cancel
Showing results for 
Search instead for 
Did you mean: 

IP Profile - 88.2% method

JayG
Pro
Posts: 1,145
Thanks: 143
Fixes: 6
Registered: ‎30-10-2011

Re: IP Profile - 88.2% method

Quote from: nozzer
My line is averaging around 54000 CRC errors daily and the DLM selected fastpath after a SNR reset, and it stays there no problem at all! Interleaving is set to "auto" and is currently switched off.

Almost exactly the same error rate as mine, with the same outcome from the DLM!
I think there are other weird and wonderful things happening with the DLM which we are not privy to - on the few occasions I have disconnected via the GUI and then rebooted to try and get above the "magic" 10Mbps sync the SNRM jumps up to about 9.5dB (whether I've improved or worsened my sync speed!) and then drops slowly over the next few days to the "normal" 7.5dB (daytime) and around 5dB (evening peak.)
I assume this is the DLM lowering the output power to minimise interference at the exchange rather than deciding I can achieve a reliable higher sync...................?  Roll_eyes
nozzer
Hero
Posts: 3,298
Thanks: 676
Fixes: 3
Registered: ‎04-08-2009

Re: IP Profile - 88.2% method

Found this...  see attached.
I think this means that if the mean time between retrains is, say, 33600 seconds (for a "stable option" line), then the DLM will hold the initial target SNRM of 6dB. If the Mean time between errors is better than 6000 seconds, the DLM will reduce the SNRM to the min target of 3dB, otherwise it will stay at 6dB.. In other words, retrains come first, then errors.
A higher retrain or error rate will cause the DLM to choose another setting of SNRM to test the retrain rate, and then the error rate. This would explain why my line, which has an apparent high error rate, but which never drops out, is stable at 6dB.
That's my theory any road up!   Smiley
WWWombat
Grafter
Posts: 1,412
Thanks: 4
Registered: ‎29-01-2009

Re: IP Profile - 88.2% method

An interesting find!
I guess a red threshold is one where, if worse than, DLM downgrades you, while if better than the green threshold, then it upgrades.
I wonder what timescale the 'mean' is calculated over.
Plusnet Customer
Using FTTC since 2011. Currently on 80/20 Unlimited Fibre Extra.
jojopillo
Plusnet Alumni (retired)
Plusnet Alumni (retired)
Posts: 9,786
Registered: ‎16-06-2010

Re: IP Profile - 88.2% method

HI itsme,
I have raised your profile to 4200 to allow for your new BT IP profile. This should take effect shortly.
Jojo Smiley
itsme
Grafter
Posts: 5,924
Thanks: 3
Registered: ‎07-04-2007

Re: IP Profile - 88.2% method

Thanks but I have been experiencing higher speeds in excess of 4mbps as soon as the banding was removed even though PN profile was at 2mbps. Strange.
nozzer
Hero
Posts: 3,298
Thanks: 676
Fixes: 3
Registered: ‎04-08-2009

Re: IP Profile - 88.2% method

I've noticed my speeds creep up recently even though the PN profile hasn't changed. Is it actually doing anything I wonder?
jojopillo
Plusnet Alumni (retired)
Plusnet Alumni (retired)
Posts: 9,786
Registered: ‎16-06-2010

Re: IP Profile - 88.2% method

Hi itsme,
I've heard people report that before, but it's doesn't last long.
@nozzer,
Your profile is correct at 10500 for a sync rate of 12012.
Jojo Smiley
WWWombat
Grafter
Posts: 1,412
Thanks: 4
Registered: ‎29-01-2009

Re: IP Profile - 88.2% method

@nozzer's DLM threshold image...
I just used some of the terms on that image in Google to see if I could see any descriptions of DLM.
And up popped this patent to a few guys in Suffolk (ie BT's research labs). If you can do the verbal equivalent of walking through treacle, you might find it interesting. It is dated 2011, but from the references inside, I suspect it might just go back to 2006.
Happy reading  Wink
Plusnet Customer
Using FTTC since 2011. Currently on 80/20 Unlimited Fibre Extra.
itsme
Grafter
Posts: 5,924
Thanks: 3
Registered: ‎07-04-2007

