I've been meaning to get around to playing with this... and SWMBO is away at the weekend, so I might get a chance!
Are the plotting scripts available anywhere? If so, I'll try doing my own plotting.
In the meantime, I wanted to point this out...
asbokid has graphed whatever it was I sent him and produced this:
http://img846.imageshack.us/img846/3836/statsop.jpg
Now i'm not sure what I'm looking at but does it appear a bit strange to anyone else? Seems the first image roughly correlates in shape to the 2nd one, except for the big "gash" in it.
I'm just trying to get my head around it.
...
BT are using VDSL profile 8c, annex B and according to this document band plan 997 truncated at 7.05MHz.
...
3) Line noise. In the higher tones this nicely mirrors the SNR as you'd expect. In the lower tone it doesn't - I suspect this is due to spectral power management.
I noticed the difference between the 2 graphs as being the location of the "gash" (although I'm not sure everyone has the same idea of which bit is being called the gash!): Orly's gash (or dip) appears to end around tone 250, in the SNR graph (ie top graph), while griff_90's gash appears to finish around tone 500 in the same graph (but second down in their image).
I suspect this is indeed caused by the spectral power mask applied at the cabinet - and that the 2 cabinets have different masks.
The document linked above:
BT's VDSL Primer from 2009, shows the masks BT intended using at the time. See pages 31 and 32 (of 66).
The masks may put the top of the gash at 600kHz (for cabinets far from the exchange), up to around 2MHz for those closest the exchange.
So... I bet that Orly's cabinet is further from the exchange than griff_90's.
However... I'm not convinced that this document represents the actual deployment at Openreach today. Some of the screen-grabs here show that BT are using US0 (25kHz - 138kHz) which isn't shown in the band-plan 997 in that document, and the spectral mask makes no allowance for it either. I suspect there are some differences in reality...
On a separate note: this mask is a way in which the E-side distance can still affect the bandwidth you see as a customer; the overall behaviour of VDSL2 isn't *just* dependent on the D-side distance.