Inphi Celebrates Life After 40-Gig
The company announced this week that it's shipping a 100-Gbit/s coherent transimpedance amplifier (TIA), a chip that represents one key step towards serial 100-Gbit/s optical transport. (See Inphi Intros 100G Parts.)
Inphi claims the TIA is the first one compliant with Optical Internetworking Forum (OIF) standards for coherent 100-Gbit/s detection, which would make it a milestone in the development of the 100-Gbit/s generation. The chip is being used by u2t Photonics AG , which announced its 100-Gbit/s photoreceiver today. (See u2t Samples 100G Receiver.)
Both companies were part of the startup wave that was poised for the surge of 40-Gbit/s demand that was going to arrive... well, sometime soon after 1999. Obviously, it didn't work out that way.
Inphi started as a chip company specializing in indium phosphide (InP), a material better suited than silicon for high-speed chips. (See Inphi Pitches OC768.) Founder and vice president Loi Nguyen says the company even had Broadcom Corp. (Nasdaq: BRCM) lined up as a partner to build multiplexer/demultiplexer chips that would work with Inphi's; the two-level mux/demux combination would distill 40-Gbit/s signals down to a speed that electronics could easily handle.
But as the market crashed and burned, Broadcom dropped out of that partnership. Inphi leaned on the 10-Gbit/s generation and spread out to different markets, such as selling chips to memory module vendors. It also started working on regular silicon chips, uncertain InP was going to pay the bills. (See Inphi Moves Beyond InP.)
Along the way, Inphi went through its share of money. The company raised $12 million in 2000 and another $24 million in 2002. In 2004, it picked up another $18 million. And some investors held on for all or nearly all of the ride: Mayfield , Tallwood Venture Capital , and Walden International Investment Group . (See Inphi Wins $24M and Inphi Lands $18M.)
The thaw began in 2006, when AT&T Inc. (NYSE: T) began building a 40-Gbit/s backbone. (See AT&T Readies 40-Gig Backbone.)
"We took our drivers and TIAs out of the icebox and started selling them, and since then, the 40-Gbit/s market has been growing quite well," Nguyen says. He claims Inphi has grown at 30 percent per year for the past five years.
U2t Photonics deserves a nod as a survivor, too. The company set out during the bubble to build 40-Gbit/s receivers, and back in 2003, it was already using Inphi chips. In the years since then, u2t has managed to keep afloat with the help of some research work. (See U2t, Inphi Team on Receivers and ECOC Roll Call.)
What's nice for Inphi is that most of the competition has dropped out. In 1999, Applied Micro Circuits Corp. (Nasdaq: AMCC), PMC-Sierra Inc. (Nasdaq: PMCS), and Vitesse Semiconductor Corp. (Nasdaq: VTSS) all would have been big-deal competitors, but all three of those companies have scaled down since then.
The thinner competition includes GigOptix Inc. (OTC: GGOX) , a reborn startup that makes laser drivers and modulator drivers, and eGTran Corp. , which offers high-speed chips but also uses them in its own 40-Gbit/s transponder modules.
Inphi has already been talking about 100 Gbit/s to some extent. Its 28-Gbit/s reference design, for downshifting 100-Gbit/s signals to 10-Gbit/s streams usable in test equipment, got demonstrated at OFC/NFOEC last year. (See OFC Goes to 100.)
A coherent receiver could be a key element to the development of 100-Gbit/s transport, though. Coherent optics, used with dual polarization modulation schemes, allow for transmission of four bits per optical symbol. In other words, to send 128-Gbit/s signals -- enough to fit 100 Gbit/s with Forward Error Correction (FEC) -- the modulator only has to "go" at 32 Gbit/s.
Inphi also announced its second-generation 40-Gbit/s TIA today. The goal in designing it was to reduce cost, something that wasn't a priority 10 years ago when every 40-Gbit/s effort was proprietary and therefore expensive. "One of the mistakes the industry made in 40 Gbit/s during the dark days of the market collapse was that everybody was working independently, so there was no standardized solution," Nguyen says.
— Craig Matsumoto, West Coast Editor, Light Reading