Print

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

[AArdvark]

Sorry to ask but I am very confused what is happening to my line.

I have been syncing at 79987 Kbps since I got the line installed, apart for a brief DLM event when I first installed my Zyxel Router.
A few days ago I rebooted the router as the SNRM Ds had been getting lower and lower for at least the last 90 days.
I also had the situation where the sync was higher than the Max Attainable and SNRM Ds was dropping to circa 3.5db.

(Note: Crosstalk etc has been hitting my line for months and the Max Attainable 'headroom' dropping like a stone)

I expected the router would resync at some point so I decided the re-boot and see what I got.

Now I have the router re-syncing regularly.
The Sync is now circa 62000 kbps with Max Attainable circa 77000 kbps & Interleave depth: 1329   421    +  INP: 3.00      4.00

Looks like DLM is hitting me () BUT the regular re-syncs are a puzzle as they are not changing the overall sync rate etc.
MyDSLWebStats is indicating 4 re-syncs in 3 days (12 -14 Feb) while DSLstats v5.3 is indicating at 3+ resyncs (3 between 21:35 14/02/2015 & 01:47 15/02/2015 alone) plus others I noted before having to restart DSLstats.

Is my line in some fatal nosedive to god knows what speed ?

I have attached the STATS logs from DSLstats v5.3 for 1st - 15th Feb 2015 if it helps.
(There are some gaps due to re-booting PC's and forgetting to re-start DSLstats  )

Thanks for your help.

[burakkucat]

Very quickly, as the warm and sleepy spot is calling me, N*Star had a sort of "oscillatory" effect visible in the graph of one of his set of statistics. I suggested that he try my old trick of allowing the line port to thoroughly "relax" by powering down the modem (a dying gasp was thus sent), disconnecting the cable linking the modem to the SSPF and leaving the equipment in that state overnight. When an active broadband circuit was eventually required the next day, to just reverse those two steps.

Upon examination of the latest set of statistics, once the circuit had been re-established, the "oscillation" was no longer present. 

[kitz]

Ive had a look at your SNRm and attainable.  Re your SNRm since 3pm yesterday, I'd do as suggested by b*cat.

The larger picture for you attainable looks like it could possibly be one of your crosstalk interferers who has turned off his router for a while and then back on.. the SNRm box-spike at around midnight on the 13th certainly looks like that. 

[AArdvark]

B*cat/Kitz,

Thanks for the quick reply.

I tried this, leaving the line off for 8 hours.

I am back at the same sync/attainable I had before switch off (no re-syncs yet):

Max:    Upstream rate = 26647 Kbps, Downstream rate = 77427 Kbps
Bearer: 0, Upstream rate = 19999 Kbps, Downstream rate = 62463 Kbps

Looks like I need to wait for DLM to run its course (hopefully) and see if the Downstream rate gets any better.

[AArdvark]

Spoke too soon 

Re-syncs have started again.

Uptime:                    55 min 39 sec
Resyncs:                   1 (since 15 Feb 2015 12:32:07)

I note that I seem to be getting bursts of High FEC errors then a re-sync.

I have alerts set in DSLstats and this pattern is new.
The last alert I had for High FEC errors was on 01/10/2014. (High FEC Alert set to 1000 in DSLstats)

At what point would I be able to report a problem to Plusnet ?
 

[kitz]

Ive just had a look at your MDWS and I cant see anything strange on there.

The one resync recorded appears to be you restarting your router at 12:32.  From looking at your line 1000 FECs isnt unusual nor would it be for some interleaved lines.   Your CRCs are at 0..  and more importantly your ErrSecs are also at 0 which is good

The DLM kicked in yesterday at 4am to apply interleaving & INP which is why your sync speed is now lower.  From what I can see, atm your line is behaving normally.

