Connection keeps dropping

[#1]

My connection has been dropping sporadically for a minute or so, which has been an irritation, but relatively minor. Yesterday it started doing it with a  vengeance, dropping more frequently and staying out for longer. The stats from the DSL modem indicate lots of errors, but they don't mean much to me - I've attached some details to the bottom of this posting because I'm sure they'll mean something to others in here.
I'm happy to replace the modem if that's the likely cause, but I'd prefer to be able to eliminate a service/line fault first if possible. Is there a convenient way to do that (given that it seems to be working OK at the moment)?
I've checked that there's nothing (other than the modem) connected without a filter. I don't have a BT 'master socket' (just an old style junction box in the loft) so I can't just plug in to that to eliminate the internal wiring, but there have been no changes to any of that for years.
Edit: Oops, forgot to attach the stats - here they are, sorry about the lack of formatting:

ADSL STATISTICS

Mode: Autosense 
Type: Interleave
Line Coding: Trellis
Status: SHOWTIME

Downstream Upstream
Rate (Kbps): 896 448

SNR Margin (dB): 11.6 17.0
Attenuation (dB): 58.0 31.5
Output Power (dBm): 15.8 11.9

Super Frames: 3462119 3462062
Super Frames Errors: 1388059 17
RS Words: 58856026 58855054
RS Correctable Errors: 49754782 89
RS Uncorrectable Errors: 1825146 0

HEC Errors: 987513 6
OCD Errors: 3 0
LCD Errors: 0 0
Total Cells: 123198510 0
Data Cells: 1280364 0
Bit Errors: 0 0

Total ES: 50665 0
Total SES: 50041 0
Total UAS: 3361 0

[#2]

Quote from: Boxersoft

I've checked that there's nothing (other than the modem) connected without a filter.

That's not right.  The modem needs a filter, just the same as everything else!  The only difference is the modem plugs into the ADSL socket on the filter.  Some people make a mistake and use the wrong type of cable but the cable between the modem and the filter should have the same (smaller) connector on each end. 

[#3]

No, I don't think so. Analog devices (phones, faxes etc.) have to be filtered, but the ADSL modem shouldn't be. What you're describing is a filter/splitter, in which the phone socket is filtered but the ADSL side is just an unfiltered passthrough as far as I am aware.

[#4]

From those results it would appear that you are getting a very high proportion of uncorrectable errors in relation to throughput and that is almost certainly the cause of the problem.
Can you carry out a simple test over night when you don't need the phone
Disconnect all other items from the phone line and connect the router to the test socket and report the stats in the morning. If you have a spare router cable can you try that. Note that you will probably need to use a filter due to an incompatible connection
If the errors are as bad then it is a line problem but if they are very low it is something in your property.

[#5]

Quote from: Boxersoft

No, I don't think so. Analog devices (phones, faxes etc.) have to be filtered, but the ADSL modem shouldn't be. What you're describing is a filter/splitter, in which the phone socket is filtered but the ADSL side is just an unfiltered passthrough as far as I am aware.

Maybe you are correct, electronics is not my forte.  I had always assumed that there was a high pass filter on the ADSL side of the filter/splitter as well as a low pass filter on the phone side.  However the cable you are using to connect your ADSL modem to the phone socket cannot be the one that was supplied with that modem.  

[#6]

Thanks for the input. I can't easily carry out that test because, as I mentioned, this house doesn't appear to have a BT master socket, just an old style of junction box in the loft. I could rig something to that, I guess, but I think that strictly speaking I'm not supposed to touch it (although there's no cover on it, so I guess a previous owner did!).
However, I tried leaving the modem switched off last night. This morning it seemed to take a long while to reconnect, but since it did it's been fine and surfing feels as snappy as it ever did (I only get about half a megabit due to distance from the exchange). Checking the modem logs now (after 2.5 hours uptime) it's showing far fewer errors (details below). With a more stable connection I was more easily able to check the PlusNet site and noticed several 'Planned 21CN Broadband Network Maintenance' items with the advice "Any customers having problems connecting once the maintenance is complete are advised to switch off their equipment for 30 minutes before powering it back up and attempting to reconnect". I was wondering whether this might be what was causing the instability here and, if so, whether the overnight rest might now have settled it somehow. Does that seem at all likely?

