Print

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

[sevenlayermuddle]

As mentioned in another thread, I recently got stuffed by a 15dB target caused by (I suspect and hope) a duff router that I'm no longer using.  Past experience with DLM suggested I held little hope of the margin coming back down again anytime soon.  Last time it stayed at 15dB for a whole year, before I tried to 'trick' it into action.  I've just repeated the same little trick...

Exactly three weeks ago I tweaked the margin up to circa 18dB.   That resulted in an ES rate of about 5 to 10 a day, but somedays I was seeing maybe 20 ES, and my hunch was that wouldn't be good enough.  A week later (two weeks ago) I tweaked it up further to  circa 22dB, which brought the error rate down to about 5 ES per day on average.

This morning, guess what...?  Yep, target margin's been reduced to 12dB.  

I know, I can't exclude the possibility that it was pure co-incidence, and the target would have come down without my trickery.  Personally my hunch is that it wouldn't but, understandably, I suspect not everybody will be convinced.   Anyway, I'll tweak it back up again, and see what happens next.....

- 7LM

[waltergmw]

@ 7LM,

VERY WELL DONE INDEED !

With your continual perseverance you might just have found a solution to the problem - that is until the DLM logic is changed again!

Kind regards,
Walter

[roseway]

Yes, I'm really quite impressed with your dedication, 7LM. There aren't many people who would voluntarily tweak their noise margin up, and it does very much look as though it pays off.

[BritBrat]

Could you explain how you did it.

Thanks.

[sevenlayermuddle]

Could you explain how you did it.

Thanks.

I currently have a DG834GT, running standard Netgear firmware.  The process for tweaking target margin is described here http://www.kitz.co.uk/routers/dg834GT_targetsnr.htm

There's probably other ways, such as using the GUIs in Routerstats etc, but I'm more at home with a command-line (something to do with my age!).   In my case, adslctl configure --snr190 resulted in an actual margin of about 21dB immediately after reconnecting (at night), drifting up to 22dB in the daytime.

Obviously I can't guarantee it'll work for anybody else, try it at your own risk!

And you do have to put up with life in the s-l-o-w lane during the experiment, my connection speed was 1792 at a margin of 22dB, vs upwards of 2700 with the 15dB margin.

[jeffbb]

Hi
Lucky you I  had very low errors for months ,< 2 CRCs and ~1 ES /24hr(tweaked to 9db from 15 db) . Target SNR did not drop from 15db.

Current situation Had some bad weather a few days ago had about 250 CRCs over a period of about  6Hrs ,since then back to normal low errors .

Regards Jeff

[sevenlayermuddle]

Jeff,

Sorry you've not been so lucky

A possible explanation for the difference is that DLM maybe keeps an eye on the actual SNR margin and, to qualify for a target reduction, expects it to be 'stable', in addition to low error rates.  By 'stable', I mean that  the actual SNRM reported by the router should rarely be seen to dip below the target set by DLM.  If DLM sees an SNMR that's been tweaked downwards, so it's always below the target DLM has set, it may decide the stability criteria have not been met, regardless of error rates.

In contrast, a target that's tweaked upwards may lead DLM to believe that the SNRM is highly 'stable', in as much as always well in excess of the target.  That, combined with good error rates, may possibly help?

But I've lost count of how the number of 'maybes' in above.  It's just a theory, and maybe I'm just lucky. Maybe (second 'maybe' this paragraph) l should buy a lottery ticket this week, just in case that's the case ?  

- 7LM

[jeffbb]

Hi
Good luck with lottery 

Regards Jeff

[sevenlayermuddle]

@Jeff,

You're obviously unconvinced.

But why not put it to the test and try it for yourself?   From the sound of things, all you have to lose is a few weeks (give it three weeks)  at a low rate, plus the time for the profile to recover if it doesn't work?

- 7LM

[jeffbb]

Hi

Might well do . I am still trying to read up more on how the DLM works . As you say it may look at the current snr margin and compare it to Target SNR margin . However if this were so there would have to be some Very clever programming that allows for "normal"  drop in SNR margin during the connection time  as the  ADSL connection get noisier during certain times of the day. The SNR margin is negotiated when first connecting to the exchange then by definition most times the actual SNR margin will probably be lower than the target  especially during "busy" periods  . Most people will also optimise  their synch rate (regardless of the target )by synching when the line noise is at its lowest.This of course means that the current SNR margin is almost bound to be lower than the Target SNR margin  . That been the case the DLM would never see the line at the target SNR margin setting.

