Print

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

[William Grimsley]

I think the not banding the speed until later may have been overlooked, but with the HH5b, you may not have been able to see all the steps the DLM made.

I did multpile tracerts and the interleaving was not applied until the DLM resync.

[roseway]

The point is that you can't extrapolate your own personal experience to a general conclusion about the way DLM works.

[William Grimsley]

The point is that you can't extrapolate your own personal experience to a general conclusion about the way DLM works.

True, but it doesn't say anywhere in the Kitz documentation about DLM intervening out of normal night time hours for FTTC only ADSL.

[Weaver]

@sevenlayermuddle I re-read Kitz' piece about MSR, the lowest speed achieved during the period. It seems however that my most recent "SNR reset" operation (ie. me hitting the button in A&A's clueless.aa.net.uk server to bring in a new downstream target SNRM) was something other than what I thought it was. The fact that the MSR shows up as a magic number, 2272, is highly suspicious. Firstly, it's magic, secondly it hasn't actually been that low at all, not even close, during the ten days in question. So basically, I don't know what on earth I'm talking about. I'm thinking I've initiated a training phase when possibly I haven't. I would presumably have to get the ISP to do an MSR reset for me and that's all I can say. ISPs presumably don't want to, because an increase in FTR increases their liabilities and therefore their costs.

[roseway]

The point is that you can't extrapolate your own personal experience to a general conclusion about the way DLM works.

True, but it doesn't say anywhere in the Kitz documentation about DLM intervening out of normal night time hours for FTTC only ADSL.

Our knowledge of DLM is still incomplete because it isn't publicly documented. The Kitz documentation is the result of many, many hours of study and observation, and is periodically revised as new information becomes available. But it's not humanly possible for it to keep up with every detail of new information, all the time.

In the case under discussion, Kitz made an observation which is correct, namely, that the FTTC version of DLM does intervene during the daytime. It does, sometimes. You made a firm assertion which is not proved to be correct, based on your personal experience. I challenged you on this because other people may read what you said and believe it to be fact, whereas it's only a proposition.

[William Grimsley]

The point is that you can't extrapolate your own personal experience to a general conclusion about the way DLM works.

True, but it doesn't say anywhere in the Kitz documentation about DLM intervening out of normal night time hours for FTTC only ADSL.

Our knowledge of DLM is still incomplete because it isn't publicly documented. The Kitz documentation is the result of many, many hours of study and observation, and is periodically revised as new information becomes available. But it's not humanly possible for it to keep up with every detail of new information, all the time.

In the case under discussion, Kitz made an observation which is correct, namely, that the FTTC version of DLM does intervene during the daytime. It does, sometimes. You made a firm assertion which is not proved to be correct, based on your personal experience. I challenged you on this because other people may read what you said and believe it to be fact, whereas it's only a proposition.

Fine. Apologies, but I can only tell you information I have to hand from my own experience.

[kitz]

What makes you believe that?

On FTTC, DLM does not work on the fly. It only makes DLM retrains away from night hours only if it's a positive change. DLM will not intervene negatively on the fly like ADSL.

I never said it worked on the fly    It doesnt.    What we have noticed over the past year is that any changes (positive or negative) which previously only happened over night has slowly got later and later in the day for some people. 

Its also untrue that it only makes positive changes away from night hours.   
We've been monitoring lines and DLM changes for years on FTTC. 
There are plenty of cases where positive steps have occurred during night time.    We've also seen many instances of negative changes also occurring during day time.

It could be something to do with now that more people are on FTTC, then there are more changes for the NGA 'RAMBO' to apply.   
The NGA DLM system can only roll out so many at a time.  It doesnt do them all together.     There is now one hell of a lot FTTC DSLAMs in the cabs than there are 21CN MSANS.   This is a plausible explanation why changes for some people are now occurring later in the day.   Its not fact..  I dont have proof, but just a reason why.      Ive found that any positive or negative changes for me usually tend to occur at between about 10 until 1pm.    I cant recall ever getting an overnight change.   Yet there are plenty of people who do.   

[kitz]

The fact that the MSR shows up as a magic number, 2272, is highly suspicious.

2272 is the default.   It is sometimes used if the line for some reason hasnt completed a full FTR period.    I think sometimes things go wrong with this within the 20CN RAMBO.
Many years ago when on adslmax I also found myself with that stupid FTR and had to go through a fresh training phase.    I also used to get a lot of stuck 2000 IP profiles. -   Yet I could dead easy sync at the full 8Mbps. 

[kitz]

The point is that you can't extrapolate your own personal experience to a general conclusion about the way DLM works.

True, but it doesn't say anywhere in the Kitz documentation about DLM intervening out of normal night time hours for FTTC only ADSL.

Our knowledge of DLM is still incomplete because it isn't publicly documented. The Kitz documentation is the result of many, many hours of study and observation, and is periodically revised as new information becomes available. But it's not humanly possible for it to keep up with every detail of new information, all the time.

In the case under discussion, Kitz made an observation which is correct, namely, that the FTTC version of DLM does intervene during the daytime. It does, sometimes. You made a firm assertion which is not proved to be correct, based on your personal experience. I challenged you on this because other people may read what you said and believe it to be fact, whereas it's only a proposition.

Fine. Apologies, but I can only tell you information I have to hand from my own experience.

Im a lot more cautious about what goes on the main site.   It doesnt go on there unless Im pretty certain of the facts.    What is said on the forums is very different.  On the forums we like to theorise.  -  You perhaps saw a convo between me and wombat last week where we were trying to find a relationship between such things as B & N values. 

