Thoughts?
I don't think there will be any correlation between VDSL2 vectoring and G.Fast. Not at first anyway, and possibly/probably never.
1. The first reason is that G.Fast, as used by BT, will only start from around 23MHz. There is no overlap.
This also means that, for the sake of backward compatibility, G.Fast in the UK will not be as fast as it could be theoretically. See the attached graph, and the difference between the aqua and purple lines; UK's G.Fast will be either 150Mbps slower than it could be, or have about 100m less range than it could have without VDSL2.
BT are, apparently researching a way to allow for some overlap, but this hasn't happened yet. They might more easily gain by reducing the "guard band" between 17MHz and 23MHz, to allow G.Fast to start from, say, 19MHz instead.
2. The second problem is that, inherently, G.Fast is incompatible with the way that VDSL2 uses spectrum.
VDSL2 breaks the frequencies into bands for upstream and downstream, and the two bands must be used identically by everyone. If some used a frequency upstream while others used it downstream, they would encounter a different (and worse) kind of crosstalk known as near-end crosstalk. Vectoring works on the standard far-end crosstalk only. Near-end crosstalk is more fatal.
Unfortunately, G.Fast works differently. Every frequency is used for both upstream and downstream together, split by time. Say 80% of the time downstream, 20% upstream.
That means overlap with a VDSL2 band would be causing, say, far-end crosstalk 80% of the time and near-end crosstalk 20% of time (or vice-versa). That would be fatal enough.
The BT research in (1) can't fix this distinction between near-end and far-end crosstalk; the best it is likely to come up with is that overlapped segments become barred (in G.Fast) from usage in one direction.
3. Vectoring isn't a magic marker for "solved". Labelling G.Fast as "vectored" does not suddenly make it compatible with any "vectored" VDSL2.
Vectoring solves the crosstalk crisis by synchronising every transmission that is being monitored, and deliberately injecting controlled (but variable) noise onto every one of those transmissions. The controlled noise is calculated in advance to be the exact opposite of the crosstalk that will be experienced on the way to the receiver, so that it cancels out.
The "crosstalk experienced" is, of course, created by all the other transmitters under the control of the chipset. The chipset needs to know in advance everything that is going to be transmitted, in order to calculate the opposition needed.
Vectoring thus only works if the vectoring chips are in communication with, and in control of, every line transmitter.
Declaring G.Fast as vectored, and declaring VDSL2 vectored, does nothing to stop crosstalk between the two systems. There would need to be a single vectoring system that controlled both the VDSL2 transmitters and the G.Fast transmitters.
Corollary: Today's VDSL2 vectoring cannot counter crosstalk from/to the ADSL2+ spectrum either.
4. It may eventually become possible for a single vectoring chipset to control line transmitters using disparate technologies, but that isn't true today.
As an example, look within VDSL2, at profiles 17a and 30a. Profile 17a (like ADSL) transmits in tones that are 4.3125MHz apart, while profile 30a went for a different scheme, with tones that are 8.625MHz apart.
Vectoring chipsets cannot cope with these 2 profiles together; vectoring chipsets cannot figure out the cancellation necessary. That is why Alcatel invented Vplus, and Huawei came up with SuperVector; these eventually became profile 35b in the ITU-T specs, which kept a 4.3125MHz separation between tones. Vectoring is then possible in a system with lines mixed between 17a and 35b.
https://insight.nokia.com/sites/default/files/vplus-allows-mixed-vectoring-with-vdsl2-17a_figure03.jpgIf today's vectoring chips cannot cope with the difference between two VDSL2 profiles, what hope does it have in tracking both VDSL2 and G.Fast?
It might be possible eventually, but today's incumbents need a solution, like right now.
5. VDSL2 will indeed be an alien crosstalker to a G.Fast system, and vice-versa. The fact the either group is vectored internally matters not to the interference it places externally on the other group.
Yes, they could probably use PSD control to make overlapped signals of a similar strength (like VDSL2 with ADSL).
G.Fast will have their own tie pairs to traverse - but they'll only be a few feet long. They'll still have some effect though.
But the show-stopper is the issue of near-end crosstalk mentioned in (2).
6. On the video, Neil McRae doesn't really cede the cabinet on technology grounds, but on financial ones. Once a cab has been upgraded with VDSL2 SLU by Warwicknet, it becomes less viable for BT to choose to upgrade to FTTC too. Without the FTTC upgrade, there is then no power & fibre for BT's subsequent G.Fast upgrade.
That's how I read it, anyway.