DSL Prime: Equipment

As some equipment makers struggle, some product lines are discontinued, but innovation continues elsewhere, especially at manufacturing companies that do not depend on the North American market—some of which are based in the U.S.

by Dave Burstein DSL Prime [February 21, 2003]

R.I.P. Nokia Diamond Lane
George Hawley designed one of the earliest DSLAMs, and it's the core of Covad's network, which has proven more reliable than the Bells. Nokia believed they had a major deal with Chunghwa, until aggressive bidding from Samsung and ADI pulled it away. After losing out on several other major bids, they've decided to close their facility in Santa Rosa but continue to service the 2 million ports sold from Finland.

Ted Appel in the Press-Democrat calculated that 3,900 jobs, nearly a third of the once booming Sonoma telecom sector, are gone. He quoted George Hawley "This is not going to be the end of it. Times are tough. I think there is still a lot of shakeout here left."

XAVi: DSLAM + Modem at $100 per port
Video intelligence
John Kasha played key DSL roles at Ascend and Lucent, and now is representing Asian manufacturer XAVi, part of the billion dollar Chicony group (think keyboards). Even at low price points, they will offer a modem designed for 16 Mbps ADSL2 speeds and quality of service features required for voice and video. The DSLAM is targeted at basement deployments.

Not right for everyone, everywhere
Speaking yesterday to a telco, I suggested strongly they were probably better off sticking to their current suppliers, even though I knew reputable alternatives that would price at 30 percent to 60 percent less. Service, support, and reliability are worth the price, if that's what they're getting in return. Customers of AFC, for example, tell me the company has come through for them in many ways. Alcatel has done a better job backing up customers than some outfits I shouldn't name.

I also know that a while back British Telecom was paying a third more than comparable equipment was being sold for in some markets, because one of their people told me they had followed up on my article. That probably couldn't happen today; the operators in Europe and the U.S. have built informal links that compare things like this even if DSL Prime didn't report on the subject. Everyone in the West would like to meet their peers in Asia as well; I hope many come to FNF in San Jose, and I know the DSL Forum and the IEC are working to bring people together.

Lies by operators on these subjects are distorting policy decisions in the U.S.. Major telcos have sworn to the press and regulators falsely and repeatedly. One notorious case claimed millions of dollars in costs which I knew to be a tenth of that, and refused to either back up or withdraw the numbers after I asked the state PSC. A year later, they wonder why the state regulators doubt their other testimony. Qwest, SBC, and Verizon have called "economically impossible" deployments BellSouth and France Telecom have made profitably.

DSLAMs in telco quantities typically cost $70 to $110 per port. Occasionally hotly contested deals add installation, test units, Redbacks or other management systems, upgraded inter-office switches where capacity is insufficient, and the like. The cost per subscriber to serve 70 percent to 85 percent is between $150 and $300. Another 10 or 20 percent require relatively expensive upgrades of small remote terminals, repeaters, or other gear, raising the per subscriber cost for the last portion of "near-universal service" to $200 to $500, and even those prices are coming down rapidly. If a central office has decent fiber connections, as most do, DSL can be installed for $10,000 to $30,000, including costs like engineering design as well as the DSLAM.

Printed prices give readers a starting point for understanding the market, where most sales are large, individually negotiated deals with details hidden. A company the size of SBC will get a better deal, but a telco buying 100,000 units anywhere in the world should be getting attractive prices.

Every time I print a price, I immediately get e-mails saying that more than price goes into the decision. Of course that's true, and operating expenses are generally much higher than the cost of the equipment. In particular, Kasha was clear this is not a unit for a U.S. central office—no NEBS testing, etc., to drive up the price. Generally, when you actually negotiate a deal, loads of gotchas come up that aren't in the first sales presentation or my headline. Test equipment, required interfaces, management control software, training of your staff, and even shipping becomes an item of negotiation.

Millions at Efficient and Westell
Efficient's Q4, Westell's Illinois plant
Tom Willie of Efficient writes me "Mexico and South America were 15 percent of our overall volume of a million modems in a single quarter." Europe is now over a third of sales, as Efficient captures the continued growth spurt outside the U.S. market. One strength is they produce their own software, which lately has "included a ton of customer software development for specific carriers." Point-topic credits 12 companies with 70 percent of the world's subscribers; Efficient has maintained close ties with several key ones.

Westell's Illinois plant now has produced a million modems over several quarters, a remarkable success for a plant in the U.S. Westell worked with the Bells and BT in the 1990s, and won many early contracts by bidding below cost in the hope of building volume efficiencies. The end of the boom caused drastic cutbacks in 2000, but I'm glad to see they just reported a third consecutive profitable quarter.

Actelis: DSP bonding beats IMA standards
40 to 100 percent better performance for 5 to 10 Mbps service
Combine four 1 to 3 Mbps G.SHDSL lines, and you can deliver speeds well above a T-1 two miles or more. That proven performance is why the Ethernet in the First Mile Committee has approved bonded G. for long reach copper. Standards based bonding has limits however. If one of the lines has a performance problem, all lines drop speed. Hamid Lalani of tells me a customer has measured dramatically better results from Actelis' proprietary DSP based system after study. Besides maintaining maximum speed on each individual line, their error-correction dramatically reduces packet loss due to temporary problems, such as interference from a vacuum cleaner near the CPE or a dimmer switch.

Actelis is adding logic to standard G.SHDSL. Voyan and others are working on DSL chips with some of these function built in, theoretically even more effective. The goal for many is 10 meg in both directions, the same as old fashioned Ethernet.

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.