Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[MikeZ]

Hi all,

One morning last week my FTTC dropped for approximately 3 hours between 02:30 and 05:30. BT's explanation was an exchange fault preventing stable synchronisation. The following morning, DLM kicked in with interleaving, adding around 15ms latency and reducing DS/US speed by about 6Mbs/2Mbs, respectively. It has now been this way since then.

I'm uploading stats to MyDSLWebStats (user MikeZ) and would be very grateful if somebody could have a look and see what my line looks line now and whether the DLM is likely to move things back again. Prior to the incident last week it had been very stable for months.

Thanks in advance,

Mike

[boost]

Can't see MDWS atm but if that outage affected everyone on your cab, then lots of modems all trying to train at the same time can cause interference.

Power cycle at your leisure to see if it makes a difference?

[MikeZ]

Thanks - after a couple of days I replaced my ECI modem with an hg612, so it has had one resync. It has been stable since, just with the increased latency and lower speed. If any knowledgable MDWS users could see how they think it's looking now, I'd be grateful...

[Dray]

Interleaving looks very high, sync looks poor and SNR margin is low. Looks like you have a noisy line.

[ip75]

I'm not an expert on this stuff - others on here are - but my observations are:

1. The SNRM, both upstream and downstream, have been pretty stable, which is good.
2. The number of errored seconds looks to be low, also good.
3. There have been a few bursts of FEC errors, of the order of 100,000. Roughly one or two bursts per day. These are errors that were corrected, so won't have caused a problem - I believe these are basically errors that are effectively corrected by the interleaving on your line. If interleaving were removed, it's possible that whatever caused these FEC errors would cause uncorrected errors on the line. However, I don't really know whether 100,000 FEC errors is a large number in the great scheme of things.

I don't know if DLM takes FEC errors into account. If it doesn't, I would think it will eventually remove interleaving and you can see how the line fares then now that you are monitoring it. However, if it does take FEC errors into account, then I don't know what it will do. Hopefully someone else who has studied this stuff will come along and be more definitive!

[MikeZ]

Interleaving looks very high, sync looks poor and SNR margin is low. Looks like you have a noisy line.

Have still, or had before DLM applied the interleaving? Line was stable and sync was good for months on end. Then there was an exchange fault (that's all BT would say) and it was trying to sync for around 3 hours. When it finally resynced it was fast and stable again for nearly 24 hours until the time of the morning that DLM kicks in based on collected stats from the day before. That's when the interleaving and lower sync was applied.

What I'm trying to determine is whether the current line stats are as a result of the DLM kicking in or whether there is still a problem. As far as I can see, the DLM has kicked in because of all the attempts to resync the day before. As I say, it was fast and stable again for nearly 24 hours until the moment the DLM kicked in.

[MikeZ]

I don't know if DLM takes FEC errors into account. If it doesn't, I would think it will eventually remove interleaving and you can see how the line fares then now that you are monitoring it. However, if it does take FEC errors into account, then I don't know what it will do. Hopefully someone else who has studied this stuff will come along and be more definitive!

Thanks very much for this. I think I've seen on this site that BT's DLM uses ES, not FEC. If that's the case then I'm hoping it will start to reduce interleaving. I also read here that it should start to do this after 9 days if errors have been below the threshold - is this still the case?

[Dray]

I think that as SNR margin is low and sync speed is low, there is currently a noise problem. If DLM had capped the sync speed then the SNR margin would be high, but as it stands it looks like the sync speed is being limited by the available SNR margin because the noise is high.

[ip75]

The noise margin stays close to 6dB, which is the target margin. So it just means that the speed is limited by the line capability rather than a cap. This is normal for lines that are not capped, I think. It's also good that it doesn't deviate much from that margin, as that shows the background noise level is fairly constant.

[Dray]

In comparison with others with similar attenuation, the sync speed is low.

[MikeZ]

The attainable looks about right - it had been stable with no (or very little) interleaving for months and was synced at about 65. Not sure what US was synced at but speed tests constantly gave me around 18 Mb/s.

I'd be happy to get back to near the attainable (but with slightly higher US - I upload to Amazon S3 every night). I'd really like the latency back that I had before, though - that was around 6ms to my ISP and now it's 22ms.

[Dray]

I suppose you may be a victim of whatever Openreach have done to line profiles on ECI cabinets.

[burakkucat]

I suppose you may be a victim of whatever Openreach have done to line profiles on ECI cabinets.

Do we know for sure that Mike is connected via an ECI equipped cabinet?    (I may have missed that detail in my speed-reading of this post.)

[MikeZ]

Do we know for sure that Mike is connected via an ECI equipped cabinet?    (I may have missed that detail in my speed-reading of this post.)

Yup:

ChipSet Vendor Id:      IFTN:0xb204

[burakkucat]

Right, got that. Thanks for the clarification. 

There is one trick that I have recommended before now and that is to disconnect the modem from the incoming line last thing at night. Power off the modem, then disconnect the cable. The next day only reconnect the modem to the line when the broadband service is required. (This should only be performed once, I'm not suggesting that it should be done every night!) By giving the port on the DSLAMs line-card "time to relax" it may clear some of the "peculiarities" currently present.

I know that N*Star has used the technique to clear what looked to be a ongoing oscillation in the SNR.