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: 10 February 2017 10 February 2017
35b was expected for 2016 but wasn't ready, the Germans say. The improved 35b vectored VDSL uses more spectrum to get speeds of 200 megabits or so up to 300-400 meters. That's not close to the 500-800 megabits G.fast can do over short loops but in some cases saves the telco a lot of money. G.fast can't share the wire binder with the old 17 MHz VDSL but 35b can. Deutsche Telekom has millions of the 17 MHz modems they don't want to replace so chose 35b. Unfortunately, the new systems aren't delivering what DT requires.
- Published: 10 February 2017 10 February 2017
200 megabits+, old modems still work. G.fast is headed over a gigabit on short loops 35b was introduced in 2014 as a simple, fast, low cost upgrade that could more than double speeds on existing VDSL lines. It uses 35 MHz of spectrum rather than the 17 MHz of most European VDSL builds but otherwise is nearly the same. Chipmakers - several now out of business - promised chips ready to deploy by 2016.
Simulations suggested 35b speeds could go up to 350 megabits over short distances. Adtran shared with me test results showing 200-250 megabits+. Alcatel/Nokia had similar results. Both were very confident a year ago. Until DT commented it would be 2018, that seemed reasonable to me.
- Published: 31 January 2017 31 January 2017
It's really cheap to just hang a box on an existing remote terminal. "We're rolling out those side pods on the cabinets a quite a place now we have the production hardware, we have the production firmware and we're building a footprint quite quickly now," said Clive Seeley on the quarterly call. "The indications are that the performance of the product over the new equipment is pretty much spot on what we had predicted from the labs in the earlier few trials and I am very pleased with that. So I'm very pleased with where we stand right now on G.fast."
They are on target for 150,000 or so homes passed in the next couple of months and a fast ramp towards the 10M they've promised for 2020. BT is running G.halffast, with a target of ?300 down, ?30 up at three hundred meters. It would be impressive if they hit that; the results I've seen elsewhere are 10%-20% lower. The results at 100 meters - the plan before the beancounters jumped in - will probably be 500-800 megabits split between upstream and downstream. The early trials of G.fast have shown 5%-15% of the lines do not reach the expected speeds. BT and everyone else in the industry are scrambling to find and fix the problems.
The more interesting question is whether Huawei and Nokia have delivered the external vectoring boxes expected about six months ago.
- Published: 23 January 2017 23 January 2017
5%-10% disparity between ECI & Huawei. Tests in December show Huawei VDSL DSLAMs in Britain have a mean download speed of 32.6 Mbps. ECI DSLAMs measure 30.5 Mbps. The chart below, by Andrew Ferguson of ThinkBroadband, has a series of readings over time that are fairly consistent. Ferguson notes, "Huawei has been slightly ahead since July 2016 after years of being neck and neck." He goes on to speculate the difference may be due to the implementation of G.INP, a very fast error correction method. Broadcom introduced G.INP years ago with some fanfare, but the results were very modest. I had heard things had improved, so this data doesn't surprise me. Andrew writes me he will also be watching the lower target noise margins BT is testing. They now use 6dB but will be testing 3, 4, 5dB. These are log scales, so even a few dB makes a difference. Lab tests suggest a significant speed increase.
- Published: 19 January 2017 19 January 2017
Slow in the first half, probably ramping in the second half. AT&T is being officially coy, but a senior source confirmed they are definitely moving ahead on G.fast. Tom Starr played an important role in the standard and they were one of the first in trials. Their top executives have been enthusiastic several times. They have been making quiet moves in D.C. to get out of their commitment to 12.5 million fiber homes, presumably using G.fast instead.
G.fast is logical for them. They've built an enormous amount of fiber the last few years, going first to businesses. In many places, fiber can be extended inexpensively to nearby buildings. They are also expanding their trial of WTTR - Wireless to the rooftop - beyond Minneapolis. Wall Street is riding them for the lines lost to cable, now going to a gigabit in most of the U.S.
They have millions of lines of coax connected to rooftop antennas from the DirecTV buy.
- Published: 15 January 2017 15 January 2017
The Chinese chipmaker has been almost invisible for two years, at least in the West. They are a substantial company, now ten years old. Their CEO is a U.C.L.A. Ph.D who was successful in American chip companies until he founded Triductor in 2005 in California. The funding and market developed in China and they moved. They have a substantial business although I've seen no figures. Huawei, by far the largest telecom supplier, has a close relationship. I covered Triductor a while back, when they were considering extending their VDSL line to G.fast. Good to see another supplier confident of G.fast demand.
Lincoln Lavoie's University of New Hampshire Interoperability Lab is crucial to the future of G.fast and is strongly supported by the Broadband Forum. Broadcom and Sckipio are working hard to make their chips fully compatible because their customers want more than one choice. As so many vendors cut back, any smart buyer wants the protection of a second source.