At the same time, promoters of passive fiber, pitted against DSL in the market but sharing the DSL community's desire to make headway against cable MSOs, are promising low-cost support for gigabit passive optical networks.

While last week's Voice on the Net/Fast Net Futures conference here nominally centered on packetized-voice applications, conferees devoted more fervor to the topic of winning over customers to bundled "triple play" video, data and voice services. Some chip and system vendors are promoting broader use of very high-bit-rate DSL, although VDSL entails deployment of fiber deeper into the network. Others are touting a smorgasbord of ADSL upgrades-ADSL+ and ++, ADSL-2/-2+ and extended-reach DSL-to achieve the optimum mix of higher speed and longer reach.

John Cioffi, Stanford University professor of engineering and founder of ADSL pioneer Amati Communications Inc., fears the DSL permutations won't be enough. At the conference, Cioffi presented Stanford's work on dynamic spectral management, a means of adjusting DSL modulation on the fly to support speeds up to 100 Mbits/second. DSM is based on iterative water filling, an algorithm commonly used in communications signal processing.

When 50 lines at 100 Mbits/s are carried in a binder group, the resulting, 5-Gbit/s aggregate speed presents an appealing economic alternative to a 1-Gbit/s, Ethernet-based passive optical network (PON), Cioffi said. The method also can take advantage of recent FCC rulings on unbundled network elements, since "the physical layer need not be unbundled."

The spectral management can be combined with a topological shift to a vectored DSL access multiplexer (DSLAM), where signal processing is optimized for the particular mix of loop lengths coming from a central office or remote terminal, he said.

In early simulations of the system, shorter loops show a migration of modulation to higher frequencies, while the data rates on even longer loops improve. Cioffi said that the baseline case for promoting DSM economically had to follow the Hippocratic principle "First, do no harm"; that is, it had to be proven that DSM would bolster worst-case copper loops.

"You could mix 10-Mbit DSL with ADSL, or even provide ADSL at the central office and MDSL at the remote terminal, something that is not allowed today," he said.

Cioffi proposed a two-step process for using DSM today, allowing ADSL rates of up to 20 Mbits/s downstream or 10 Mbits/s symmetrically. Such an arrangement could be implemented totally in software, although hardware support would allow faster adaptive processing. DSM later could be combined with vectored DSLAMs to allow the co-generation of several DSL service types from central offices and remote terminals.

Timothy Waters, vice president of marketing and business development at Celite Systems Inc. (Austin, Texas), said Celite proposes a DSL aggregator at the local cross-connect to provide the equivalent of a cable headend fanning out from the central office.

The current Celite system already provides more flexibility than traditional remote terminals, Waters said, while the adaptivity Cioffi described could yield "a level of flexibility in DSL systems that carriers just can't get today."

With 5 Gbits/second of aggregate throughput per 50-line binder group, vectored DSL could take on Ethernet passive optical networks for last-mile access.

Last week, TeleChoice president Claudia Bacco hosted a conference panel examining the latest cost studies for PON and VDSL. Gary Lee, chief executive officer of FlexLight Networks Inc. (Kennesaw, Ga.), said the ideal interface for higher-speed PONs might be Sonet's Generic Framing Protocol (GFP) rather than Ethernet. To date, only FlexLight and startup Mangrove Networks Inc. have talked about using GFP over passive fiber.

In many cases, carriers can't justify deploying Sonet rings to remote customers, Lee said. But if passive point-to-point fiber links are deployed for up to 32 customers using coarse wave-division multiplexing, the economics become much more compelling. While a typical office park might require $3.78 million in initial hardware costs for Sonet rings or $4.88 million for point-to-point Sonet, a GFP-based PON protected ring with a CWDM overlay could be built out for $2.4 million, he said.

Many in the Ethernet PON community say the advantages of Gigabit PON and the Generic Framing Protocol can be duplicated in EPONs through changes in network topology. Wave7 Optics Inc. (Alpharetta, Ga.) offers residential and business EPON solutions that use a hardened Layer 3 Ethernet switch between the metropolitan active-fiber ring and the last-mile tap. "The economic model we looked at was not benchmarking against other PONs but comparing our fiber provisioning to the cable hybrid fiber/coax network and the active-fiber alternatives from telephony," said Tom Tighe, chief executive officer of Wave7.

Semiconductor vendors appear prepared to supply the emerging PON markets. John Egan, access marketing director at Infineon Technologies, said the expected price drops for EPON components look as good as those for more traditional fiber-to-the-curb and fiber-to-the-home active topologies.

Phil Grove, marketing director of voice-over-broadband products at Legerity Inc., said his company is working on derivatives of its broadband voice chips to target PON applications.

Lee of FlexLight is chairing the PON Forum, a coalition that is uniting the efforts of the Ethernet PON and Gigabit PON communities to displace ADSL and VDSL installations.