Dark Blue: Firm commitments from incumbent: BT (10M), Belgacom, Australian NBN, Swisscom, Austria, Bezeq Israel, Chunghwa Taiwan, Telus Canada, Telekom South Africa, SK Korea, (U.S.) AT&T, Century, Frontier, Windstream, Belgium, Omantel
Mid Blue: Smaller carriers in Germany, Norway, Finland, Japan
- Published: 15 August 2016 15 August 2016
More than a year late but now in acceptance trials at a large European carrier. 8/16 Everything in the article is accurate as far as I know, except the claim "First results" in the title, now removed. A reader just informed me that Fastweb in Italy is actually in production with 35b and there are some results on Italian websites 8/20 I thought I had "First results" but in fact Alcatel had published results from KPN testing I had missed. Below.
Original: From chipmaker Broadcom to Nokia and Huawei, everyone promised 35b "Supervectoring" for about a year ago.
Now it's here, and Kevin Schneider’s lab tests at Adtran found “mid to high 200’s @ 1,000 feet.” Talking with Kevin, I didn't sense any major anomalies in his testing at other distances. Until now, 35b speeds were estimated based on simulations like the chart here. These early test results are very encouraging. The hope is they can move to field trials in Q4 and soon after to deployment.
35b is essentially vectored VDSL 17b MHz extended to 35 MHz. Initially, the industry thought G.fast from close-in distribution points would be the standard upgrade. With volume production and some competition, the price of G.fast equipment will fall to very little more than VDSL. 35b is close to G.fast speeds from around 300 meters to around 450 meters, but much slower for shorter distances.
The savings won't come from a big difference in modem and DSLAM prices down the road. Moore's Law continues to bring chip prices down while performance increases. However, DT won't have to replace millions of 17b modems because 17b and 35b can co-exist in the same binder. Questions remain about vectoring efficiency, especially with older modems, so everyone is anxious to get field results.
Someone, perhaps at Alcatel, realized that DT wouldn't have to replace as many customer modems if they upgraded to a new system that could work with some of the older modems. That's not practical with G.fast, with a line code incompatible with the older modems. They decided to go with the modified VDSL, extended to 35 MHz. 35b sacrifices a great deal of speed but doesn't require replacing as many old modems. In addition, 35b is faster for those beyond the efficient reach of G.fast. DT signed on, especially after Broadcom promised chips would be ready in 2015.
In theory, the 35b chips should have been an easy design modification, I was told, because they were so similar to the existing chips. If any reader knows what held them up for a year, please let me know. Anonymity assured. 8/20 Update Alcatel's Paul Spruyt's presentation last October said they had 8 trials already underway and four customers, I just noticed. Ten month later, I would expect some of these to be in production if they hadn't found problems. <end update>
The U.S. and Europe mostly deploy VDSL using 17 MHz, not yet vectored. Many customers get less than 50 megabits downstream, although it's often advertised as "up to 100 megabits." With vectoring, the Belgacom data is that most short loops see doubled speeds, with most over 80 megabits and many over 100. Belgacom, under heavy pressure from cable, has invested in many cabinets close to customers. DT, AT&T, and Century/Qwest have fewer cabinets, longer distances, and ultimately lower speeds.
17b runs from 2 MHz to 17 MHz; 35b runs from 2 MHz to 35 MHz. The former covers 15 MHz, the latter 33 MHz. That's a 120% increase in spectrum used, so 250 megabits to about 300 meters is plausible. We now have a confirming datapoint. Thank you Gary Bolton and Kevin Schneider for what I believe is the first reported test speed. Adtran in the last few years has gone from reserved with the press to being one of the most helpful companies.