We covered the news that BT had G.FAST running in its research labs back in September, but we have now actually seen G.FAST running for ourselves in the labs over in Martlesham Heath, Suffolk. For those who are worrying already that BT is betting the world on copper, don’t panic as there is a lot of work going on with fibre too, including improvements in how the light carrying data is packed into fibre allowing BT to demonstrate 3 Tbps running over fibre alongside over live traffic.

Trackway used for testing distance and RF crosstalk on copper

Trackway used for testing distance and RF crosstalk on copper

BT has been at the forefront of the standards for ASDL, ADSL2+, VDSL2 for many years and sometimes the adherence to standards has meant we are not the first country to market, with some places rolling out the pre-standards and fixing issues with firmware upgrades. Not unlike the problems people have seen over the years when buying pre 802.11g/n wireless kit.

The debate over whether BT should abandon research into G.FAST and concentrate on a full fibre future is often a heated one and the debate could probably fill a conference centre for a week and there still be people on both sides who disagree. What we will say is that we believe, G.FAST adds another tool to the armoury and we are confident that as competitors show they can make FTTH work economically we will see BT deploy more of it in the next decade, G.FAST with its short reach of 50 to 150m likely may be used more, but the accountants may find that for certain distribution points going full fibre is ideal.

G.FAST weatherproof kit mounted on a short pole in lab.

G.FAST weatherproof kit mounted on a short pole in lab.

G.FAST really powered up and running

G.FAST really powered up and running

On the right you can see an actual G.FAST node, this unit can deliver four G.FAST lines and is a very early prototype and has been used to look into the real world performance using the extensive copper test facility that is outdoors and with its maze of track allows BT to reconfigure cable runs for distance and crosstalk performance. The rugged good looks are because the units are designed to be weatherproof and be mounted on telephone poles or in pavement chambers, though this unit is not fully sealed as shown by the plain yellow fibre entering at the bottom. A closer peek of the status lights reveal the unit really was running, and over the run of copper was delivering well over 500 Mbps downstream and around 200 Mbps upload. The balance between downstream and upstream frequencies can be adjusted , so if higher symmetry is needed it can be delivered.

I will mention Gigabit just once today, since while the summed downstream and upstream of G.FAST do reach Gigabit speeds, and has led to sum sections of the press calling G.FAST a gigabit service we won’t be doing that until it breaks 1 Gbps in any single direction..

The most interesting demonstration of the short visit to this copper lab, was showing a slightly newer remote node unit that was able to run VDSL2 and G.FAST to over four lines and a seamless upgrade from VDSL2 to G.FAST was demonstrated by simply unplugging the RJ11 lead from one modem and plugging it into another. A digital spectrum analyser was running and the waveform can be seen below, with the grey VDSL2 on the left and the G.FAST running all the way up to 106 MHz as the yellow trace, the sync period was pretty quick at 15 seconds too.

Comparison between the frequency utilisation of both standards.

Comparison between the frequency utilisation of VDSL2 (grey left) and G.FAST (yellow trace) standards.

In this case VDSL2 and G.FAST were co-existing in the same bundle, and this means that potentially Openreach could in the future once this dual-mode hardware is ready to deploy offer VDSL2 with the option of upgrading to faster G.FAST for those who feel the need for real speed. We should highlight that while Openreach is very soon about to start deploying VDSL2 nodes that will look similar as yet the G.FAST chipsets are not ready for massed production, so if you do see a grey metal box appearing on a pole in North Yorkshire it will be VDSL2 only for a couple of years. Pole mounted VDSL2 nodes might even allow for a speed boost anyway, since VDSL2 can go faster than the current 80 Mbps product speed, the lab kit was running at 110 Mbps a speed that would be achievable over the distance many properties are from their distribution point.

Is G.FAST the limit of copper? Well apparently not, work is underway to understand how doubling the spectrum used by G.FAST would behave, so a G.FAST2+ might be on the cards.

Grid work allowing cross talk and xDSL distance testing.

Grid work allowing cross talk and xDSL distance testing.

This means that the outdoor test grid is likely to stay at Martlesham Heath for sometime as the staff and the manufacturers work to stay ahead of the mean demand for broadband speeds.

The big unknown is what price will services that run at 300 Mbps be priced at, if the current pattern is followed we can expect to see existing services dropping in price with the new faster products replacing them, so 500 Mbps over copper at £30 per month plus voice costs may be something we will see advertised on TV in a few years.

The other copper network which is the Virgin Media fibre/coax solution, does have the ability to increase its speeds to compete and in the past has always tried to pre-empt higher download speed products from the BT camp, the question for the future if cloud based services continue to grow will be whether DOCSIS can meet the challenge on upload speeds.

One interesting point, the labs have shown that even existing micro-filters are sufficient for G.FAST, though we would add that just as with VDSL2 to get the best speeds you will want to place the modem as close to the entry point in your home as possible and avoid the signal hitting extension wiring. Though it is possible that the vastly improved error handling from vectoring may counteract this slightly.

G.FAST is going to be attacked as a stop-gap measure towards a full FTTH future, but just as FTTC has brought fibre in some cases 17km closer to properties, leaving several hundred meters of copper, G.FAST and FTTrN will bring it ever closer and while the FTTP on Demand product is too expensive for consumers when the final fibre splitter is just 100m away we may a much more reasonable installation cost to take the fibre direct to the property.


G.FAST node on pole

BT Copper Lab

Mini Telephone pole with 8 lines radiating from it.

4 Responses

  1. Hugo Pickering on 20 Oct 2014

    I guess BT’s publicity on G.Fast is to counteract last week’s share report from Redburn encouraging investors to sell their BT stock due to shares underperforming and their network not being well invested. The report cites Virgin, Gigaclear, CityFibre and other altnets who are providing symmetrical FTTP broadband services that significantly out-perform copper based final mile, which will ultimately be phased out anyway.

    • andrew on 20 Oct 2014

      Visit was on Wednesday 8th October and had originally asked BT for the visit on the 25th September in response to their G.FAST release on 25th September. So not sure how this fits into the timetable for the Redburn sell recommendation.

      But if that was last week as in 13th to 17th October, then setting up this trip pre-dates that.

      By the way have not covered the 3 Tbps over 300km of fibre demo’d in a different lab the same day, since that is more core network.

      As for whether G.FAST will get rolled out and its price and if FTTP would be cheaper will all depend on what the accountants work out. That visit was much a tech demo, than a hear is a finished ready to deploy GEA-G.FAST product.

  2. RichardD on 21 Oct 2014

    A wireless connection speed does not sum up and down links, it represents one direction, both independently negotiable.

    The reason you see about half-throughput is overheads between useful data and other layers, primary the PHY layer.

    You were explained this in your forums ad nauseum, but still you misunderstand it and misreport it as a ‘trick’ of wireless marketing.

    • andrew on 21 Oct 2014

      Have now committed the overheads of signalling to non-volative memory which are a lot higher on wireless (existing standards anyway).

      Reworded the paragraph appropriately