Chip startups are drawing their battle lines for next-generation passive optical networks (PON), as Passavé Technologies Inc. has announced its entry into GPONs, while Teknovus Inc. says it's onto something better with a clandestine effort to upgrade Ethernet PONs.

Those startups, along with BroadLight Inc., sell PON controller chips, which sit inside both the optical line terminal (OLT) at the central office and the optical network terminal (ONT) in the field.

What's different is the choice of technologies for each of the three. Broadlight has stuck with the ATM-based PONs, developing chips for BPON and recently announcing chips for GPON, the gigabit-speed successor to BPON (see BroadLight Intros GPON Chips). Passavé and Teknovus got their start in the EPON camp.

For the next generation of chips, Passavé is joining Broadlight with a GPON line announced this week (see Passave Unveils GPON Chips). Passavé will produce chips for the OLT and ONT, with prototypes appearing this year and commercial availability set for the first half of 2006. "We don't expect to see volume deployments of GPON until 2007," says Dror Sal'ee, Passavé's vice president of marketing.

That could put the two chipmakers at odds if the U.S. RBOCs do adopt GPONs, as some have theorized (see RBOCs Cast Wide GPON Net).

Teknovus won't join the GPON game, though. Sticking to its Ethernet plans, the company is putting its next-generation efforts into a new development for EPON -- one that will be "ahead of GPON by one year," says Rex Naden, the company's CEO.

"There's a lot of work going on that hasn't been announced. There are a lot of people working in this area," Naden says. He won't divulge who, how many, or how far along they are, although he says the plan includes "the largest carriers in the world and the largest OEMs in the world" and "will be announced soon."

The key would be to offer backward compatibility with present EPON deployments -- something GPON can't provide -- and to keep up with the demand for video traffic. "They're not going to make money with voice and they're not going to make money with data. It's all about IPTV," Naden says.

For now, no Institute of Electrical and Electronics Engineers Inc. (IEEE) groups have formally begun work on a new EPON standard.

It's rumored that Teknovus is working on a faster speed grade for EPON, matching GPON's 2.5-Gbit/s symmetric delivery versus the present 1.25 Gbit/s for EPON. But given Ethernet's penchant for advancing in powers of 10, Teknovus might be looking well beyond a double-speed EPON.

"Definitely the IEEE standard will come up with higher rates. The question is whether they go with a double-rate option or go to 10 Gbit/s," Sal'ee says.

There shouldn't be many impediments to a 10-Gbit/s Ethernet PON, "particularly now that 10-Gbit/s Ethernet demand in the data center and other places has jammed down the price of the optics, particularly at short range, to reasonable levels," says John Freeman, an analyst with Precursor.

Of course, a "reasonable" optics price still exceeds $1,000, but that should continue to drop. While Freeman isn't aware of any higher-speed EPON developments, he figures the next step for the technology should be to 10 Gbit/s.

"You might as well leapfrog it, because by the time this stuff scales in 2007, 10-Gbit/s [Ethernet] might be deployable," Freeman says.

As for why Teknovus is sticking with EPONs, Naden cites the Asian market. "Eighty percent of the lines in the world by the year 2010 are going to be from companies that are currently planning to use EPON," Naden says. That includes carriers in Korea, Japan, China, India, and Australia, he says.

The prospects in EPON were enough to garner PON vendor Alloptic Inc. another $30 million in funding, announced just yesterday (see Alloptic Allotted $30M More).

Japan is a primary EPON target considering the success of the technology there. Highly competitive CLECs such as KDDI Corp. and Softbank have gone with EPON, and incumbent {dirlink 5|135} began deploying EPON this year (see PON & FTTx Update).

"Even if there's a bias towards ATM at NTT, the most successful fiber-to-the-home guys are going to be the CLECs, and EPON is fundamentally cheaper," Freeman says. The ATM-based PONs use three wavelengths versus two for EPON, and that extra wavelength adds to the cost. "The price of that extra color is $30 to $40 per transceiver," Freeman says.

— Craig Matsumoto, Senior Editor, Light Reading