Prospects weighed for 40-Gbit optical networks
By Craig Matsumoto
EE Times
(10/04/01, 12:39 p.m. EST)
SAN JOSE, Calif. — The challenges of OC-768 (40-Gbit/second) optical networking run beyond the physics of light into areas such as power and cost, panelists at the Communications Design Conference said Tuesday (Oct. 2).
"There's been an alarming lack of power considerations at lower speeds before," Perna said. In particular, carriers wanted OC-192 designs to emphasize the integration of more features in silicon, but Perna said that "power has to be considered as a much more important variable as you go from 10 Gbits/s to 40 Gbits/s."
An exact power budget is difficult to derive on paper, however. The permissible power dissipation of a line card varies depending on whom you ask, he said. Perna gave 8 watts per chip as a rule-of-thumb figure used at PMC-Sierra. Paul Liesenberg, vice president of chip maker ZettaCom Inc. (San Jose, Calif.), mentioned the same figure.
Network Equipment Building Standards compliance sets a limit of roughly 1,950 W for power dissipation in an entire seven-foot rack, said Kris Shankar, vice president of marketing for systems OEM Metro-Optix Inc. (Santa Clara, Calif.). That puts an upper bound on power, but the budget for particular systems and line cards will vary throughout the industry, he said.
.
.
.
The problem is exacerbated by carriers' insistence on preserving the size and shape of the chassis for equipment, Perna said. Likewise, the line cards holding the chips aren't growing in size. "What you cannot do is populate your card with four times as many chips as you had for OC-192," said Liesenberg of PMC-Sierra.
In fact, carriers want network processors and other chips to become even smaller as they move to 40 Gbits/s, Liesenberg said. For some devices, such as network processors, he said, standard-cell blocks will be too large to reuse, and custom designs will rule the day. "That's a scary world where chip vendors like ourselves bet the farm on a new architecture," Liesenberg said.
Panelists also noted that carriers want 40-Gbit/s systems to honor old interfaces. "There's still a lot of vestigial DS1 and DS3 [lines] out there that you have to address," said Shankar of Metro-Optix.
Most of the attention on 40-Gbit/s challenges has focused on the physics of such high speeds, where effects such as chromatic dispersion become problematic. Chromatic dispersion exists in all fiber transmissions, but isn't a serious problem yet; an OC-48 (2.5-Gbit/s) signal can travel 1,280 kilometers before the dispersion renders the signal unusable. For OC-768, that distance is 5 km, said Mark Barratt, vice president of business development for LaserComm Inc. (Plano, Texas).
For reaches beyond 360 km, OC-768 signals will need broadband dispersion compensation at every amplifier as well as tunable narrowband compensation at the receiver, he said.
Most 40-Gbit/s deployment is some years off, delayed in part by the economic problems among all carriers. But trial installations will be emerging during the next year. Barratt said he knew of three trials gearing up, and he predicted carriers would see OC-768 revenues by the end of 2002. Shankar was less optimistic, predicting revenues would come by the end of 2003.
eet.com |
