In reply to a post by Chrysalis:

it sounds like FTTC DLM is worse than ADSL DLM in that it constantly ...

I guess it depends on your definition of "constantly".

DLM seems to intervene within 24 hours of the MTBE (or MTBD) going beyond the "red" threshold, and seems to take up to 4 weeks to undo the intervention when MTBE/MTBD holds better than the "green" threshold.

By the strict meaning of the word "constantly", then yes - DLM works forever. By the implied derogatory meaning (where "constantly" implies that changes happen too often), then No - I think the FTTC DLM has pretty serene response times.

If BT set the red & green thresholds sufficiently far apart, the hysteresis should prevent DLM from regularly flip-flopping. However, even if they aren't quite right, a flip-flop period of 28 days isn't too bad (though a 27:1 mark:space ratio isn't good).

Note: I don't know the thresholds for FTTC, but I saw something suggesting that for ADSL2+ on a "standard" profile (not the "stable" or "superstable" profiles) the red threshold was an MTBE of 60 seconds, and the green was an MTBE of 600 seconds.

In that setup, DLM would leave the line alone with a 24-hour error count of between 144 and 1440.

In reply to a post by WWWombat:

In that setup, DLM would leave the line alone with a 24-hour error count of between 144 and 1440.

FWIW, I'm currently remotely monitoring a new VDSL2 (FTTC) connection (Interleaving is still OFF - for now) where Error Seconds in just 12hours have exceeded 13000.

Sync speeds are 11397kbps DS & 563kbps US (yes, only 563k US).

I imagine Interleaving will be turned ON at quite a high depth by tomorrow afternoon.

When I've finished monitoring this connection, I'll be watching another where VDSL2 DS sync speed is currently around 4Mb (usually lower than 3Mb).

This is NOT an Infinity connection & speeds were not expected to be "superfast".

In reply to a post by Bald_Eagle1:

FWIW, I'm currently remotely monitoring a new VDSL2 (FTTC) connection (Interleaving is still OFF - for now) where Error Seconds in just 12hours have exceeded 13000.

Sync speeds are 11397kbps DS & 563kbps US (yes, only 563k US).

I imagine Interleaving will be turned ON at quite a high depth by tomorrow afternoon.

Yup, I imagine so.

I guess the theresholds need to be scaled by the relative number of CRC blocks being transmitted.

In ADSL, a superframe takes 17ms or so, which implies 58 CRC codes per second.

I don't know about VDSL2, or even if there is a fixed rate, but I guess you get to see the OHF rate by monitoring the stats.

yes, but I had never had DLM force a resync, it would only change settings if the line dropped for another reason.

You understand the difference? also am I correct that FTTC DLM forces a resync to change settings or is it like BTw adsl where it only works when the line drops for other reasons?

Is there a variant of SRA for VDSL?

SRA was amazing for my adsl experience. That adapts but the difference is it doesnt interrupt connectivity, it also doesnt toggle interleaving just works on noise margin. According to this doc is also a RRA feature.

http://www.embedded.com/design/connectivity/4007197/...

Edited by Chrysalis (Mon 26-Nov-12 03:02:56)

ok so there is a threshold where its neither too bad or too good.

To give you an idea, my adsl line had 1000s of crc errors a day, this was normal for the line, I did experiment with the SNRM not long before cancelling ukonline (where I had full control of the noise margin) and to get my error rates down to what others typically got maybe less than 100 errors a day I had to goto a noise margin in the high 20s, and had a pretty low sync speed. With the 1000s of crc errors tho the line was still very useable, it had issues when I got 1000s a minute tho.

So the DLM setup you described would have probably bumped the line settings way too conservative, and indeed prior to switching to LLU I had issues of over conservative line settings combined with instability.

I have read there is speed banding now, although I have not heard of any isp using it. Is this a way for DLM to be overriden and a line locked to a speed?

Edited by Chrysalis (Mon 26-Nov-12 03:15:08)

In reply to a post by Chrysalis:

Is there a variant of SRA for VDSL?

