Intel Pushes Silicon Modulator
Thirty is such an awkward number in telecom. Intel assures the world it expects to push to 40 Gbit/s (OC768) soon -- the device does have the headroom for that -- and the goal is to produce a version that does 100 Gbit/s, says Mario Paniccia, director of Intel's photonics technology lab.
The results are being published today in Optics Express, a peer-reviewed journal (with a URL that provokes giggles) from the The Optical Society (OSA) .
Intel's main point is that the technology is catching up with high-end needs. The 30-Gbit/s model represents the latest generation of the 1-Gbit/s modulator introduced three years ago. (See Intel's Modest Modulator .)
Intel's goal isn't to displace long-haul optics. The target is rack-to-rack connections, and even chip-to-chip connections eventually, which means the paramount factor is cost. The idea is to create photonics devices that not only are made in silicon, but can be built using the relatively cheap complementary metal-oxide semiconductor (CMOS) manufacturing processes used for the majority of chips.
"The goal is not to get better performance than the III-Vs," Paniccia says, referring to the class of materials including indium phosphide (InP). "It's about getting 90 percent of the performance."
But is a 100-Gbit/s modulator the right approach? Startup Luxtera Inc. , also pursuing silicon photonics, says it's best to stop at 10 Gbit/s, relying on multiplexing to pack together a 40- or 100-Gbit/s signal.
That's more practical, Luxtera says, because electronic signals in CMOS can't comfortably surpass 10 Gbit/s yet.
"If, in a real system, you can't build transistors and circuits that run at that speed, it seems kind of a moot point," says Alex Dickinson, Luxtera CEO. "That's led us to go down the WDM path." (See Luxtera Launches Silicon Optics.)
Besides, when it comes to CMOS optics, Dickinson says the modulator hasn't been the "limiting factor" anyway. "It would be a lot more profound if it was the modulator driver built in a CMOS process driving a silicon modulator," he says.
If nothing else, today's announcement shows Intel remains committed to its silicon photonics research, even after selling off much of what had been its optics division. Even amid speculation of a telecom selloff, Intel was presumed to be keeping Paniccia's crew, on the basis that they're doing long-term research as opposed to product development. (See Cortina's Comfortable With Intel Inside and Will Intel Trash Telecom?)
Intel has also talked about work with silicon lasers, producing first a device that required an external pump laser, then a hybrid silicon/InP part. (See Intel Claims Laser Breakthrough and Intel Fires Up Silicon Laser.)
And Intel is working on pieces beyond the modulator. "We have a silicon germanium [SiGe]-based photodetector at 10 Gbit/s that comes pretty close to GaAs [gallium arsenide] performance," Paniccia says.
Generally speaking, a lot of progress has been made in silicon photonics in just a few years, much of it by university teams. Intel has been working with the University of California, Santa Barbara on silicon photonics. Other university efforts include programs at UCLA and the University of Surrey, Guildford.
— Craig Matsumoto, West Coast Editor, Light Reading