DSL Prime: AT&T Could Deliver More

The new U-verse system is capable of delivering close to VDSL's top speed of 100 Mbps, but AT&T is capping it at 6 Mbps.

by Dave Burstein of DSL Prime and Future of TV [August 25, 2006]

97 Mbps down measured at AT&T
U-verse user still can't access more than 6 Mbps, however
Chad Brantly in San Antonio syncs at 97 megabits on his U-verse gateway, close to VDSL top speed of 100 meg. This is a remarkable demonstration of what's possible, only four years after John Cioffi's groundbreaking presentation about 100 megabit DSL. Presumably, the DSLAM is 100 meters or less from his home, and few other users are in the binder group. Unfortunately, AT&T is refusing to sell more than 6 megabits to the customer.

Now that I, Light Reading, and DSL Reports have reported on Brantly's site, expect it to become very visible. Wall Street and reporters are all starving for information about AT&T, a company that hides what they are doing. Brantly, a Macintosh programmer, confirms an AT&T problem with screen flickers due to packet loss in the Alcatel equipment. Some of this is corrected by better installation practices, but some is proving tough to troubleshoot.

Tut Systems last year suggested the solution would be forward error correction, and designed a unit for that purpose. Researching this two months ago, I got a firm "FEC isn't needed" answer from a senior engineer, but it may be back on the table. Microsoft unfortunately believed the sales talk that packet loss, jitter, and latency problems were totally solved in new IP networks, and only now is scrambling to develop an FEC program if needed.

Brantly hears AT&T will launch single channel HD on October 15, presumably at the 8.5 to 9.25 megabit speeds I've previously reported. Anything less than that with current production gear is "fake HD", simply not up to quality if you have a good set. Phil Swann reports some of the satellite HD is lousy because of low bandwidth. The difference is minor on all but the best 10 percent of sets sold to date, so it's tempting to cheat.

AT&T and others will be very glad to hear that David Price of Harmonic is telling me the newest chips and improved encoding tools bring 6 Mbps live programming close. Price has credibility, because he was (accurately) saying 9 to 12 Mbps for live HD when his competitors were saying 6 Mbps is enough. Expect demos at IBC in September, and units that may replace multi-million dollar headends as soon as next year, including some newly built.

User reports trickling out make clear AT&T is in an early beta of U-verse, not a "deployment." There are constant changes to the operation while field personnel are just starting to gain the experience they need. Brantly's install took over four hours, which if not improved virtually doubles the cost of Lightspeed. Another expensive change rumored is expensive equipment at the local exchange. Lightspeed originally was designed to send virtually everything from nationwide "headends," with only local programming and ads in each city. I haven't been able to determine whether that's due to the widely reported scaling problems of the Microsoft IPTV software or packet loss at the Alcatel edge routers. Inside AT&T, everyone is pointing fingers at the other guy.

Despite the problems, I expect AT&T will honor their pledge to Wall Street to have a volume of equipment installed at yearend and move wider from December to May. They may even surprise Wall Street with customer counts. A marketing machine is ready to drive demand as soon as they can handle it. Presumably, everything will be in selected territories with ideal network layout and controlled numbers of users.

Moving the DSLAMs costs more than fiber all the way home, so is constantly under consideration at AT&T, where Chris Rice has discussed the upgrade. BT, Telfonica and others face similar choices.

The "Fiber to the node" plans are particularly cheap because they use existing field cabinets (cross-connects), but the network was designed for cross-connects at 5,000 or 6,000 feet, and only a handful of customers are close enough for higher speeds. Rice of AT&T is carefully keeping expectations down, telling reporters consistent speeds at 3,000 feet are 20 to 25 Mbps down. One of the biggest problems at AT&T is vendors promised to deliver Mbps over 5,000 feet, but haven't come close. The original budget called for very little bonding; now, to get the speeds for HD, AT&T will need to bond many lines.

Copyright 2006 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.

Reader Ryan Renneberg replies (from Canada): Just because a user can sync to the DSLAM at 97 Mbps doesn't mean there is 97 Mbps available on the link to the DSLAM. And that's just one user...can you imagine if multiple users were going for 20mbit+ packages? The fibre running to the DSLAM wouldn't have the available bandwidth to support it.

To which we reply: thanks! Note posted. I believe Dave Burstein understood this. He wrote the article; I wrote the deck.