Dear All,
First of all, thank you for the many helpful, not to say hilarious
suggestions. Much as I appreciate these, I'm afraid that - apart from the risk of physical violence towards me in my search come REIN or SHINE - I live in Glasgow and there are probably several thousand premises within reasonable distance of the 1.7km route of my D- & E-side cables between my house and the exchange. So, I think the prospects of tracking it down with some of the suggestions are regrettably vanishingly small.
On a more technical point or two: earlier in this thread, I posted the QLN tests the HG612 captures at (re-)sync time. When the router resyncs because of this 'problem', you can see that in fact the noise floor rises across the entire ~2MHz spectrum by about 10-20dBm. In fact the noise characteristics of the line (LW/AM station spikes & all) is very well preserved, and there is no sign of any particular frequency or set of discrete frequencies that you could point an finger at. So this noise has a very broadband effect.
Secondly, the drop cable is ~10m or so the pole-top DP and then u/g to the PCP 170m way, and then to the exchange. Sunset here is at ~20:45 and Sunrise at ~06:00, so as B*K said it doesn't fit the street lighting scenario well. In fact, although there is clearly something causing a marked bathtub in snrM reduction, its time is not precisely the same every night. Indeed the same bathtub occurs sometimes during the day.
Occasionally it spills over into the audible voice band, but tends to disappear before I can get a voice fault tested and reported.
But to keep the debate alive here is a 5-day snrM plot - many thanks again everyone, especially fot the light relief
