In 2016, G.fast looked very promising.
But only BT & Australia's nbn remain
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: 14 October 2016 14 October 2016
Until now, only coax allowed individual speeds to be modified. G.fast allows changing the downstream/upstream ratio from 90/10 to 10/90. While most of the time higher downstreams are better, Sckipio has demonstrated 750 up. Before DTA, you could only deliver at a fixed ratio to all the customers connected to a DPU; the first DTA systems, from Adtran and Calix, only worked on low-interference lines, either coax or an isolated pair of phone lines. In Booth 20D at BBWF, Sckipio and Calix will have a live demo of Collective dynamic time assignment, their name for DTA over twisted pair in a typical binder. I have unofficially seen some test results; they look good.
When I first saw DTA, I immediately knew why I wanted it. I need faster upstream because we're backing up terabytes of Jennie's video footage to Amazon's unlimited cloud. At $55/year, that's a great deal. I've spent two months so far at Time Warner's top speed of 20 megabits, 200 down. I'd much rather have 75/75, but FiOS doesn't come to my building. If we had 500 megabits from fiber to my basement, I could have 100 down and 400 up. If my neighbor installed 12 virtual reality headsets for gaming friends, she might prefer 400 down, 100 up. (Wish I had the speed today.)
Rami Verbin of Sckipio told me a while back he was working on this, but I didn't expect it to be working so soon. It requires brilliant mathematics for signal processing and state of the art electronics.
- Published: 14 October 2016 14 October 2016
Supporting 30A should win Japanese & Korean market. The biggest problem for most telcos deploying DSL is that until now, no one made a DPU with more than 16 ports. HFR of Korea has the first 24 port DPU and a giant customer ready to go. $15B SK Telecom is Korea's largest wireless carrier and #2 in fixed. There are 9M apartments in Korea served with VDSL and copper LAN, most running at 100 megabits.
Korea Telecom is actively deploying GIGA Wire, their version of G.hn, for higher speeds. They intend to upgrade 95% of those units by the end of 2017. SK needs to respond quickly. Japan is similar; millions of their "fiber" lines are fiber to the basement + VDSL. Korea and Japan used an advanced version of VDSL, using 30 MHz rather than the 17 MHz common elsewhere. Ikanos, now gone, had long been the preferred vendor. Sckipio is first with G.fast chips designed to work with VDSL 30a.
Curtis Frankenfeld of Century recently told me, "I would like G.fast to support a larger vectoring group than the current 16 ports."
- Published: 12 October 2016 12 October 2016
WTTR + G.fast for coax. In an interview with Sean Buckley, Small said “The other thing that we have separately talked about is we’re exploring the use of G.fast. That technology can work over coax or twisted pair so that’s an essential companion way to deliver service where a property does not have Cat 5 or Cat 6.” DirecTV has millions of lines with an antenna on the roof and either coax or ethernet cable throughout the building. Their first homes, in Minneapolis, are only offering 100 megabits. They want to use G.fast and take that to 500 megabits or more.
AT&T continues to be coy about whether they will turn their G.fast trials into volume deployment. They have vendor proposals on the table; G.fast is ready to go. As Century says, it works as promised. It's not very expensive, according to the quotes I'm hearing. Vendors are forward pricing to win the business. Century has made the leap to G.fast; 12 telcos are in early deployments, 60+ in trials. When I saw AT&T was doing WTTR (wireless to the roof), I assumed it would be G.fast. So far, it isn't.
Small carefully said, "we’re exploring the use of G.fast." (Emphasis added.)
- Published: 07 October 2016 07 October 2016
If fiber to the basement is profitable, why not wireless to the rooftop? I broke the story that a small N.Y. outfit, Skywire/Xchange, is doing G.fast over WTTR, Live from New York. It's G.fast with wireless backhaul. AT&T has been vocal they plan using G.fast, which is already in early deployments elsewhere and would work over both coax and existing telco twisted pair. I believe they are not using G.fast at this stage, although it will be a natural choice going forward.
AT&T's WTTR looks to be the same as Google's new Webpass division. Companies like Towerstream have been beaming mmWave to rooftops for commercial customers for more than a decade. Hundreds of wireless ISPs rely on mmWave backhaul as do most mobile towers around the world. The technology is old but few have done WTTR for consumers.
For $3,000, you can buy a pair of Ubiquiti radios to carry a gigabit. Siklu, which supplies Google/Webpass, and Ericsson have five gigabit units available with low latency.
- Published: 04 October 2016 04 October 2016
Longer reach, reverse power, downloaded upgrades for the customer equipment, DTA over coax, and a dozen other improvements. The updated G.fast standard was pretty much agreed last week at an ITU SG-15 meeting in Geneva. The major chip vendors, Broadcom and Sckipio, are already hard at work. The carriers are hoping for equipment in the second half of 2017.
I have hundreds of pages of proposals from some very good engineers so this is just the first pass. Amendments one and two were approved. Amendment three was consented but very few changes are likely before final approval.
BT needs longer reach. Their finance guys insist they use existing cabinets rather than building to the distribution points closer to the customer. G.fast was designed for 50-200 meters but existing cabinets are often 350 meters away or more. The new standard increases the maximum transmit power up to +8 dBm, with a practical goal of 300 megabits 300 meters. See BT G.fast Musts: ~ 350 meters, 48/96 ports. I didn't (yet) see anything about more ports, perhaps because that was possible without any updates. Broadcom had promised 96 port systems in Q4 2016 using external vectoring engines but the schedule has slipped.
AT&T has been vocal they want a true gigabit to compete with cable, not "up to a gigabit." Comcast offers gigabit downstream in parts of Atlanta, Nashville, and Chicago, with Detroit scheduled next.
- Published: 02 October 2016 02 October 2016
Fiber to the basement + G.fast is the obvious choice today to upgrade apartment buildings. Telecom Austria A1 did some of the first testing of G.fast in the field, an early proof of concept. In 2014, they reached 536 megabits with equipment from Alcatel and Sckipio. See Live G.fast at Telekom Austria. The A1 press release speaks only of "hundreds of megabits"; I expect most customers will see 500-800 megabits.
A dozen telcos have deployed 10's of thousands of G.fast lines. It works. Curtis Frankenfeld of Centurylink reports, "No disappointments on performance. The results in the field on real cable approach the lab results." Adtran alone has 65 customers in trials. I expect most of them will soon be announcing deployments to MDUs.
The eventual cost of G.fast should be little more than VDSL, based on the fill of materials. It's early days and competition is weak. I hear rumors of higher prices, which may be discouraging some.