Im guessing here, but a resync whilst your x-talk disturber was offline enabled you a decent sync with plenty of SNRm, however when the disturber turned his router back on then it took your SNRm down, which will in turn cause errors up to a point when it forces your line a resync to gain more SNRm.   Unfortunately DLM has noticed this though and it looks like youve been penalised.    I can see similar sometimes when one of my disturbers switches his router off and his actions result in 5Mb changes each and every time.  Fortunately for me though, I have sufficient spare SNRm to be able to cope so DLM isnt bothered.   However as each new line come on my cab, that margin is deteriorating further and there will come a day, when its possible that I could find myself in that situation too. 

[AArdvark]

I must/will stop stressing over this 

I cannot do anything to change what is going to happen anyway.

I will talk a deep breath and resume my 'Zen-like' state once more.   

Thanks.

[les-70]

I must/will stop stressing over this 

   
   I know the feeling.   You might want to try going back to the HG612 as the modem and bridge to the Xyxel or another router.  Most seem fine with the Zyxel but I had resyncs and very high error bursts with it.  The HG612 shows smaller error burst for me and no resyncs.  One or two other also find higher error rates with the Xyxel.

[kitz]

Im not sure about that ^.    iirc Aardvarks line normally performs better with the Zyxel than the HG612.  Putting the HG612 on will give him an even lower sync speed.

The mystery with aardvarks line is that his E/S shouldnt have been sufficient to trigger the DLM...(see below)  but look at his SNRm graph the day before interleaving kicked in..  look how low his SNRm was and look how it changed through the results of what could have been the result of an x-talk disturbers action.  Could it be that the DLM kicked in through MTBR rather than MTBE..  If you look at his CRC's then theres just one huge burst at 11:30 on the 13th... (which can often happen when a x-talker joins the cab)...  in which case I think the best course would be leaving things alone for a while, hoping they settle and hoping that his disturber doesnt go messing with it's line.

If you go back further on aardvarks SNRm graph you'll see its been running happily for weeks and weeks at around 4dB SNRm without producing any significant amount of E/S that the DLM may be interested in.

[kitz]

@les-70,  should mention that aardvarks stats have always been similarish to mine, bar at times the effects of new users joining the cab.
His similar line conditions to mine could be the reason why the Zyxel usually performs better on his, just as it does on mine.   I cant recall the exact figures now, but aardvark got quite a large chunk of speed back when using the Zyxel than the HG612.   I use the Zyxel because its the only one that allows to to continue to sync at 80/20 and be stable with some to spare.   Some other modems only allow me to sync at 74-75Mb.   If you recall when I had a fault last Oct I put the HG612 back on for a short while, but took it off because it was performing much worse than the Zyxel.

What I think may have happened is that his line has been stable for a while, but the SNRm decreases as each new user is put on..  hence it being down to 4dB for quite a while.   What looks like a recent new user added, just seems to have tipped the scale and with it already being at 3.5dB the line couldnt stand another disturber and be stable.   I'll admit Im guessing but to me it looks like the combo of the not unusual CRC spike that occurs when a new user is added, and then the new user possibly messing with his router has caused the jumping around of aardvarks SNRm and subsequent resyncs. 

Im hoping that if aardvark can now maintain a stable 6dB and his current error rate then the DLM will eventually relent.    I think the lesson learnt here is if you see SNRm doing the slow decline due to crosstalkers being added, then you are best performing a manual resync to bring your SNRm back up to 6dB, rather than wait for another user to be added to the cab and the modem being forced to do a resync itself.

[WWWombat]

The mystery with aardvarks line is that his E/S shouldnt have been sufficient to trigger the DLM...(see below)  but look at his SNRm graph the day before interleaving kicked in..  look how low his SNRm was and look how it changed through the results of what could have been the result of an x-talk disturbers action.  Could it be that the DLM kicked in through MTBR rather than MTBE..  If you look at his CRC's then theres just one huge burst at 11:30 on the 13th... (which can often happen when a x-talker joins the cab)...  in which case I think the best course would be leaving things alone for a while, hoping they settle and hoping that his disturber doesnt go messing with it's line.

If you go back further on aardvarks SNRm graph you'll see its been running happily for weeks and weeks at around 4dB SNRm without producing any significant amount of E/S that the DLM may be interested in.

