G.fast Could Use a Boost
Amendment 2 will usher in techniques such as higher bit loading and increasing the transit power to bring about further improvements. But the real prize is expected to come with Amendment 3 in 2017. Besides doubling the frequency range, from 106MHz to 212MHz, this iteration of the technology will also bolster vectoring, which cuts out crosstalk interference, and support the introduction of distribution point units featuring many more "ports" than today's models.
"At the moment we are stuck at 16-port vectoring," says Joyce. "That is fine for distribution points but with cabinets it's not really enough -- we want to be going into 48- and 96-port vectoring solutions for cabinet deployments."
While vendors seem confident they can support Amendment 2 using existing first-generation silicon, Amendment 3 will definitely require a new set of second-generation G.fast chips from Broadcom Corp. (Nasdaq: BRCM) and Sckipio Technologies , the market's two leading G.fast chip vendors. These second-generation chips seem likely to become available next year, but it remains unclear exactly when. "They are right around the corner in telco time," says Michael Weissman, Sckipio's vice president of marketing. "That doesn't mean next week but it doesn't mean in 2018, either."
BT will obviously be very keen to get delivery of technology based on those chips to pursue a more widespread cabinet-based deployment of G.fast technology. In the meantime, its current G.fast rollout plans, which involves connecting up to 140,000 premises by the end of March 2017, will rely on enhanced first-generation chips that support Amendment 2 of the G.fast standard, with new higher-power line drivers. For this stage of its access network upgrade, BT recently announced supplier deals with Huawei Technologies Co. Ltd. and Nokia Corp. (NYSE: NOK) for G.fast equipment. (See Huawei, Nokia Land Initial G.fast Deals at BT's Openreach.)
Broadband equipment vendors may also be keen to weigh up the differences between Broadcom and Sckipio once second-generation chips become available. Sckipio has long been regarded as the real G.fast pioneer, but there is a perception that Broadcom has been closing that gap during the past year.
Currently, the headline difference between the two silicon vendors is that Sckipio does not support fallback to VDSL, an older copper-based broadband technology, while Broadcom does. Sckipio insists this is not an issue, and that it supports "co-existence" with VDSL, but it also acknowledges that some operators have seen the lack of VDSL fallback as a potential problem. Nevertheless, it denies this has driven BT towards Huawei and Nokia, which are using Broadcom chips, and away from Adtran, which has worked closely with Sckipio during BT trials. (Both Adtran and Sckipio are still involved in those trials, it should be noted.)
With second-generation silicon, Broadcom has a dual-chip strategy, according to Adtran's Joyce, whereby a second chip handles vectoring needs when more than 24 ports are in use. Sckipio, by contrast, does not rely on this dedicated "vectoring engine" but can instead link chips together to provide so-called "distributed vectoring." There is a possibility that Broadcom's dual-chip approach drives up space requirements and costs. Joyce reckons this is unlikely because larger "pizza box" units would, in any case, be used to support 96-port deployments. But until physical chips materialize, Adtran cannot verify any of the differences between Broadcom and Sckipio.
Next page: Regulatory and competitive heat