Thanks guys, to answer some of your Qs:
- The ISP is Plusnet, the problem started around January this year.
- Off-peak download speed tests are always fine. Things dip in the evening and are back up to full speed before midnight.
- Upload speed tests are always wavering, peak or off-peak. I reported it to Plusnet but I don’t think they took much notice of it.
- The BQM is the same when the line is totally idle, I switched off wireless with no devices connected one day to check this. Same packet loss and avg latency hump - starts in evening, resolved by midnight.
- The strange regular spikes on the BQM appear to be a quirk of the Netgear modem-router in use at the time, they are not there with the Technicolor router.
- I swapped routers, cables, filters, phone, am using the test socket and tested wired with diff machines (PC and Mac) - made no difference. The only thing I noticed is when using the Netgear, the upload speed tests are less wavering than with the Technicolor, but they are still wavering.
- Interleaving is off (left to DLM) location is West Yorkshire.
- The pings reported by BQM are under 15ms off-peak which matches a command line ping. (The TBB speed test never gives a reliable figure in my experience).
- From looking manually at the SNR margins, they are always rock-solid e.g. 5.9, 6.0 or 6.1 even at night and problem periods.
- DLM in fact previously always selected 3dB down target SNRM for my line and this was solid too.
- Plusnet reset DLM in January when they saw this, target has since stayed at 6dB (down and up) and interleaving off.
- I don’t know why the MTBR indicates 6 resyncs in a day, this is definitely not accurate.
- The line is very stable and only drops sync if I cause it, or Plusnet test the line.I know this from monitoring the uptime in the router stats, the BQM graphs, and the RADIUS plots which Plusnet attach the same time as the line reports.
- At the time of the line test, it was over 5 days since a resync, which in turn was caused by Plusnet testing the line previously Before that it was over 5 days again due to the DLM reset.
- I was wondering what do the errored seconds in the line test refer to exactly - over what time period (0 upstream, 8 downstream)?
- In fact, now I look more closely, Plusnet posted complete results for 2 line tests, on 2nd Feb and 8th Feb. They both have absolutely identical “MTBx” figures - surely this can’t be right, especially as MTBR is so wrong, it’s like the MTBx figures are stuck? Other figures are different between the 2 tests, and the max stable and fault threshold rates are totally different.
- Plusnet’s notes to the engineer said “low MTBE causing slow speeds in evening”
- The engineer said his tester picked up “only 4 errors” in 5 minutes, which he said was nothing.
- He then rang some department who could access stats he couldn’t, they told him the line was absolutely fine and to report back to Plusnet slow speeds were probably due to congestion.
- The current state of play is I am waiting for Plusnet’s update after the visit.