but I expect that using the Broadcom MSAN, they will be following the BT design but perhaps with different parameter settings.
Much of the technology and design will be the same.
Despite searching long and hard in the past I never could find any definite information about the PSD masks and which tones they affect other than there are a few different ones in use by BTw depending upon your distance from the exchange.
They will certainly affect up to tones 60 on most lines to some extent. If your line is good enough this is what causes the inverted U shape when graphing upstream and the lower end of the downstream bins.
Be* appears to be unusual in that it also blanks out tones 476 - 499 for some weird reason. A bit annoying if youre on a line which could make good use of them.
----
One of the earlier parts of the sync process is channel analysis. This is when the modem checks which tones are good enough to use and this happens
before the setting of any sync speed.
In your case the modem will have gone through channel analysis and knows which range of tones it has available to load and the
maximum bits that it can load in each of those tones.
adsl1 standards mean that a minimum of 2 bits per available tone will be loaded.
(the cap of 384kbps means its highly likely this restriction cap will also be set to G.992.1 standards).
So after its gone through channel analysis the next stage of
radsl is setting the sync speed. Normally the modem would fill all the bins (determined as available in the previous analysis) with as many bits as it can so that the sync speed is determined.
However in your particular case.. the modem knows it has a whole range of bins that are good.. yet now the enforced cap of 384 kicks in. Now it It can only load a total of 'y' bits in all of the bins that are sufficient to meet the speed of 384 and its got a whole range to play with.
Logic would imply that the modem would start at tone 33 and fill to the end.....
but in your case its almost looks as if the modem has started filling the bins backwards.. and it ran out of bits to load by the time it reached tone 117.
I suppose its also very possible that its actually loaded the middle set of tones available. Ive not done the actual maths to check up on this - but its something I suppose that could be done if you had the patience to do so... but as a rough guess seeing that tone 500+ is available from your 2nd set of results.. this is exactly what it appears to have done... and dumped 2 bits in each of those middle tones of the available range.
-----
I think it will be a long time before we see a similar graph again. I dont think there will be many lines that are capable of 16Mbps capped to a lowly 384 kbps.
This didnt even happen in days of old when you may have had a good line on a 512kb line. The technology was different - the additional tones werent open for use.. and it wasnt radsl.
Here we are seeing the difference between a fixed rate line with the use of power cut back and a full adsl2+ rate adaptive line that is capped with a profile set at the MSAN.