Regards Jeff

[geep]

Hi,
As well as the "intelligence" in the DLM program, I wonder if there may also be some "intelligence" (or stupidity) in the router which is aiding (or fighting) with the DLM.

I have observed that my 2Wire BT27000HGV sometimes retrains (= resync?) in the evening as the SN Margin drops. It starts out happily at 9dB and about 6.8 Mbps sync in the afternoon. Evening comes and SN Margin drops. CRC errors increase so it retrains. Again at 9dB, but now syncing at about 7.5 Mbps. CRC rate is horrendous, but it doesn't seem to care. Then in the afternoon it decides to retrain again as it thinks the noise margin is too low - seemingly at about 6dB - so we're back again at 6.8 Mbps & 9dB and the cycle restarts.

But this cycle isn't consistent. Sometimes it does it for a few days, and other times it never retrains at all over several days.

My ST546 doesn't do this - it just maintains synch, and I watch the noise margin go from 9 to 6 in the evening, and back to 9 in the day. Maybe it's less "intelligent" than the 2Wire, and a tad slower too.

Cheers,
Peter

[jeffbb]

Hi
@Geep
quote : I have observed that my 2Wire BT27000HGV sometimes retrains (= resync?) in the evening as the SN Margin drops. It starts out happily at 9dB and about 6.8 Mbps sync in the afternoon. Evening comes and SN Margin drops. CRC errors increase so it retrains. Again at 9dB, but now syncing at about 7.5 Mbps.

I have not used any routers that have this facility(thank goodness).
The idea seems good so that you maintain a stable connection. It does seem that its not achieving that. You would expect that if it retrains because the SNR margin is too low then resynchronise at target  then the synch rate should drop thereby giving a more stable connection. How it can both increase the SNR margin and synch rate is difficult to say. Must be changing bit loading ,reducing the number Spare bits for bit swapping  .This would probably cause more errors I guess.

Is it not possible to stop this function ? as it seems to be counter productive .

Regards Jeff

[sevenlayermuddle]

@Jeff,

I seem to remember seeing, in one of the BT papers or patents, a statement, vague and all-encompassing, but suggesting that all sorts of errors and stats, as well as just reconnects, are included in DLM's stability assessment.

I've been wondering whether maybe (there, I said it again!) DLM may also include corrected FECs in it's equations?   Corrected FECs don't affect the data, don't need retransmissions, and so have no impact on user experience or perception of stability.  But a high number of FECs may still indicate that corruption is occurring, regardless of whether it's being corrected by FEC.  The authors of the DLM algorithm could have decided high FEC counts should not occur on a 'very stable line', and so take them into account before reducing target.  I've always tended to ignore FECs in my own line stats, but I bet there's a lot fewer FECs with a massive margin than with a normal margin, and that could be another reason DLM apparently responds favourably?

Just a(nother) thought.

- 7LM

[geep]

@Jeff

There's no control over anything with the 2700HGV.
One just looks on in awe via its comprehensive monitoring facilities and wonders "why the hell did it do that?"

Cheers,
Peter

[jeffbb]

Hi
@7LM quote I seem to remember seeing, in one of the BT papers or patents, a statement, vague and all-encompassing, but suggesting that all sorts of errors and stats, as well as just reconnects, are included in DLM's stability assessment.

Yep CRCs ES etc are all part of it . One other important item is the doubler counter . This is the counter that increments by 1 up to a max of 5 when a line has a problem after having a less restrictive Target SNR margin,causing it to be increased again . The max for this counter is 5 which equates to about 14 months before a reduction in the Target will be available

quote from  BT pattern European pattern application
Furthermore, a delay doubler is used to increase the delay
(i.e. by increasing the good threshold) required before a line which has moved down from a more aggressive profile to
a less aggressive profile level is allowed to re-
transition back up to the more aggressive level. The delay doubler is
therefore incremented (in the present embodiment up to a maximum of 5) whenever the line is re-
profiled to a less
aggressive level and then the delays are reset (as in the case where the line is re-
profiled to a more aggressive level)

### each time the line’s profile is transitioned to a less aggressive profile the
DELAY DOUBLER is incremented until after 5 such transitions, each time the DELAY is reset it is reset to a value of
448 (i.e. equivalent to approx 14 months). In the present embodiment, if a user’s stability policy or level is changed the
delay doubler is reset back to zero; furthermore, the delay doubler and even the delay counter may be manually reset
by an operator to cater for exceptional circumstances.

Basically it can take up to 14 months to improve Target SNR margin if there has been some Up and Down resetting of Of the target SNR margin during some previous period .

Regards Jeff