We cant always go off one persons experience.   If I put up on the main site that the NGA RAMBO only makes changes between 10am and 1pm then that would be false.   Yet it would be entirely true if I applied it to my line and observations on my line only.   See the difference?

[William Grimsley]

We cant always go off one persons experience.   If I put up on the main site that the NGA RAMBO only makes changes between 10am and 1pm then that would be false.   Yet it would be entirely true if I applied it to my line and observations on my line only.   See the difference?

Yes, I see the difference. *Removes blindfold*

[Weaver]

Kitz, the MSR = 2272 thing has now happened on the most recent two occasions. It's just as you say, as if this value were set at the start of the procedure, and if the procedure never completes, then that's the answer you're left stuck with.

[kitz]

I think for some reason 20CN DLM has always been more problematic for some reason.    21CN DLM is certainly more robust and doesn't appear to have the same issues with people being lumped with default figures.

TBH the 20CN DLM was one of the reasons why I was so glad to move to LLU.   If you look back at my post history pre LLU, you would see the amount of times I would suddenly find myself stuck with the default 2Mb profiles.  It would happen if we say had a local power outage,  there was once even when an electrician was working on something in my home and turned off my mains without me powering off the modem.   It got to the point whereby I'd simply contact James at Plusnet and say 'its happened again' and he'd know what I meant without me having to say anything more.  Back then BT didnt like doing a full FTR reset at all, although they were a bit more relaxed to resetting IPprofiles.    I believe BT have supposed to fix some of it, but even Plusnet admit more recently that the default profile thing seldom happens with 21CN, but can still crop up from time to time on 20CN.   

[Weaver]

I just seem to do rather better with DLM these days, but it used to drive me mad. I find that it of course helps being very careful not to switch my modem off, ever, the UPS plays its part in that. And also I definitely take care not to unplug the DSL line from the modem while it's turned on, and that way I seem to keep the DLM gods at bay quite successfully.

I wonder how much in the way of software changes has been applied to DLM for plain old ADSL2 over the past decade?

[aesmith]

DLM doesn't seem to have reacted when I've had a few disconnections (self inflicted) within one day.   These have come about because of our variable noise margin, if I need to shut the router off during the day, it's likely to reconnect at an unsustainable speed, so I tweak the SNR margin until I get just over 4 meg.  Took three attempts last time, even though the margin was 4.5 before disconnecting, I needed to raise the target to 7.5 to get a sensible speed.

What does annoy me is the one-way profile battles - if I connect just once at less than 4meg I get knocked back immediately to the lower profile, then it takes days and days before the profile is reinstated, then more time before Plusnet picks up to the increase, then it needs a disconnect/reconnect to bring the profile into effect.

By the way is that PPP reconnection needed for the BTW profile, or is it just for Plusnet?

[aesmith]

Just completed the reset/relearning via A&A's control panel.   Before the learning process it showed the "magic" default values ...

BT Test xDSL Status Check:Pass OK.pass OK. Circuit In Sync
BRAS=3500Kb/s FTR=1817Kb/s MSR=2272Kb/s ServOpt=1 I/L=A
THIS LINE IS STABLE AND WITHIN ERROR/RETRAIN LIMITS FOR THE CHOSEN SERVICE OPTION (ILQ=A)
Up Sync=448Kb/s LoopLoss=30dB SNR=20dB HIGH ErrSec=0 HECErr=0 Cells=18409
Down Sync=4480Kb/s LoopLoss=54.5dB SNR=4dB ErrSec=3 HECErr=0 Cells=46539

During the learning period these were reset to 288/288 presumably the lowest possible values?

BT Test xDSL Status Check:Pass OK.pass OK. Circuit In Sync
BRAS=3500Kb/s FTR=288Kb/s MSR=288Kb/s ServOpt=1 I/L=A
THIS LINE IS STABLE AND WITHIN ERROR/RETRAIN LIMITS FOR THE CHOSEN SERVICE OPTION (ILQ=A)
Up Sync=448Kb/s LoopLoss=30.5dB SNR=20dB HIGH ErrSec=0 HECErr=0 Cells=205027
Down Sync=4352Kb/s LoopLoss=55dB SNR=4dB ErrSec=2 HECErr=0 Cells=5060518

Then replaced by real values at the end of the process ...

BT Test xDSL Status Check:Pass OK.pass OK. Circuit In Sync
BRAS=3500Kb/s FTR=3200Kb/s MSR=4000Kb/s ServOpt=1 I/L=A
THIS LINE IS STABLE AND WITHIN ERROR/RETRAIN LIMITS FOR THE CHOSEN SERVICE OPTION (ILQ=A)
Up Sync=448Kb/s LoopLoss=30dB SNR=20dB HIGH ErrSec=0 HECErr=0 Cells=7626
Down Sync=4032Kb/s LoopLoss=54.5dB SNR=6dB ErrSec=1 HECErr=0 Cells=7464

Interestingly during that 10 day period when we had some weather the line was disconnected a few times, and reconnected at slower speeds a couple of times, once as low as 2144k.  The learning process seems to have ignored these low speed episodes.

The other question is the stated ILQ, which is given as Amber.  Is from that BT status report, is the ILQ supposed to be the current status, or is it a record of the status recorded at the end of the training period?    It's still showing ILQ=A today even though at least the last three days have been nice and clean and well into the green threshold.