Sorry - I'd normally do more analysis than this, but I have almost no time right now...
What I can say, from looking at the graphs on the original Dropbox page, is that you appear to have very high QLN values - especially at the higher frequencies.
QLN= Quiet Line Noise
QLN is measured before synchronising, and the graph will not change until there is a new sync.
I understand that a QLN of "-140" represents really good silence. Values of between -120 and -140 seem to be pretty reasonable (and are what I get). I suspect that values around -100 are not very good.
On my line, I get values starting at -130, dropping to -140 at tone 350, suddenly rising to -120 and then gradually dropping to -135 at tone 3900.
In your case, it suggests that the modem has measured "high noise" throughout the spectrum (above tone 500). This means the SNR values will be correspondingly lower, and thus that the number of bits that can be carried on each tone are lower.
Your SNR graph is 40dB in higher D1, 30-35dB in D2, and 25dB in D3. Mine runs at 45-50dB in higher D1, 38-42dB in D2, and 25-30dB in D3. I'm getting better signal levels throughout.
The good news is that, because you are so close to the cabinet, the modem can use all of the tones.
You broadly seem to get 6-8 bits per tone over the full spectrum. I get 14 bits around tone 500, dropping to 4 bits at tone 3900.
Your line is 50 metres (ish), while mine is 400 metres. To back up the fact that my line is a lot longer, the three attenuation figures are: D1=11.4dB, D2=27.4dB, D3=41.7dB.
So, it is all down to noise. Presumably the noise is crosstalk.
Again... sorry. I won't have chance to add more to the conversation.
Edit: Added SNR and length comparisons.
Edited by WWWombat (Fri 22-Mar-13 01:15:09)