Kitz Forum
Broadband Related => Broadband Technology => Topic started by: NewtronStar on November 07, 2014, 08:42:50 PM
-
It has been a long journey for me to find why my line has never been moved to fastpath on the DownStream side it's nothing to do with Errored Seconds or CRC's or a high SNRM it looks like it's down to FEC counts and I have said in past FEC (forward error correction) is a good guide to see the level of noise on the end users line.
If the line is to noisy it would be stupid for the DLM to move you onto none-interleaved until the FEC count gets to a low level of 0 - 25 per minute over a 14 day period this not going to happen on my line as it's well above that level more like 300 - 1200+ per min.
It's just my theory into why some End Users get fastpath and others never do.
http://en.wikipedia.org/wiki/Forward_error_correction (http://en.wikipedia.org/wiki/Forward_error_correction)
-
Hmm . . . An interesting theory based on your own observations. I wonder if others can supply evidence that agrees with you proposal? :-\
-
Sadly newt your daily ES is simply too high. It has to be extremely low to go down a profile.
Not saying FEC gets ignored, I have no idea, but I am pretty sure your ES per day is too high.
-
Sadly newt your daily ES is simply too high. It has to be extremely low to go down a profile.
Not saying FEC gets ignored, I have no idea, but I am pretty sure your ES per day is too high.
My errored seconds are 4 X lower than BaldEagle1 ;)
-
Hmm . . . An interesting theory based on your own observations. I wonder if others can supply evidence that agrees with you proposal? :-\
I hope so B*CAT after 3 years with FTTC and still the DS is interleaved it must be the level of noise on my line that is stopping the DLM moving it onto fastpath it's the only reason that I can see, errored seconds look fine crc's quite good and the snrm looks ok for the line sync with an undulating attainable over 24 hours.
Fastpath on the downstream is not my holy grail it's more to do with the why this line is not getting the none interleaved option :)
-
Hmm . . . An interesting theory based on your own observations. I wonder if others can supply evidence that agrees with you proposal? :-\
Zen's FTTC training document mentions that DLM takes FEC into account, this looks like it is true.
-
Hmm . . . An interesting theory based on your own observations. I wonder if others can supply evidence that agrees with you proposal? :-\
Zen's FTTC training document mentions that DLM takes FEC into account, this looks like it is true.
Would love to see Zens FTTC training document :) did see your thread and the download link but the format of the Doc was unusable with the software I have installed on the PC for example don't have MS office only MS Works :-[
-
Point me to the link and I'm fairly sure I could convert it into a PDF file. :)
-
Sadly newt your daily ES is simply too high. It has to be extremely low to go down a profile.
Not saying FEC gets ignored, I have no idea, but I am pretty sure your ES per day is too high.
My errored seconds are 4 X lower than BaldEagle1 ;)
the threshold to stay on fast path is not the same as the threshold to return to fast path. The latter is much lower, and I think it is 10x lower.
-
My errored seconds are 4 X lower than BaldEagle1 ;)
Thats an unfair comparison to make. You cant compare the number of Errored Seconds on an Interleaved line, to one that isnt Interleaved.
Theres more than one algorithm used to get an interleaved line back to non-interleaved.
BaldEagle is still below the limit for Interleaving to be applied.
Your line isnt below the limit for Interleaving to be removed.
The other algorithm is the length of time before the DLM will act on removing interleaving, and this is based on line history. Its why after my recent bout with DLM, it only took one day of stability before it took me back down a notch again.. yet on some other lines it will take months. It takes line history in to account to stop DLM flapping between interleaved and non -interleaved.
BE hasnt been Interleaved since the 26th of Aug. If you look at his Err Secs before that they are much less than they are now.
-
My errored seconds are 4 X lower than BaldEagle1 ;)
Mine tend to be around 8 to 9 times lower when interleaved.
-
When I was last interleaved I used to get over 1500 CRC a day on fast path (cannot remember ES sorry).
When I got interleaved my daily CRC literally dropped to single figures if not 0. It was extremely low. Which obviously meant my ES was also extremely low.
If Openreach allowed a line back onto fast path which was getting say 100ES a day whilst interleaved, that line will very likely be bouncing back to interleaving after a single day.
