AT&T leading the efforts for an updated standard. Field tests of current G.fast gear have seen speeds of 400-700 megabits downstream, not the gigabit promised in early pr. 400-700 is plenty for almost everyone but the marketing department is worried. Cable is going to one and two gig (shared.) Google is just one of hundreds around the world offering a symmetric gig. The engineers have marching orders: get our G.fast speeds to a true gigabit ASAP!
Fortunately, the engineers are confident they can reach a gigabit.The ITU group already is working on a second appendix to the G.fast standard, aiming for 2016. A chip CTO tells me it will be easy to do more efficient coding (non-linear precoding) and transmit more bits for a given frequency (constellation.) Those two changes he believes enough to do the trick.
Trevor Linney of BT is running the most advanced tests of G.fast, connecting 4,000 homes. He reports "Our research shows that even more capacity can come from copper."
He also predicts more bits per tone and a lower noise floor (i.e. <-150dBm/Hz). He supports that with the test data in the illustration. Increasing the transmit power, perhaps to 8dBm, is also under consideration.
ITU-T agreed to specify non-linear pre-coding in an amendment to G.9701. Dong Wei of Huawei presented the data below at the excellent Upperside Conference. Interference is different at different frequencies a better coding scheme can make a large difference. Simulations show a difference of 100-300 megabits. Wei presented three different possible coding schemes, the three upper lines in the graph. Deciding on which is right for the standard can be challenging, one reason standards take time to develop.
Bonding two lines doubles the speed, also new to the standard.
ITU-T agreed to specify non-linear pre-coding in an amendment to G.9701.