Line trained but stuck at 12dB SNRM, what do I do now?

[#1]

OK, I'll keep my eye open. Bad time of day really to be swapping, but you know that. At least the v4 will start with a clean set of error figures.

[#2]

@Anotherone, what don't you like about the bits/tone?  It doesn't look that much different to the ones I was getting back in October.
There is something very odd about the CRC count.  See attached graph.  Not exactly the sort of graph you would expect from noise generated events.
Do you know whether the row marked CRC(FAST) on the Linksys data is the same statistic, and what the second undefined row is?
As always, many thanks for your time.
Colin

[#3]

I would guess the second row labelled undefined is actually CRC(INTERL) and the second one marked HEC(FAST) is HEC(INTERL). Since your line is interleaved the INTERL lines apply for you. Not sure what the problem is with your line though.

[#4]

I think spraxyt is spot on there with the errors explanation. So at least the two modem/router are consistent with the figures.
As for the CRC count (and considering the Bits/Tone) this looks like some steady continuous interference. Check and see if the rate is the same in daylight hours.
With the Bits/Tone, I've looked back at the October ones - I'd forgotten quite how horrible it was really, but comparing to now there's some subtle differences and more obvious ones. The SNR per tone is somewhat less for quite a few tones, reduced bit loading and there are more missing tones, noticeably 57 and nothing above 116.
Can you post a daytime Bits/Tone so we can see if the differences are significant as well as checking the CRC rate?
This does look like an interference issue, assuming you don't have audible noise on the line and if the line attenuation plot is accurate (it should be with the v4) there wouldn't seem to be any significant change in the line, so it's a case of eliminating everything local again.
I'm suspecting a PSU has got noisier unless you've got any new electrical/electronic equipment / lighting etc.
There's also the question of what's causing what would seem to be some spike/surge at 0530 which we never discovered before. If it's spot on the same time every day it has to be on a time switch. It might be worth checking what the Economy 7 switching times for your vicinity are. Also see if there's a CRC jump at that time.

[#5]

@spraxyt thanks for that insight, I am sure you are correct.
@Anotherone as far as the 0530 issue is concerned I am sure that was just the DLM adjusting the SNMR during the training period.  I have not seen any more of these since the training completed and I saw the same issue during the training in October and after that it ran for three months without a resync.
The CRC errors are astonishingly consistent.  Looking at the rate over the 6 days uptime for the stats I sent the first time they happened every 1.3 seconds.  Overnight since I restarted yesterday they are also every 1.3 seconds and the line on the chart is absolutely straight.  I don't think it is anything at this end , at least if it is then it is something that is on all the time since it did not change overnight.  I will shut the house down later however just to check.
Is not the reason that the bits/tone are a bit lower due to the fact that it is running at a higher SNMR?  I am really not worried about the background noise level, I am confident that if I push the SNMR down in the router then it will go at 2016 again which is as good as it ever has been and not bad for 7km I think.  It would be nice to get to the bottom of the 1.3 second CRC errors though if that is what is causing the DLM to want the high SNRM.
As I said, I will shut the house down a bit later and see if the CRC errors stop.

[#6]

Well whatever is causing the CRC errors every 1.3 seconds as reliable as clockwork is not in the house.  I have shutdown everything in the house except the router and it still clock up 6 errors every 8 seconds.

[#7]

I assume that included noting the CRC count and switching off the computer/removing it from the locale (& not connected by ethernet) for a defined period and checking the CRC count when you switched on/reconnected again just to be sure it's not the computer
The missing tones are likely to be caused by the noise,  but the bit loading may recover completely when you push down the Target SNRM, but I wanted to see a daytime Bits/Tone to see how much of the last one was possibly affected by the general dusk to dawn issues.

Quote from: clanla

@Anotherone as far as the 0530 issue is concerned I am sure that was just the DLM adjusting the SNMR during the training period.  I have not seen any more of these since the training completed and I saw the same issue during the training in October and after that it ran for three months without a resync.

