After the DLM trial - small puzzle

[#1]

After the recent DLM trial ended, I decided to stick with a fixed profile of interleaving off and a 6dB target SNR both ways. On the final day my line was syncing at 8519 / 816. Fine.
A few days later, on 1st March, I had to force a resync late in the evening after my connection dropped. The line came back at 7684 / 808. Also fine, as I know that my line is noisier at night.
Since then, though, there have been two resyncs from the BT end, one on March 4th at 2pm and the other today at 8.55am. Both of these were in broad daylight, and I would have expected the sync rate (with a constant target SNR) to go higher. However, each time the line has resynced at exactly the same speeds: 7684 / 808.
(Another minor puzzle: after the latest resync I seem to be getting more near-end CRC errors on the line - about 50k in 25 minutes, whereas previously it was taking about three days to reach that number. That's despite a relatively high SNR of 8 at the moment. The weather is bright and sunny, so I can't see any problems there.)
Any ideas about what's going on here? Previously I might have thought I'd been put on a banded profile, but that shouldn't happen now, and anyway 7684 isn't at the top of any band, as far as I know.
Graham

[#2]

Hi Graham,
Profile shows exactly as it should:
WBC 160K - 24M No delay (INP 0) 6dB Downstream, UC No delay (INP 0) 6dB Upstream (ADSL2+)
Have you tried a resync when the SNR shows as 8dB (or isn't 6dB) to see what happens? Maybe there's some local interference or something causing the noise to be higher and thus the higher number of errors?

[#3]

Thanks for that info, Dave.
I haven't tried a resync when the SNR has been higher than 6. It's been a solid 7 during daylight hours today, but that's not different enough from 6 to bother with a resync, I think.
The error count seems to have quietened down considerably. Currently the line's been up 1 day 9 hours and the count is 59k, so only 9k more than after the first 25 minutes.
I'll stick Routerstats Lite on to monitor what happens between 6pm (start of dusk here) and when I go to bed.
Graham

[#4]

Just in case anyone was on tenterhooks to see what happened after my previous post:
The line's still up, despite the usual night-time volatility.
I'm attaching two Routerstats graphs - a day and night comparison. I really don't think the difference is down to any lighting or appliances in the house - I've tried controlling for just about everything. We have 'heritage' street lighting and that's my prime suspect, though it could of course be something else.
Graham

[#5]

If you think the lighting is the cause, then the 'longer' days we are experiencing now should help to confirm that    Unless you council has money to burn and insist on having them on during the day 

[#6]

I confirmed last year that the noise problems are dusk till dawn - independently of when we switch on lights, TV, etc.
Hmm. I wonder how much this will affect speed when we eventually get fibre to the cabinet (end of 2011 maybe). I'm intending to get new house wiring then, but there might still be problems if the street lights are causing interference.
Graham
ps. since 2am resync on 11/3 I'm now getting speeds of 7562 down (a bit lower) and 816 up (a bit higher). Daytime SNR is 8, so the line shouldn't have too many errors - so far just 9048 CRC errors in 1 day 7 hrs. Pretty good.

[#7]

Dusk 'til dawn very probably means street lights or something very similar. At least it seems better now.
FTTC is still a DSL technology and will be suspect to noise etc. just with less copper length there's less places the noise can be created that will interfere. We've only got 30 odd lines on FTTC so far so not really able to give you a idea of what will happen. We do know that speeds on FTTC will reduce as more people get enabled on a cab because of the cross talk and interference so it may well cause some lower speeds. The fault threshold rate though is 15Mbps so if sync speed drops below this we can raise a fault.

[#8]

The line's been up for 5 days 16 hrs now, and the near-end CRC error count has gone up to only 25600. Daytime SNR has dropped slightly, to 7, but otherwise all's fine. I'm resisting the urge to experiment with resync times.
Graham