier lexicon.

Meanwhile, two years ago, the U.S. government's Defense Advanced Research Projects Agency began urging researchers to look into OCDMA as a way of increasing efficiency in wavelength-division multiplexed (WDM) networks. The resulting research efforts are ratcheting up spectral efficiencies slowly and remedying multi-user interference with time-gating, making OCDMA increasingly more interesting to the private sector once again.

“OCDMA may yet have its day,” said RHK co-founder John Ryan at this year's OFC/NFOEC show. “This is not something that's going to remake PON economics next year, but over the next several years, one could see a rethinking of these networks.”

OCDMA operates by “encoding” wavelengths at the transmitting end and “decoding” them at the receiving end. The technique's obvious security implications are actually a matter of debate among telecom professionals. The more interesting aspect for telecom engineers is OCDMA's potential to increase bandwidth utilization in multi-user networks. Unlike WDM, in which each signal gets to hoard its own separate wavelength, OCDMA lets all the different signals crowd the pipe in a capacity free-for-all and sorts them out at the other end by reading their “codes.”

This increased capacity utilization could be useful in metro and long-haul networks, but perhaps especially in broadband aggregation, where some users (with DSL or fiber to the home) might leave some capacity unused.

“[Broadband networks are] at the front of our brains,” said Joseph Bannister, who is researching OCDMA for the University of Southern California's Information Sciences Institute.

At OFC/NFOEC, researchers from Telcordia Technologies and Nomadics Inc. described a low-loss method of encoding signals by passing them through wavelength-selective micro-ring resonators (picture tiny optical cul de sacs) and heating them slightly, thus changing the circumference of the cul de sacs and slightly shifting the phase of the signals. The technique, which the researchers say is cheaper and more integrated than previous methods, wasn't possible a few years ago. And the community of micro-ring resonators, now sparse, should grow with demand in plenty of non-telecom applications.

With its current momentum, when might OCDMA be deployed in commercial telecom networks?

“I don't know,” said Telcordia's Shabab Etemad. “I've been giving talks to major network providers, but at this stage, they look at it as a laboratory curiosity. Which it is, in some ways.”

With DARPA's requirements driving efficiencies ever higher, OCDMA may grow more curious.

Microelectrical mechanical systems (MEMS), the tiny tilting mirrors used to switch optical signals without electrical conversion, may have a tarnished name in the U.S. market since Lucent Technologies and Nortel Networks ceased development of their MEMS-based switches in 2002 (Nortel having famously paid $3.25 billion in stock for its OPTeraConnect PX only two years earlier).

But did the technology get a bum rap? Marlene Bourne, former In-Stat analyst and vice president of research for the microtech publication Small Times, thinks so.

“A lot of the negative aspects of MEMS — the fears that they're fragile, that they've got moving parts that are going to stick — a lot of that was coming from [vendors of] competitive technologies who were seeing MEMS gaining a strong foothold and trying to stop the train a little,” Bourne said. “Customers quickly moved past those issues. Now it's clearly a proven next-generation technology. It's just that no one's willing to invest in it.”

One of the only new platform launches at OFC/NFOEC was the combination dense WDM (DWDM) and MEMS-based optical switch from Lambda Optical Systems. Also at the show, Lambda Optical's MEMS supplier, Glimmerglass, unveiled what it called the “world's largest fiber switching solution,” a switch the size of a microwave oven capable of connecting up to 1360 fibers.

Glimmerglass and Lambda Optical both acknowledge (and Bourne agrees) that not much has changed in the MEMS technology itself during the past few years. Rather, the two companies learned a thing or two about what to do with it. Glimmerglass holds a long list of patents for robotics that govern the precision of MEMS' moving parts and intelligence for inter-mirror communication. Lambda Optical distinguished itself by combining MEMS with DWDM and focusing on metro and regional networks.

Lambda Optical chief executive Irfan Ali admitted that his company hasn't found many eager customers in the U.S. so far, aside from the U.S. Naval Research Laboratory, which uses its gear to transport satellite images terrestrially. But trials are heating up in other parts of the globe, he said.

While Lambda Optical, together with Calient Networks, carries the torch for MEMS as an alternative to the high price of optical components in optical-electrical-optical conversions, another newly launched vendor, Infinera, is promising to lower the cost of those conversions by integrating the components into chips. Michael McLaughlin, Glimmerglass' vice president of corporate development, says the two technologies will complement rather than contradict each other going forward. To support the claim, he pointed out that one of Glimmerglass' investors, Mobius Venture Capital, is also an investor in Infinera.

Silicon-based optical communication could be viewed as the telecom equivalent of the ancient alchemist's long-pursued dream of turning lead into gold. Mass-produced silicon is much smaller and cheaper than indium phosphide and gallium arsenide, the materials most commonly used in today's optical networks. But silicon can't emit light; that is, its electrons cannot be stimulated to release photons. Still, the drawback hasn't stopped researchers from striving to find out just how much can be done with silicon.

In February, the science journal Nature published an article submitted by Intel researchers claiming to have demonstrated the world's first “continuous-wave Raman silicon laser.”

In 2001, scientists at UCLA created the Raman silicon laser, a few centimeters of silicon that amplified light in a way that would take thousands of meters of fiber. But it was a bursty signal, not useful in communications. To achieve the first continuous-wave version, Intel had to overcome the problem of “two-photon absorption” (TPA). As light travels through a silicon waveguide, occasionally two photons will hit a silicon atom at once, knocking loose one of the atom's electrons. It's rare, but the higher the laser power, the more it happens. And if enough electrons get loose, they can absorb some of the laser light and diminish the power of the signal.

Intel designed its silicon waveguide like a semiconductor, putting two different types of doped silicon on either side of the undoped silicon that formed the waveguide. When researchers applied voltage to this device, it swept the stray electrons out of the laser's way, avoiding TPA.

Intel's invention doesn't obviate optical components by any means. To get the laser started, it still relies on the same kind of pump — an external cavity diode laser — used commonly in today's telecom networks. But the company envisions its achievement as a milestone on the still years-long path toward a variety of silicon photonics applications such as amplifying signals in long-haul networks and generating multiple lasers with different wavelengths from the same pump beam.

Bahram Jalali, professor of electrical engineering at UCLA, doubts silicon's application in optical communications will grow far beyond niche applications such as channel equalizers (which even signals by amplifying weak ones and attenuating strong ones) and free-space optics. Silicon economics demand either high-priced products or a lot of volume, Jalali said, so while silicon lasers might do well in, say, military applications where money is no object, they will probably be too expensive for telecom carriers.

“Unfortunately, the volumes are not high enough in the optical [communications] market,” he said, acknowledging that this point of view puts him in the minority of researchers in the field. “They just don't sell anything in millions of quantities. That might change as Internet traffic increases, but it's still far away.”

Also, no one knows what the next achievement in silicon might bring. Luxtera, a fabless semiconductor company that focuses on silicon integration of high-speed fiber optics, declined to comment for this article, in preparation for a major announcement this week.

At the OFC/NFOEC show, RHK's John Ryan wondered if the next several years could bring a new ecosystem in silicon-based optical components.

“I hope so,” Jalali told Telephony. “But we've been saying that for years.”