Internet connection keeps dropping

[#1]

Well I posted the good news last Friday (16th), so now for the bad news.
The problem continued after the REIN engineer left.  There was a drop of about 1.5dB on the line that day at 16:45 and since then there have been regular 1.5dB increases at about 10:05 and 1.5dB drops at about 17:00.  In addition there is the killer event which is a 6dB change in the noise margin that has been occurring at widely varying times.
The first graph attached is when the REIN engineer was here.  It shows a peak where the noise margin was at 16.5dB between 12:03 and 12:18, a level 1.5dB higher than it was outside of that time period.  The REIN engineer called 10 minutes later to say that the problem had been identified.  So I am assuming that when that peak occurred the engineer was switching the problem device on and off, as a test.  The difference in noise margin that the device made was 1.5dB.  On a 6dB profile my line could tolerate such 1.5dB rises and falls but it wouldn't be able to tolerate the 6dB drops that have also been occurring.
Since the 16th, my line has synced between 13000kbps to 14400kbps @ 6dB, 7800kbps to 8900kbps @13dB and is currently at 15105kbps @3dB.  The current line speed can't possibly be maintained because the next 6dB drop will cause the router to resync.  I suspect that unless the source of the problem is eliminated my line will end up syncing at about 8000kbps with a noise margin that goes as high as 14dB in order to tolerate the variations that are occurring.
I've checked my ticket and the PlusNet team have spotted the ongoing problems and have 'raised an escalation with our suppliers'.  I will add a link to this posting onto my ticket so as to provide additional information.

[#2]

Hi pnf,
I rather suspect that Mr Neighbour came home and said "Stuff the neighbour's internet connection, I'm not buying a new Freeview box".  Given that you do not know which neighbour, you cannot even extend the offer you suggested of buying a replacement... unless you go knocking on doors asking if BTOR called there recently... but if they are not good enough to not use the problem box, they are not likely to tell you it was them.  If you offer a free box then everyone will claim that it was them.
If you do find the neighbour and offer that box, make sure it is a swap and you take the duff one away so that they cannot keep using it.
Looking at your useful analysis, it would be helpful if you could distinguish between resynchs which deliver BENEFIT (lower SNRM and higher synch) and those which deliver DETRIMENT (lower synch speed).  The former might well the effect of the DLM 10 day training period.
It is also useful to post the graphs relevant to analysis of the resynchs.
Note that away from resynchs, a RISE in SNRM reflects the REMOVAL of REIN, where as a SLUMP in SNRM reflects the START of REIN.

Kevin

[#3]

Hi Kevin,
I don't think I'll be knocking on any doors.  If the box was taken out I'm now not certain that it would completely solve the problem because it didn't appear to be the cause of the 6dB rises and falls on the line.  If it could be proved that the box was the sole cause of the problem then it would be worth the cost of a new box, within reason, just to get rid of it, particularly if it belonged to an elderly person worried about the cost of replacement.
With respect to the resync analysis, I've now added to the pdf file the speed the line resynced at and the SNRM immediately following the resync.  I've also attached the graphs for each resync. 
I believe there have been a number of disconnections from the Internet, where the DSL connection was maintained.  Unfortunately RouterStats doesn't capture those events.

[#4]

Here's a good example of what is happening ...
At 08:44 today the router was syncing at 15,105 kbps with the noise margin at 3dB.  It had been stable on this setting from 15:50 yesterday and right through the night.  At 08:45 the interference started and because the line was on a 3dB profile the CRC graphs captured it quite nicely.

[#5]

NPR,
Cool graphs!
Let's hope this gives DCT something to chase!

Kevin

[#6]

Our Faults Team have full visibility of this thread so they'll be able to review your graphs and they'll be back in touch as soon as they can.

[#7]

Thanks Chris.