I'm pretty sure the equivalent of SRA is built in to the VDSL specification as I'm sure I've had resyncs but you just don't notice them. Not having a static IP, I've not set up a BQM.

Compared to ADSL/ADSL2+, its a non issue.

Edited by jchamier (Mon 26-Nov-12 08:13:45)

In reply to a post by Chrysalis:

To give you an idea, my adsl line had 1000s of crc errors a day, this was normal for the line, I did experiment with the SNRM not long before cancelling ukonline (where I had full control of the noise margin) and to get my error rates down to what others typically got maybe less than 100 errors a day I had to goto a noise margin in the high 20s, and had a pretty low sync speed.

My (very) old ADSL line (20CN) had a DLM-event very much like the system I just described. Normally it would have almost no errors (10 CRC per hour), but one day got 1,000's in 3 large bursts... and next day had slowed down (via an automatic resync). Some days later, it recovered.

Our first FTTC line suffered 4% packet loss at initial connection due to the error rate (though the modems were locked at the time) DLM acted after 48 hours there, adding interleaving & slowing things down, with the change requiring a resync. The line then stayed like that for months... until the 17a profile was introduced

Our current FTTC line works perfectly at full speed, but the phone line had problems one day in August. Again, DLM reacted (next day) by slowing things down, requiring a resync, and took 3.5 weeks to restore full capability.

Every single one of these changes was made with a visible resync. I suspect that all of them *needed* a resync - I don't think SRA or RRA can cope with turning interleaving on & off at will.

With the 1000s of crc errors tho the line was still very useable, it had issues when I got 1000s a minute tho.

My FTTC line with a 4% packet loss was very useable - but lost around 10% of the ultimate throughput available to it. However, I've seen a line where around 10% loss made it totally unusable.

On ADSL lines, there are 58 CRC checks per second, so it's possible to get up to 3500 per minute... but I'd expect it to be not very useable if you were running at over 350 per minute.

So the DLM setup you described would have probably bumped the line settings way too conservative, and indeed prior to switching to LLU I had issues of over conservative line settings combined with instability.

Yes, it probably would have done. BT's DLM definitely focusses on stability over speed.

I have read there is speed banding now, although I have not heard of any isp using it. Is this a way for DLM to be overriden and a line locked to a speed?

My experience, watching Plusnet handle other subscribers' faults, is that BT employ a number of DLM profiles... which set interleaving to low & high levels, and which *may* also set banded speeds.

This SIN mentions banding of speeds too (section 2.2.1)

The banding seems to be purely a DLM feature, and nothing that an ISP can affect, or select, or even reset.

This Openreach description shows that ISPs can select 1 of 3 settings for DLM: Standard, Stable or Speed. I've never seen an ISP offer to change this setting for a user, so I'm not sure how widespread this setting is but note that this is an *Openreach* setting.

This copy of a 'BT Wholesale' description has slightly different information (section 11). It mentions a Standard, Stable and SuperStable option choice. Again, not seen much at the ISP.

That document (section 13) mentions that DLM policies affect Interleaving, Capping and Impulse Noise Protection, and that the noise margin is set to 6dB for all policies.

But most importantly, it states that these policies are adjusted "to meet target stability".

So it looks like the only thing that an ISP could set is to specify a stability level - but not by directly locking the line to a specific speed.

Note: On 21CN ADSL2 systems, there is a similar setting that BT talks about (the options are Standard, Stable and SuperStable), but few ISPs mention. The information I had shows that the different settings here are in the MTBE thresholds - where Standard has the 60/600 thresholds, the Stable has 600/6000 thresholds, and SuperStable has 6000/60000 thresholds.

FTTC DLM appears to force resyncs but it only takes 20-25 secs to resync and most of the time they occur in the early hours of the morning.

sorry I meant I still had 0% packetloss and full throughput.

usually my line either worked fully or completely collapsed, I cant remember having situations with 10% lost throughput.