Try checking ronski's line the day before he got returned to fast path, that may be a fair indication of what is needed.
-
Newt try capping your downstream to 27000 or even 25000 that might make enough impact to get DLM to act.
-
Just wondering if I should cap my line speed. Still stuck on 49M down, but the upstream did go up from 15M to 20M after a short time. What do you think, worth trying or should I accept 49M is the best I can get with a downstream SNRM of around 15dB to 16dB? I'm guessing this probably only works best with interleaved lines though, it probably won't help my banding prediciment.
-
every time I have been synced with even 1 extra db snrm, I found a significant change in error count. Its amazing how effective it is. Worth trying, but do not cap it too much, because it may band you lower. The issue you have is banding not interleaving, I think if you not syncing at the same speed as the upper limit of the band then it wont be removed.
-
I Have been reading this, I too want to get back to FP, I had some issues with noise up to about a month ago, but that's gone. Do we actually know what ES value is need to persude the DLM to try FP again, mine are less than 20 / day
-
What is your FEC count per day?
Like NS i believe the DLM takes that into account. Mine for yesterday was 252312 so not much hope for me.
-
955328 ! So on that basis not for me either.
However I'm not sure how this works as the FEC's are as a result of the error checking. No Interleave and then no FEC (I think !)
Its a black art this BTOR DLM :(
-
Its a black art this BTOR DLM :(
Same as that.
As you say FEC’s are corrected errors, so without some level of interleaving I guess they would be CRC’s.
Like me you need to track down where the little buggers are coming from or how to prevent them – either task will probably be impossible.
-
Im not convinced about FECs. Everything Ive seen implies only one parameter for Errors recorded in the Element Manager & Data Collectors. Ive been doing a hell of lot of reading recently on the DLM on anything I can get my mitts on that comes from BT. Im trying to put it all together as Ive got tons of stuff its just taking me time to do so as the whole thing will encompass several pages on the site, because it really is that massive of a subject and far more involved that it first appears.
Nor would the inclusion of FECs work for MTBE calculation. However there are occasions when if a line is performing very badly then RAMBO will directly monitor the line - in those instances it will look at other parameters not normally used by the normal day to day monitoring of most lines. For those lines RAMBo for sure looks at SNRm variances - which it wouldnt normally do. Most of the monitoring for most lines is the responsibility of the Element Manager and its Data Collectors.
The only algorithm that I havent got my mitts on is the up to date figures for MTBE and MTBR. Ive come across several but not sure which is correct. MTBE < 30 seems to be the one that Ive seen that is most appropriate and its the one I used for the DLM calculator.
-
Thanks for the interesting comments. Im sure a whole gang would be interested in the full info....Something tells me its not going to be easy to get. If we are to believe the ISP's they say they don't know either, although I'm not sure I believe that ;)
-
Thanks all for the advice if i was gonna get move onto fastpath it should of happened years ago, if it's down to my 180+/- errored seconds per day then it's look like fastpath it is out of my reach, and banding my line seems such a loss of sync with a small attainable of 32000 - 35000 kbps
:(
-
Hi Kitz
Since being interleaved (2 months) I’ve seen between 100 & 150 US ES per day upstream and 0 &2 DS ES per day, but FEC’s are high and I’m sure it was high CRC’s that got me interleaved. So far even though the ES are low the DLM isn’t letting go of me and I believe (possibly wrongly) it’s because of the FEC’s.
-
The problem may be banding related. Ive seen something that implies the DLM intervenes it goes something like INP -> banding, but when it drops back down again it goes banding -> INP. I havent got all the stuff to hand, but Im pretty sure it mentioned about how lines would drop through banding before removal of interleaving. That however is one area I havent gone into in depth yet.
-
Hmm . . . An interesting theory based on your own observations. I wonder if others can supply evidence that agrees with you proposal? :-\
Zen's FTTC training document mentions that DLM takes FEC into account, this looks like it is true.
The DLM info is immense, (as Kitz alludes to herself). I can clarify that FEC's are taken into account, however I can't find info on the threshold values ??
The bulk performance metrics DLM uses, appears to have 39 values reported every 15mins (such as FEC, Errors, attenuation, Traffic Count) ............. and 26 values reported every 6hrs (such as bits-per-tone, SNR-per-tone, INP).