I agree - it is noticeable that the ES rate climbs a little as the SNRM drops over the 90 day period, but none of it appears significant at all. The only sign is that one (very big) burst of CRC errors on the day before DLM got involved.

[les-70]

  The HG612 will certainly lower the sync a bit but resyncs to get 6db don't seem to have helped the Xyxel.  The Xyxel is clearly fine on lines without much noise but if as well as cross talk he has extra noise then the Xyxel may not be happy.  The HG612 is a suggestion  as if if it helps it would tend to suggest extra noise and not just crosstalk. 


    That said the DLM action is puzzling.  The ES were low and a succession of resyncs surely can't be explained by low SNRM due to changing crosstalk.  It should only take one resync to respond to a low SNRM.

   

[AArdvark]

All,

I have been getting a very good sync with stability since swapping to the Zyxel and would be loath to change to the HG612.
As Kitz states my line is very similar based on past performance with what was a large Max Attainable circa 110Mbps.

A typical sync for the last few months was:

Max Attainable :    Upstream rate = 27066 Kbps, Downstream rate = 74357 Kbps
                              Upstream rate = 19999 Kbps, Downstream rate = 79987 Kbps


The 1st re-sync was ME, I rebooted the Zyxel to try to improve the sync when SNRM was 3.5db.

It then bounced between 79.9/72 <--> 77/62 a few times and I wrote the OP.

The FEC peaks seem to be at 10 min intervals when they happen, so I think something is happening elsewhere affecting the line ?

I think it is a 'Keep Hands-off & wait ' moment for me and I will wait to see where it settles.

Many Thanks for the help and suggestions.

[kitz]

I wouldn't worry too much about the FECs atm,  its CRCs and Err Secs that Im more interested in.

FEC's are weird and its false to make the assumption that 1 x FEC would equal 1 x CRC.  FEC's run at a much higher rate and from all observations a single FEC doesnt seem to equate to a non-interleaved CRC**

I don't know the actual reason for this, but there's many lines that can rack up thousands of FECs yet still be perfectly fine when it comes to CRCs. **
My obvious line of thought is it may be something to do with this:-

FACTS:-

- TCM (Convolutional Error Correction) is usually permanently on for all non-interleaved lines.  Trellis uses a Viterbi algorithm has very little overhead and no counters.
- When interleaving and error correction are switched on, DSL uses a Concatenated Convolutional (Viterbi) and Block coding (RS) algorithm. 
- When interleaved, its the concatenated code that is now doing all error correction. Its acknowledged that the Viterbi code usually catches most of the errors whilst RS code cleans up the rest.   

Speculation:-

If both Viterbi & RS algorithms are being used in the concatenated FEC code, could the count of FECs be higher because its either catching or counting some errors which would have ordinarily been caught by TCM alone, and not triggered any counters.
-----------------------------------------

** As an example of this I attach a copy of Ronski's stats. For the sake of simplicity I'm ignoring spikes.. but here we can clearly see that when interleaved the graph shows circa 1000 FECs.  When not interleaved the CRCs are substantially less maybe a hundred or so. Even a spike CRC shows only a fraction compared to FECs. 

Obviously the ratio between FEC/CRC would be likely to change depending upon individual line conditions, the type of noise it usually experiences, and whether that noise is usually caught by TCM (white noise) or RS (bursty).

I think BaldEagle may have also said its similar on his line in that he can rack up lots of FECs, but hardly any CRCs in comparison.

[Bald_Eagle1]

FEC's are weird and its false to make the assumption that 1 x FEC would equal 1 x CRC.  FEC's run at a much higher rate and
I think BaldEagle may have also said its similar on his line in that he can rack up lots of FECs, but hardly any CRCs in comparison.

Just for the record, I have attached recent typical examples of DS FEC (RSCorr) & CRC (OHFErr) graphs from my connection.

On only one occasion (not much before 03:00 this morning) do the FEC spikes come anywhere near to matching the CRC spikes (timescale-wise).

These FEC counts are from a fastpath connection where INP & delay are zero for both DS & US.