DSL Prime: The VDSL Standards War Drags On

DSLAM prices are now so low that even industry holdout SBC may invest in new equipment. Meanwhile, the VDSL standards storm continues with a new development: lawsuits.

by Dave Burstein DSL Prime [July 3, 2003]

EFM goes for DMT
"VDSL Olympics" test results crucial
DMT will be the primary VDSL line code for the U.S., as the Ethernet in the First Mile committee (previously partial to QAM) joined T1E1.4 in supporting the standard. T1E1.4 in parallel will propose a Technical Requirements (TRQ) for QAM, and maintain both going forward.

One comment that supported the T1E1.4 resolution that created two documents suggested "One choice got the standards track gold medal. The other got the TRQ, which was described as roughly equivalent to a DSL Forum specification." To me, DSL Forum specifications are pretty significant, as the Forum leads the industry internationally. The dual recommendations (and the careful language) might be read as a compromise supporting both in different ways, but the T1 leadership and the subsequent vote in EFM made the sentiment clear. Even more crucial is the support of the bells, who have not been buying VDSL but may offer it as part of the fiber build. All the big players (Lucent, Nortel, Siemens, Cisco, Motorola) are likely to bid for the Bell fiber deal, so are likely to develop or bundle a DMT board to include in their bid, even if it doesn't go beyond a lab demo unit.

Did the U.S. listen to China, Japan, Germany, Korea, Britain and France?
A remarkable request came from British Telecom, Korea Telecom, NTT (Japan), China Telecom, China Netcom, Deutsche Telekom, France Telecom, and others to allow a free choice between QAM and DMT. My initial opinion was that "T1E1.4 intends that both documents would be maintained going forward" was a respectful acknowledgement of that position, but key members of the committee tell me international opinion had little impact. That amazes me, because international co-operation is more crucial than ever.

Until recently nearly everyone looked to the U.S. for telecom innovation, including DSL, but those days are past. Lucent has decimated Bell Labs, killed the Lambda router, and drastically cut R & D. Nortel is a shadow of what it was. Telcordia is scrambling to survive. The Bells have cut research much more than reported, rarely investing in work that only pays off over time. The President of SBC is personally embarrassed by how they've lost the world leadership to Korea Telecom; it was superfluous for me to point out that France Telecom, Telecom Italia, Yahoo BB, and NTT had all done better in DSL than his company.

Kevin Meyer and others in Electronics News have a strong series on how the world's best engineers are staying in their home countries, no longer moving to the U.S.. TI's most advanced DSL chips are designed in India; China's developing their own video compression and CDMA standards to avoid ridiculous royalties. In this VDSL battle, two key designs came from Israel, one from Sweden and France, and the fourth from Indian born engineers in California.

The ITU, perhaps in two weeks, will work towards an international standard. If the major telcos above don't change their opinions, the logical ITU conclusion is to allow both line codes. The politics are very complicated.

Now, choosing the best for your deployment
No one wants to be dependent on a single source, Ikanos. Their engineering is obviously exceptional (they beat Alcatel/ST to market), but it's a small company whose board members have a history of selling the firms they invest in. The price they are looking for after this decision is presumably very rich, and it's not obvious who will pay that much for a company without earnings.

But 2.5 million users have proven QAM works well and inexpensively, and many deployments are from remotes and basements where QAM is the obvious current choice for price and field-proven reliability.

What the Olympics prove
Two DMT chips did better than two QAM chips in the T1E1.4 sponsored "VDSL Olympics", lab testing at BT and Telcordia released June 17 and available to all on the T1 website. None of the four chips, from Infineon, ST, Ikanos, and Metalink, delivered as much as hoped. DMT chips proved beyond a doubt that they can deliver comparable performance. It's puffery, however, to say this very limited test proved DMT is better. No engineer would accept claims based on only four datapoints loaded with confounding variables. Behrooz Rezvani's team at Ikanos proved their enormous skill by delivering DMT chips far before mighty Alcatel/ST; it's certainty plausible the results came from a better design team, not a better line code. The DMT chips are newer designs, and Francois Crepin of Metalink told EE Times their nextgen chips (with Reed Solomon coding) will more than catch up later this year. Some very smart people tuned the systems for the lab test; a suggestion that C.'s optimizing for the test made a major difference should be taken as a compliment to his ability, but not implausible. It's also always dangerous to extrapolate from the lab to the field, even careful work from the best labs in the world. The dramatic example in our field is that early ADSL chips had the range for far fewer actual customers than the lab claims.

I often wish I had independent results to check manufacturer's claims, and hope to see them more often. To encourage that, the next three chips that provide me tested results will get a feature article, even if they are otherwise uninteresting. I hope other reporters—and purchasers—also ask for results. Otherwise, I'm in a bind when I get a press release with extravagant claims. One current release claims improbably more reach. I've asked the company for test results that back up the claim, and will report it when confirmed.

Pricing of VDSL
$100 for DSLAM+ modem is a common Asian price for QAM, as Infineon is selling 10 Mbps symmetric VDSL chips for $10. The 50 Mbps chips are somewhat more expensive, but everything is falling rapidly. Metalink and Infineon know their customers need high volume pricing, while Ikanos initially asked more for the chip than some charge for an ADSL modem.

There is no "market price" for DMT gear, because so little has shipped. I believe Ikanos has significantly dropped their price in some bids, and I'd be interested in hearing whether their current effective monopoly is letting them charge more. ST's chips aren't yet deployed in systems visible in the field. Broadcom's likely to make an announcement soon but hasn't given me a date for shipping chips. TI two years ago was promising imminent delivery of current spec VDSL chips, but they pulled the engineering team off VDSL. They cannot give me a date for chip availability to justify their rhetoric. Currently, there's effectively one source for DMT chips if you are deploying in 2003, and no one except ST firm for 2004. So if you want DMT, you'll need to bargain hard.

Neither the U.S. nor Europe has seen significant volume, so prices have been much higher. That international discrepancy breaks down rapidly these days, however, and at least two companies have suggested to me they'll offer Asian prices in other markets. Key Bell supplier Alcatel doesn't even have a VDSL card for their U.S. DSLAMs yet, much less price details.

Neither fear nor favor
"I hope never to hear the words 'line code' after this week" J. told me, and it is boring so many of you I almost killed the piece last issue. Some companies involved currently hate me for not seeing things completely their way, but that usually passes. But I couldn't get an answer on royalties, a key question that exploded with the Globespan TI court case. I couldn't kill the story after that, and got a dramatic letter telling me it was the right move "Private—please do not quote me (you could get me fired). I and many others are anxious to see fair play and I am very reassured that you are covering the proceedings."

One e-mail began was "I know your 'opinion' on VDSL (DMT vs QAM)." I was surprised, because I have actually not been sure of my own opinion. Anyone who thinks I didn't report the positive aspects of DMT needs a course in remedial reading.

Copyright 2003 Dave Burstein.
The DSL Prime Newsletter is reprinted with permission.

"The power of the printing press belongs solely to those who own the presses"
—A.J. Leibling

The Internet is the cheapest printing press ever invented.