News & Analysis
DMT backers claim victory in VDSL competition
Robert Keenan
6/18/2003 10:52 AM EDT
WAYNE, N.J. Backers of discrete multitone technology claimed victory in the VDSL market Tuesday (June 17) after a standards group released results of a much-hyped VDSL Olympics. But while the DMT camp smiles, quadrature amplitude modulation (QAM) backers are questioning the results, claiming that the results did not provide a real-world view of the technologies.
Sponsored by ANSI's T1/E1 committee, the VDSL Olympics pitted two chip set developers on the DMT front, Ikanos and STmicroelectronics, against two QAM chip vendors, Infineon and Metalink. All four agreed to have two independent test labs, BTexact and Telcordia, perform tests to check power spectral density, reach, data rate performance and other metrics. After weeks of waiting and hype, the DMT camp could be the big winner
"All the tests are done and DMT was proven to provide the highest performance," said Richard Sekar, vice president of marketing at Ikanos. "Overall, in every aspect of the tests, DMT did well."
"The general point regarding the results was good," said Jean Beka, deputy general manager of ST's Access Division. "We are really satisfied with the outcome."
One of the reasons that DMT backers are grinning and QAM backers are quiet is the power spectral density (PSD) tests performed during the second testing set handled by Telcordia. In results posted online, chip sets from ST and Ikanos received a "yes/yes" rating in both the uplink and downlink direction during PSD tests run at various data rate configurations.
The QAM-based designs, however, struggled during the PSD tests, with Metalink failing to get a yes/yes rating on three tests and Infineon failing to receive the rating on 11 tests.
According to the online report from Telcordia, Metalink's problems were primarily the result of unexpected energy in the downstream transmitter. In Infineon's case, the VDSL chip set fell below the lower band notch in the upstream and downstream paths on several instances.
According to Telcordia's results, PSD tests on Infineon's chip set were also "consistently above the M1-notched template +1 dBm/Hz maximum level" in several instances.
The 16 Mbit/s downstream/1 Mbit upstream and 22 Mbit/s downstream/2 Mbit/s upstream were problem bands for both Metalink and Infineon. Both companies had PSD problems in these bands. "These are important configurations for today's carriers," Sekar said.
Metalink President Francois Crepin did acknowledge that Metalink's chip set encountered some out-of-band problems. "This was due to an anti-aliasing filter not working properly. This has nothing to do with QAM or the performance of our device," Crepin said.
Reach tests also appeared to favor DMT. In 16-Mbit down/1-Mbit up configurations, the DMT chip sets reached peak rates of 3,450 feet while the two QAM-based chips delivered a peak reach of 3,000 feet. At 22-Mbit down/3-Mbit up, the DMT chips hit a peak reach of 2,850 feet while the QAM chips delivered a peak reach of 2,550 feet.
Metalink and Infineon disputed claims that QAM struggled in testing and took issue with the test format. "Our opinion is that the tests focused only on low-bit-rate services and that these services are not applicable to the Asia Pacific market, which is driving VDSL deployments," said Mark Tyndall, vice president of business development in Infineon's Wireline Communication Division. Tyndall said valid testing should have been included configurations with 50 Mbit/s downlink speeds and higher.
"Some of the results that are being quoted are based on optional tests," Metalink's Crepin said. "Some of these tests are influenced by Reed-Solomon coding schemes that are not standardized and are already excluded by the Ethernet in the First Mile standards body."
If Metalink implemented the non-standard Reed-Solomon coding scheme as well as a constellation providing 1-bit/Hz, known as QAM2, Crepin said the QAM chips sets could achieve up to a 600-foot increase in reach. The new constellation can be added through the addition of firmware to the chip set. Metalink will introduced a version of its chip set in the third quarter that supports the Reed-Solomon coding scheme in question, Crepin added.
Infineon also said implementing QAM2 and the higher-efficiency. Reed-Solomon coding technique could compensate for the differences in the tests. Tyndall said Infineon can support these features through firmware upgrades.
Metalink's Crepin also said the tests performed included prototype chips from the DMT camp while the QAM manufacturers provided deployed chips. Hence, DMT performance could vary once those chips move from the prototype stage to production.
"This is not a real-life assessment of the technologies," Crepin said. An accurate assessment should include a comparison of transceiver chip sets to evaluate power consumption, footprint and silicon die size.
Industry observers questioned whether the DMT chips were prototypes. While ST's chip has yet to hit the market, the Ikanos chip set is available, Sekar said. Ikanos said it has shipped 500,000 VDSL ports based on chips used in the Olympics, Sekar added.



