Another interesting point - the dip in the QLN around 1307kHz - Any guesses? There's transmitter the other side of London on 1305kHz near Enfield: https://www.google.co.uk/maps/@51.6504011,-0.0083601,3a,69.4y,275.8h,89.84t/data=!3m6!1e1!3m4!1skVgXliuIoE9C0jOvjmFNtA!2e0!7i16384!8i8192?hl=en However, that should couple equally into each side of the line pair and be rejected as a common mode signal.
That's not daft at all, but seems to be what is going on. Everywhere that the quiet line noise rises, if I tune a MW radio to that frequency, there is a radio station there.
That means the line is behaving as an antenna, which probably means my DSL signals are also being broadcast over the mediumwave spectrum!
There are radio stations receivable on: 558, 720, 909, 1053, 1089, 1107, 1152, 1215, 1305, 1458, 1548kHz. The one on 1305 is by far the strongest, receivable on the neighbouring frequency channels. These correspond to VDSL2 tones: 129, 167, 211, 244, 253, 257, 267, 282, 303, 338, 359. On each of those tones and neighbouring tones (dependent on the frequency) there is an uplift in the QLN figures, a drop in the SNR figures, and as expected a drop in the bit loadings.
Looking at the ADSL line (which only goes up to tone 256) I see a similar correlation there - so both lines seem to be affected by AM radio stations.
Looking at the shape of the curve, it does have the broad base characteristic of a relatively high-powered broadcast transmitter. (Relative to the tiny power of the xTUs at either end of xDSL link.) I agree, one would expect such a signal to be rejected as common mode but we have already speculated that the AC balance of the circuit is poor. Hence poor rejection of common mode signals.
Thanks, and yes, that is where I found the grid reference to then locate exactly where the transmitter is - almost the diametrically opposite side of London to me. Given that both of our lines seem to have very similar QLN profiles matching the AM stations that are available, either both lines are similarly unbalanced, or it's entirely normal for the quality of cabling around here.
I'm not surprised given that the various underground joints in the copper cables that I've seen engineers working on have been a mass of untwisted wires - I saw a number while an Openreach engineer was investigating audible noise issues on this line last year. He gave up and switched the line onto a different pair back to, I presume, the PCP next to the FTTC cabinet.