That thought did cross my mind, I had a look back through the previous thread etc. and I couldn't find quite enough info to be certain, but I knew you'd have the data for that. In which case we can forget about it then in the sense that it appears to be DLM not liking the CRC errors.
Are you still on the SSFP ADSL v1.0 filter? I suppose you ought to do the obligatory filter swap with your best filter (i seem to recall you had one that was not so good).
Whilst you say you aren't concerned about the noise, it is that which is causing the CRC errors and whilst you can push the Target SNRM down, that or a worse error rate will be affecting the actual throughput speed and possibly latency although strangely the Interleaving depth isn't especially high!  Unfortunately I don't think the error rates are quite high enough for BTw to investigate a fault despite DLM not liking them, but you could try running through the faults checker to see if it leads anywhere after you've tried a filter swap.
Other than that, it's a case of get your radio set out and see if you can track down a source of noise that runs 24/7.
Is that neighbouring line still on a DACS?

[#8]

Yes, I tried it for a period with only the router powered up and CRCs still arrived at 1.3 second intervals (though actually I could only tell the average rate over the period in this case of course).
Daytime bits/tone attached and also SNMR.  The signal is very quiet during the day as you can see, and only drops a couple of dB at night.
I have two Speedtouch filters as well as another filter that did not seem as good.  I will swap over to the other Speedtouch one in the morning when it is quiet.  Or perhaps I should try the not-so-good one first in case there is a design flaw in the Speedtouch one, then go to the other Speedtouch (assuming the not-so-good is still not so good).
Do you think that what one might call 'ordinary' noise could give such regular CRC errors every 1.3 seconds day and night week after week?  Presumably whatever it is could be picked up anywhere along the line or at the exchange, in which case I would not pick it up on the radio.  Would it not vary through the day if it were normal radio interference?
The neighbouring line is not on a DACS.
Colin

[#9]

This is not "ordinary" noise to cause errors at this high rate. And it's not "normal" radio propagation which is generally pretty low in the day but increases after dark. You'll get the odd day where there's a bit of a lift on MW/LW but such propagation is never that steady.
The chances are you will pick it up on a radio, especially close to the phone line or mains, the tricky bit is identifying which interference to listen to when it's 24/7. Whether there is some lighting on somewhere nearby that's 24/7 - a road sign? Any electric fence units nearby?

[#10]

I tried the not-so-good filter and immediately got pushed up to 15dB SNRM (even though I followed the 10 minute rule).  The DLM really doesn't like me.  CRC errors still there.  Then tried the second Speedtouch filter, stayed at 15dB, speed 960.
The electric fence seemed like a good idea but none near the house as far as I know.  No lights or street signs either.
Unfortunately my AM radio seems to have broken so I can't try that either
To get the check done on the line do I go through the PN speed fault reporting system?

[#11]

If it's a REIN issue then it'd need to get raised as a fault, yes - I'd recommend summing up this thread and your issues in the comments box though rather than just linking to the thread as that'll make things a lot easier for our faults team to look into.

[#12]

Thanks Matt, that's a very helpful suggestion.
Colin, anyone in the household got an HTC mobile phone?
Whilst I was whiling away the hours waiting for SWMBO in the Eye Hospital waiting area, I noticed someone using their HTC mobile and it had a thin slot of light at the top that pulsed roughly every 1-2 seconds

[#13]

When the routers were swapped, was the new power-supply used as well, or was the old one just connected to the new router?
When you powered down the house, were other power supplies like 12v lights and DECT phones unplugged from the mains and TV/HiFi/set-top box equipment turned off not just placed on standby?
I must say I like the mobile phone hand-shake thought. If I stand near an AM radio that picks up the blips of that, but at rather longer than 1.3 second intervals I think.

[#14]

@Estragon and Anotherone
Thanks for the suggestions, yes everything in the house was switched off at the wall, all lights off and all mobile phones off, and when I swapped routers I used a different PSU.
I have found another AM radio so about to have a go with that.

[#15]

I haven't picked up anything on the radio so have reported a fault.
It seems to me that there is a fault with the fault reporter.  I can't find a feedback link on PN's website to point it out.  Having taken the basic details it says that it sees that there have been dropouts on the line and proceeds down a path that assumes that the dropouts are the problem, whether they happen at the same time each day and so on, whereas the dropouts are due to me doing all the things it says I have to do, plug into the master socket, try different filters, different routers and so on.  So it never actually asked me what the problem is.  Still, I put a description of the problem in the Comments box so hopefully someone will read it.