In 2016, G.fast looked very promising.
Thousands worked at developing and deploying.
It wasn't enough.
Most carriers are investing
in fiber or 5G instead.
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: 21 February 2016 21 February 2016
XG-FAST, a potential future development of the technology, is in the early stages of lab testing, but has exceeded expectations in trials at Adastral Park, BT’s global research and development campus in Suffolk, and Alcatel-Lucent’s labs in Antwerp.
It delivered aggregate speeds of 5.6Gbps over 35 metres of BT cable, a record for full-duplex data transmission over a standard single BT line at this distance. The technology also performed well over longer distances, with aggregate speeds of 1.8Gbps over 100 metres; a significant result, as most UK homes are within this distance of their local distribution point, be that a pole or footway box. (below, full duplex excerpt from an article by Jochen Maes and team.)
At the start of the G.fast work, they investigated the possibility of sending data both upstream and down in the same frequency. It wasn't included then. There are trade-offs: complexity, reach, robustness.
A respected DSL engineer confirms to me DSL can also go full duplex. He emails:
ADSL1 had an echo-cancelled mode that was an option, but the crosstalk into other lines was a problem for full duplex. This was abandoned in DSL for many generations after that. With vectored systems, it is indeed now possible again to do a full-duplex G.fast or G.vector, but the power and cost would increase significantly. So, yes, G.fast could be 2x faster.
Usable speed doubling is unlikely given G.fast variable upstream and downstream.
Traffic might be 80% downstream and 20% upstream. Even if the full duplex works perfectly much of the available upstream is not of use to the consumer.
I just sent a note to chip engineers asking, "If a large customer wants full duplex, how long would it take and how much would it add to the cost?" I doubt anyone will go on the record but I welcome all thoughts.
Going full duplex is not new. 30+ years ago, full duplex technology and echo cancellation were used to create 9600 baud modems. Moore's Law allows faster processing today and some engineers are optimistic. (At least one I respect remains unconvinced.)
Stanford and Columbia teams are looking to bring FD to market for wireless. Kumu Networks is extremely hopeful that full duplex is ready to work in wireless networks, although fitting into a mobile phone will be a challenge. It's a spinoff from the labs of Stanford Professors Sachin Katti and Phil Levis. Harish Krishnaswamy at Columbia has developed chips.
AT&T and other carriers, facing one and two gigabit cable, are pressing hard for gigabit speeds to advertise. My guess is they'd spend $5/home more for the speed just for the promotional value, maybe more.. Century just confirmed AT&T's experience that promoting higher speeds has a "halo effect" on selling current services. So they announce "70 cities" for gigafiber even if they have passed only 2-3% of homes. That's also why nearly every telco seems to be announcing 5G "tests." I think Verizon will find something to call 5G and put some boxes in the field in the nest 18 months. It isn't really ready.
Of course, almost no one can point to any common application that even maxes out 200 meg, except heavy downloading.
From IEEE Communications Magazine — Communications Standards Supplement • December 2015
Full Duplex Transmission
In multi-user DSL deployments, simultaneous upstream and downstream communication on identical frequencies is made impossible by near-end crosstalk at the customer side of the network. When used as a single-subscriber technology, XG-FAST will have no NEXT between CPEs, hence allowing for simultaneous upstream and downstream on the same frequency, which we will refer to as full duplex, doubling the spectral efficiency compared to currently used time or frequency division duplexing schemes.
It requires an analog hybrid at the transceivers that attenuates the transmit signal to the level of or below that of the received signal, not to substantially increase the required dynamic range of the analog-to-digital converter. The residual echo signal is further removed digitally using vectoring techniques.
Note that compared to TDD, full duplex transmission also allows a reduced latency and framing overhead. E Communications Magazine — Communications Standards Supplement • December 2015