It's cool to be uncool, at least when it comes to lasers. Components startup ASIP Inc. made its debut yesterday by announcing what might be a significant advance in uncooled laser technology, as well as the integration technology that made it possible (see ASIP Delivers Integration Platform).

Uncooled lasers are mostly used in datacom applications, where the fact that they don't need a power-guzzling thermoelectric cooler allows them to be put into much smaller packages.

ASIP's CEO, Mike Decelle, claims his company has produced one of the first uncooled 10-Gbit/s, 1310 nanometer electroabsorption modulated lasers (EMLs) -- a laser with a modulator integrated on the same piece of indium phosphide. This stands in contrast to most other commercial uncooled lasers, which are directly modulated -- meaning pulses are created by turning the laser itself on and off, rather than using a modulator.

The significance of this is that EMLs produce cleaner signals than directly modulated lasers, leading to fewer bit errors at the end of the link. That's because direct modulation creates small wobbles in the wavelength of the laser, or "chirp," which results in greater dispersion as the light travels down a fiber. Chirp also limits the distance the signal can travel, although with 1310nm devices fiber attenuation is also a significant effect at distances over 40 kilometers.

Right now, the longest distance that's been claimed for uncooled 1310nm laser technology is 40 km (see OFC Is Highly Transponderant). Agere Systems (NYSE: AGR) claimed this span for its directly modulated lasers, as did OptronX Inc., a startup whose technology was bought recently by JDS Uniphase Corp. (Nasdaq: JDSU; Toronto: JDU) (see JDSU Buys Part of OptronX).

And then there's Optium Corp., which claims to be first with uncooled EML technology. Its device, which was unveiled in March this year, uses an on-board microprocessor to compensate for the effects of temperature, says Optium's VP of sales and marketing, Tony Musto (see Optium Touts Tiny Transceiver).

ASIP claims to have demonstrated 60km performance in the lab, although the part that's shipping to customers today is listed as 10 km. But that's more an issue of standards than performance, according to Decelle, who notes that 1550nm devices are usually specified for 40km links.

"Until now, EMLs have been considered too complex and costly to use in enterprise applications," says Decelle. ASIP's technology challenges that assumption: The startup claims that, thanks to its technology, it can produce EMLs at a cost that's competitive with directly modulated lasers. Of course, if ASIP can't keep that price promise, its current product strategy may disintegrate entirely.

The key to bringing the cost down, says Decelle, is ASIP's integration technology platform, which, he contends, has much higher yields than standard indium phosphide manufacturing.

"If you take a look at manufacturing methods today, these tend to emphasize complex growth methods. Every time you want to construct a different optical function you need different combinations of semiconductor layers to be deposited on a wafer."

A typical technique for producing multiple optical functions is called "selective regrowth." To make an EML using this method would involve first depositing the layers required to make a laser, then etching a hole through those layers. New layers of material can be grown in the hole to create the modulator region. In practice, it's even more complicated than this, Decelle contends, requiring up to four regrowth steps to make an EML.

The problem is that regrowth techniques like this have very poor yield, and each regrowth step compounds the problem. In fact, Decelle attributes the generally poor yields in indium phosphide manufacturing to the use of regrowth.

ASIP's solution is to avoid regrowth. It does this by growing all the layers for the different optical functions in a continuous sequence -- so that different regions (e.g., laser and modulator) end up on top of each other -- and then carving out the device in the different regions using etching. Etching is a high-yield process, Decelle claims, pointing to its prevalence in the electronics industry.

The key to making this possible is being able to channel light vertically from one layer to the other. To do this, ASIP has developed what it likes to call an "optical via" -- analogous to the electronic vias, or connections, used to connect different layers on a printed circuit board. The via is a tapered waveguide: As the waveguide gets narrower, light is forced into the waveguide in the region below it.

ASIP calls its technology platform "Asymmetric Twin Guide," referring to the process of optimizing the design of the waveguides in each region. The startup plans to use this platform as a starting point to produce a range of other integrated optical devices, but so far it hasn't said what they might be.

Other startups point out that this isn't the only way of avoiding regrowth methods and improving yields. Intense Photonics Ltd. and DenseLight Semiconductors Pte. Ltd. are both backing an alternative, called "quantum well intermixing" (see Intensive Care and DenseLight Goes to the Quantum Well).

Founded in July 2000, ASIP's technology was developed at Princeton University. It scored its first and only round of funding soon after, the amount undisclosed. Its investors are Redpoint Ventures and Multilink Technology Corp. (Nasdaq: MLTC).

According to Decelle, the startup is already realizing revenues from commercial sales and sampling of its uncooled EML, but he notes that it will need to raise more funding to reach cash-flow positive.

— Pauline Rigby, Senior Editor, Light Reading
www.lightreading.com