Re: IP Profile - 88.2% method

Just because there is a patent does not mean it is being used. I have been named on 3 patents that to may knowledge are not used in a final product.
WWWombat
Grafter
Posts: 1,412
Thanks: 4
Registered: ‎29-01-2009

Re: IP Profile - 88.2% method

Absolutely. Or it is being used in a different form where the new patent is still working its way through the system. Many options...
On the other hand, where we have no public information about what DLM does under the hood, just a lot of informed (and uninformed) guesswork, this might now provide a framework. Like any work of science or engineering, you see if the theory fits the observations...
Of interest to some is the fact that DLM must pretty much be a software algorithm, employed in a processor somewhere. But... I could have sworn that the UK patent office didn't accept software patents! I'll have to read it more thoroughly.
Plusnet Customer
Using FTTC since 2011. Currently on 80/20 Unlimited Fibre Extra.
itsme
Grafter
Posts: 5,924
Thanks: 3
Registered: ‎07-04-2007

Re: IP Profile - 88.2% method

Algorithms and methods to collect information are patentable. Some of which should not be in opinion. One of the patents held by a large pharmaceutical company was storing configuration and algorithms variables in an external E2PROM. This information for years was stored in an internal E2PROM but when the price fell so it was viable to supply it as part of the consumables which had benefits for the user as well as the manufacturer it was limited by the existing patent. To me the patent should not be allowed as it was not ground breaking, or did not cost money to develop. It just stop competition.
The company I worked for were taken to court for infringing this patent as well as others. I know that it did not infringe it as the last design changed that I did before launch was to transfer the variables that did infringe from the external to the internal E2PROM.
   
itsme
Grafter
Posts: 5,924
Thanks: 3
Registered: ‎07-04-2007

Re: IP Profile - 88.2% method

Just an update. Since Jojo removed the banding I have been in constant sync with the exchange a few hours short of 7 days. The banding was removed approximately at 16:00 which was fortunate as this is the time period in the middle of the SNRM swing and gave me a sync rate of 4777kbps. I see a swing between 3dB to 9dB which give me a maximum sync rate as reported by the modem of 4200 to 5700kbps. So my conclusion about banding is not to stabilise the line but to stop the user resyncing to gain higher speeds. Which is what I could do now as I would gain an extra 1mbps but what's the point if I then lose sync at night. It would become a constant cycle of me forcing a resync during the day for it modem losing it at night.
jojopillo
Plusnet Alumni (retired)
Plusnet Alumni (retired)
Posts: 9,786
Registered: ‎16-06-2010

Re: IP Profile - 88.2% method

Hi itsme,
I have seen lines that have been much more stable with banding applied. It may be that the problems which caused your line to be banded in the first place are not evident at the moment; merely speculation though Wink
Jojo Smiley
IKnowNothing
Grafter
Posts: 27
Registered: ‎10-01-2012

Re: IP Profile - 88.2% method

Quote from: IKnowNothing
Quote from: itsme
Has anyone had the banding raised or removed automatically by the DLM?

Not yet.

So, just to drag this back up for info for anyone interested. Now had my banded profile either removed or raised automatically about 1 and a half months later. Router just resynched by itself at 4536 with an SNR margin of 10. If I remember correctly I think that's probably the next band up from what I was on before so raised rather than removed completely is my guess at the moment.
jojopillo
Plusnet Alumni (retired)
Plusnet Alumni (retired)
Posts: 9,786
Registered: ‎16-06-2010

Re: IP Profile - 88.2% method

Hi IKnowNothing,
Yes, it has moved up to the next band. Your profile looks like there's no interleaving on, so I would imagine if it lifted completely there'd be a danger of it becoming intermittent, which would probably put interleaving on. If you were in the test socket, having changed filters it may be worth lifting it. I can do that if you'd like to try it?
Jojo Smiley