I've seen a few posts from people having similar problems to myself, so thought I'd punt a question in here and see if anyone has any further ideas.
Situation is briefly:
- ADSL Max connection 4 km from exchange, rural Scotland, line 4 years old (or at least the extension from nearest BT street cabinet to our building is)
- problems with several (domestic) routers over the years, with drops and badly declining speeds, installed an 877W about 18 months ago and has been more stable (probably, one's memory being what it can be....)
- line has sync'd at 3 Mbps absolute best ever, more often ca 2300, but now generally only 1856 kbps
- usage is not heavy - 3GB/month - small business style usage - e-mails, file transfer, web, no VPN, nothing fancy at all, no external access in
- had not monitored the line very much until 2 months ago when a regular dropout started (more on this below)
- raised a Case with Cisco support and we got to the point where we believe the problem is a noisy line. This included ADSL f/w ==>> 4.0.015 (IOS is 12.4(20)T)
- no improvement noticed due to the f/w, and after tests straight into Test socket behind the master socket also showed drops and noise, we called BT out to test the line (yesterday), bad news was they found "no problem" in the final loop from exchange out to us. Currently back via ISP (Plusnet) to escalate
I've monitored the SNR Margin and sync speed using SNMP and correlated to the debug logs over the last two months :
SNRM (downstream) varies from circa 10 dB (even a later 14 dB if synced when noisy) to 0 dB and can go briefly negative. Sync speed varies (as would be expected) depending presumably on the line noise prevailing at the instant of training. Drops are often - but not always* - associated with Loss of Margin or similar. Errors are a range of LOM, LOS, LOF, LCDi, aal5 rx errors. Not sure really which error counts to monitor but have noticed that the Uncorrected Blocks errors jump corresponding to drops or threatened drops. A Quiet Line Test (17070 opt 2) can show a line as quiet as a mouse, or with quite a lot of static (sounds like when lightening discharges interfere with an AM radio), but all very intermittent. [*mind, the SNMP poll period is 5 secs so I could have missed very short losses]
One of the strangest things I've found is that the drops (when the flapping is at its worst) can be regular enough to time your watch by - every 30mins 08secs +/- 2 secs between ATM0 going down and the previous ATM0 going down. The modem always reconnects immediately in circa 20-25 secs. On the other hand the connection can stay up for 4 -5 days including with a LOT of noise, short LOM's and so forth.
I've re-tried other routers including another 877W, and a Netgear DG834G which has the reputation of hanging on well to poor lines, and it can be OK, but it can also sync way down at 128 - 192 kbps and resets the profile to unusable speeds when it does this (something I've not seen with the 877W). However, graphing the SNRM and sync sped on the Netgear I detected the same 30:08 pattern one day, not during drops, but showing as downward peaks in the SNRM figure, so I'd be fairly sure the 30:08 min events are not to do with the 877W
I still think it's a noisy line and "just" a question of getting BT to recognise/find it and fix it, but from seeing posts on these fora about incompatibilities between 877W's and the UK system (albeit much seem to be 2+), and about FW upgrades, I would still appreciate any thoughts as to residual things I might have missed in our equipment.
Thanks in advance
>> One of the strangest things I've found is that the drops (when the flapping is at its worst) can be regular enough to time your watch by - every 30mins 08secs +/- 2 secs between ATM0 going down and the previous ATM0 going down. The modem always reconnects immediately in circa 20-25 secs. On the other hand the connection can stay up for 4 -5 days including with a LOT of noise, short LOM's and so forth.
Ask to the provider to perform a 24h bit error rate test on the line (BER), this should show the problems. This if you can stay 24h without the line.
Finding the noise source(s) over the 4 km line is another matter. However, the noise source that appears every 30 min may be caused by a factory installation. This regular occurrence leads to this type of guess.
Hope to help
Thanks, Giuseppe. I have another visit from BT this morning, presumably the next level of support up from before; hopefully they'll be able to discuss this with me.
Can the bit error rate be monitored from my router? Closest SNMP variable I can find seems to be 1-day Errored Seconds (adslAturPerfCurr1dayESs)
you can use the router stats as long as BT accepts these results.
The BER test can be done with special ATM cells (ATM OAM F5) that the modem in your router has to send back to CO, so it may be not totally intrunsive and doesn't need a dedicated ATM PVC.
Hope to help