=====================
Mode: Autosense 
Type: Interleave 
Line Coding: Trellis 
Status: SHOWTIME 
 
  Downstream Upstream
Rate (Kbps): 896 448
 
SNR Margin (dB): 11.4 17.0
Attenuation (dB): 59.0 31.5
Output Power (dBm): 15.8 11.9
 
Super Frames: 561620 561623
Super Frames Errors: 12 3
RS Words: 9547544 9545602
RS Correctable Errors: 116065 49
RS Uncorrectable Errors: 65 0
 
HEC Errors: 10 1
OCD Errors: 0 0
LCD Errors: 0 0
Total Cells: 20176152 0
Data Cells: 853231 0
Bit Errors: 0 0
 
Total ES: 30 0
Total SES: 9 0
Total UAS: 237 0

[#7]

Quote from: Oldjim

... connect the router to the test socket ...

No Test Socket, Oldjim, as Boxersoft mentioned earlier.  In the absence of a Test Socket the best thing to do is to unplug all other phone equipment (in case one of the filters is faulty).

[#8]

Planned maintenance very rarely causes instability. If you still experience problems then please report this to the Faults Team.

[#9]

Thanks Ben. I'm still experiencing drops, although there now seem to be longer uninterrupted periods between them. When it drops I'm often unable to see the modem's web UI at all on my network (though it still responds to PINGs) so I think it might very well be a hardware fault. I've ordered a new modem - if I still have problems once that's installed, I'll file a report.

[#10]

Good call on the hardware, have you tried a factory reset first though? Don't forget that we can send hardware @ £4.99 subject to a 12 month contract if you don't want to pay full whack for a router elsewhere.

[#11]

Yes, I tried a factory reset - it didn't help. I hadn't realised (or had forgotten) that PN can supply modems in return for a 12-month contract until it came up in another thread, but which time I had already ordered a simple single-port unit as a test/spare - nowhere near as good a deal as £4.99 for one with 4-port + wireless, but there you go.
The replacement is behaving much the same way - connection stability varies from several hours rock-solid down to mere seconds - it's been unusable all afternoon. I've also discovered that I do in fact have an NTE5 master socket after all - an unused and overlooked socket. I tried connecting directly to the inner socket of that but it's no more stable than otherwise, so I'll go ahead and file a report.
Actually I'm beginning to suspect some correlation with rainfall (we've had a fair bit of it recently). I don't know if it's possible that a joint somewhere is suffering ingress but, in any case, stability is clearly extremely variable. Presumably if the line is tested (or whatever happens) during a good period, nothing will show up. What happens in such cases?

[#12]

If your line hasn't shown any problems in the few days before the line is tested then the we'll say 'no fault found'. However just speak to the Faults Agent who calls you and ask for them to put the ticket on hold to check again in a few days' time.

[#13]

I reported the fault ad the initial check showed 'no fault'. Not surprising really, as I couldn't kick-off the test until everything settled down. I had a call from someone at PN this evening though. Although things have been much more settled recently, I'm still seeing drops and apparently they are plainly visible from your end (along with a mysterious downstream SNR margin reportedly of zero - at my end it's a similarly odd 2147483647...).
Now being referred.

[#14]

From a KItz thread

Quote

Those large numbers for the downstream margin are the result of a small bug in the Netgear firmware. What they actually mean is that the noise margin is negative, so it's no surprise that the router loses sync soon after.

] Do you have a Netgear?
OT a bi,t but here is something about that number http://en.wikipedia.org/wiki/2147483647.

[#15]

Interesting. A bit more research on that suggests that it's not negative as such, but "A special value of 0x7FFFFFFF (2147483647) indicates the line attenuation is out of range to be represented" (RFC4706).
This is with a Solwise ADSL unit that I got to replace a D-Link that I thought might have been the cause of the problem. The D-Link was consistently reporting SNRMs around 11 - 13. The Solwise has consistently shown it as a huge positive number (presumably that same special value), but is generally working well apart from the line drops (which, although still troublesome, are less frequent and less persistent than they were).