-
Thanks, I guess that means we can never work it out !
My only issue with the whole system, that actually provides a stable service, is if you get a blip, you get correction/banding applied almost within 24hrs, but it can take forever because of the caution counter to get it back.
Really as and end user we should be able to chose between speed and reliability, I know ISP's have the 3 levels but, I could live with an opened port and mange with the odd disconnection. The latency is an issue that affects my VPN's, but I guess BT would say its not an SLA line so buy one :-[
-
every time I have been synced with even 1 extra db snrm, I found a significant change in error count. Its amazing how effective it is. Worth trying, but do not cap it too much, because it may band you lower. The issue you have is banding not interleaving, I think if you not syncing at the same speed as the upper limit of the band then it wont be removed.
Yeah :(. Trouble is I am syncing at the upper level of the banding, I guess I'll just have to be patient and see if the next couple of months it goes up, or downgrade to 40/10 at the end of my contract (despite attainable being much much higher).
-
The DLM info is immense, (as Kitz alludes to herself). I can clarify that FEC's are taken into account, however I can't find info on the threshold values ??
The bulk performance metrics DLM uses, appears to have 39 values reported every 15mins (such as FEC, Errors, attenuation, Traffic Count) ............. and 26 values reported every 6hrs (such as bits-per-tone, SNR-per-tone, INP).
Before I start.. no Im not shooting the messenger. I very much value your input BS. :)
However, that seems weird, why would the DLM function part of RAMBo need all those parameters such as atten and traffic count etc.
Everything Ive seen just says errors and retrains are used by the element managers for the DLM and sync speeds for RAP. It also records uptime in seconds for each 15 min bin.
The DLM algorithms are supposed to be
MTBR = Uptime / Retrain Count
MTBE = Uptime / Errored & Severely Errored Second Count
which are normalised over a 24 hr period.
If a line is classified as performing very poor (& mention of 10/20 retrains within a short time) , then the RAMBO will take over and perform direct monitoring of the line with the DSLAM (ie cutting out the usual process) and monitor other parameters such as SNRm variance.
---
PS also
Changing these parameters is based on two performance metrics, errors (in particular, in this embodiment, errors caused by code-violations) and re-trains (i.e. re-syncs).
FEC isnt a code violation - those are CRCs/Err Secs/SErr Secs :-\
-
Hey, no worries Kitz ...... as you know I'm not that up on DLM/Rambo and really can't be bothered pouring over the thousands of documents surrounding it. But, I did go back to the 'Metrics document' I alluded to above, and as usual ..... you are right.
Hidden away is 'Mandatory' and 'Optional' reporting values the six optional are ....... ATU-R Vendor ID, Upstream traffic count (TBD), Downstream traffic count (TBD), Power management state, Upstream Actual Aggregate Transmit Power, Downstream Actual Aggregate Transmit Power ........... the rest, (FEC's included), are mandatory reports. :)
-
lol.. no worries. Ive a lot going on right now, and there have been many nights when Ive been unable to sleep despite being totally knackered, so reading material in the hope of sending me off has all been DLM related.
I've also just shelled out and send away for something written by one of guys who works for BT about their access system which looks like it may be interesting. - Im getting quite the collection now of books on my shelf about broadband, which has been noted and commented on before by the more observant of BToR guys that have visited when Ive had problems. :D
Maybe if I can make some sense out of the all the info Ive got so far and put it into the usual more readable format, then it may be of use to others. I didnt think when I started just how big a topic this would be though... so far Ive managed to dilute RAMBo and the DLM system into one page that is almost ready. I also like to do something in graphical format which helps 'at a glance' and thats practically done too.
The full DLM algorithm though.. thats for another page.. as will be RAP - despite me already having an IP profile page. ???
-
You, lady ...... are an inspiration. Cheers for all the effort you put in. ;D :graduate: ;D
-
PS.. If any of you BT guys ever happen to spot RAMBo in the exchange I'd love to see a photo. From the descriptions it simply sounds like it may look like something similar to this
-
FFS ...... I've been trying to track this geezer down !! Now I know what I'm looking for, I shall try. :)
-
;D