BT in the G.fast Lane

The history of telecom -- at least since the rise of the cable industry -- is a story of catch up: The cable industry juices up its DOCSIS standard and the telcos respond with a new DSL approach that squeezes more bits out of their legacy copper.

The endless race shows no sign of slowing. The next tool the telcos are preparing is a DSL variant called G.fast. Perhaps the most active telco in the ramp up to G.fast has been BT. Trevor Linney, its Research Leader for Access Networks, told Telco Transformation that the technology is performing well and outlined tests and trials moving forward.

Telco Transformation: What is G.fast?

Trevor Linney: G.fast is the latest generation of copper broadband technology. It is able to deliver ultra-fast speeds of 100 Megabits [per second] and above. It is very much an improvement and evolution of VDSL2. G.fast offers a number of very significant improvements compared to VDSL. The first is the obvious one: It goes from super-fast to ultra-fast speeds over copper. In BT's context, that is up to 330 Megabits per second downstream and 50 Megabits per second upstream for the initial trials and pilots. It does that by using a far bigger spectrum.

From the exchange, ADSL1 used 8 Megahertz. The second generation -- ADSL 2+ -- used 2.2 Megahertz and up to 20 Megahertz. From the cabinet, VDSL2 from OpenReach uses 17 Megahertz.

G.fast really hits the ball out of the park. It uses up to 106 Megahertz of spectrum. That is a massive increase. It can therefore deliver transformational [performance], which for us initially is 300 to 500 Megabits over the life of the potential deployment.

TT: What other tools does it use to achieve those results?

TL: It has taken incremental improvements of DSL and made them mandatory. A good example is vectoring, or cross talk cancellation. It was optional for VDSL but is mandatory for G.fast.

Vectoring maximizes performance by removing interference across customer lines. And you eliminate fading as more people take the service. You ensure delivery of a constant customer speed over time.

TT: How does it work?

TL: Signals leak. Vectoring uses some very smart math and digital signal processing to pre-compensate to ensure when the signals reach the home the affect has been removed.

I think it's true that with each generation we take all the learning from the previous versions and incorporate it. I think it's more than organic growth, though. G.fast is fundamentally made some quite different choices in the way the protocol works.

A good example: In VDSL and previous generations [of DSL] there is an upstream band then a downstream band and then an upstream band. They are relatively fixed. G.fast divvies things up by time: Downstream transmits for a while and upstream for a while. And you can configure that ratio, which makes it more flexible over time. As customer demand changes, you can tweak the ratio.

TT: Can you describe the tradeoffs with G.fast?

TL: It's true as with any copper-based technology there is a so called "rate versus reach" limitation on how fast a speed can be attained over a given distance. Because G.fast uses much higher frequencies it can't go as far.

One of big breakthroughs that BT had on its journey with G.fast is that it realized it is not just a technology that is able to go 100 meters or so. We realized with a few improvements it can further stretch rate versus reach. We challenged the industry to deliver ultra-fast broadband not just ten or 100 meters but over 300 meters or more. I think it's very much evolved first from its original conception to what is delivered today. We see chipset and system vendors deliver hundreds of megabits over 300 meters, which is really impressive. And it will have the ability to deliver 500 megs over time because there is headroom in the protocol. It is less limited than some people expected it to be.

I think it will certainly continue to evolve. The ITU-T is already studying how to use additional frequencies, and move from 106 to 212 Megahertz …They will look at novel coding. The key is that the ever-increasing amount of compute power available from chips today unlocks capacity. Things have been too complicated for earlier implementations become practical.

At the last Broadband Forum, Bell Labs showed a version of G.fast that offers 500 megahertz on a copper pair with full duplex operation, running upstream and down at the time. It produced an aggregate of 5.6 gigabits over 35 meters of copper, which is a great result. It shows you copper has a lot more to give over time.

TT: Tell us about BT field trials.

TL: Our initial large scale field trial of G.fast went live last summer to over 4,000 homes in the UK in two locations... The vendors were Nokia via their Alcatel-Lucent acquisition, AdTran and Huawei. Those trials were really positive and show G.fast operated within our expectations. We announced the next stage, which is to move to a pilot passing about 25,000 premises total in two locations. We are preparing our more industrialized solution, with a view -- with regulatory approval -- that will approach passing 10 million homes in the UK during 2020.

In terms of results of the trial OpenReach 330 megs downstream and 50 megs up. The majority of customer received headline rates. We also trialed in longer lines and they continued to see ultra-fast speeds.

TT: How will the new approach be ushered in?

TL: I think that you going to see VDSL2 and G.fast coexist in operators' networks for a long time to come. G.fast has been optimized to deliver ultra-fast over short to medium copper loops. VDSL is excellent to deliver superfast speeds over longer distances. In the UK, we in have VDSL over loops that are in excess of 1.5 kilometers. G.fast is not designed for that.

You will see deployments that offer both. In the UK they are configured so that they are spectrally compatible. VDSL [extends to] 17 Megahertz and G.fast starts at 19 Megahertz. So they can coexist.

TT: What is the biggest challenge?

TL: There is a challenge in density. You will find for a while that VDSL DSLAMs may be significantly bigger and have more ports than those for G.fast. That was one of the other big challenges that BT posed to the industry. Eight or 16 ports was not big enough for deployment at the cabinet locations where we want to start our rollout. We needed 96 ports or larger. They responded to this excellently. Huawei has already publicly announced their roadmap for delivery of a 96-port G.fast DSLAM.

— Carl Weinschenk, Contributing Writer