Kitz Forum

Broadband Related => ADSL Issues => Topic started by: 8062282 on July 11, 2019, 06:06:00 PM

Title: A very bad day...
Post by: 8062282 on July 11, 2019, 06:06:00 PM
On the 3rd of July I came home to an ongoing spike in errors which started about 4 hours before I got home & was still ongoing until I saw it. I immediately re-booted the modem & everything went back to normal. No idea what started it.  How does the DLM react to a one off bad day? As you can see the errors for that day were massive. The FEC errors went from zero to 3000 & were just going on & on until I re-booted.

Average error rates for 03 Jul 2019
CRC erors per hour:  11383 Down,  0.58 Up
FEC erors per hour:  7569463 Down,  5.55 Up
HEC erors per hour:  17475 Down,  0.34 Up
ES per hour:  1071 Down,  0.58 Up
SES per hour:  327 Down,  0 Up
Title: Re: A very bad day...
Post by: 8062282 on July 12, 2019, 05:37:01 PM
It's happened again. The SNR shot off to 16 dB at 23.20 last night & then went down to 13.5dB with a speed of 5511 when the DLM kicked in at 03.31. It then tootled along until 11.45 when it dropped to 1.2dB due to 'something' which was creating 1000's of ES errors. It only stopped at 14.00 when I noticed it & re-booted the router. I finish at 12.30 on a Friday, if it'd been in the week, I wouldn't have picked up on it until after 16.00, by which time 1000's more ES would have been created..


Any help with what might be causing this type of interference? We've tested various things in the house & nothing causes this amount of errors. When we switch certain things on, they can cause a slight spike but it's sudden & nowhere near 300 CRC's. Nothing really creates any ES either.   


Errors so far for today are pretty bad..



Average error rates for 12 Jul 2019


CRC erors per hour:  2971 Down,  2.85 Up
FEC erors per hour:  3221902 Down,  6.89 Up
HEC erors per hour:  3435 Down,  3.47 Up
ES per hour:  473 Down,  1.54 Up
SES per hour:  0.91 Down,  0 Up

Title: Re: A very bad day...
Post by: burakkucat on July 12, 2019, 06:36:52 PM
Yuck is the appropriate word for that latest CRC per minute plot.

Unfortunately I do not have a clue as what to suggest for your next tests.  :(
Title: Re: A very bad day...
Post by: 8062282 on July 12, 2019, 07:24:13 PM

I'm learning so please bear with me. Does that look like it might be something internal or could it be something between me & the exchange? If it looks internal, that's a start. if it could be anything between me & the exchange, I might as well cut my wrists now  :lol:
Title: Re: A very bad day...
Post by: burakkucat on July 12, 2019, 08:46:54 PM
Does that look like it might be something internal or could it be something between me & the exchange?

Seriously? I have no idea.

The fact that it showed as a "elevated block", with virtually vertical sides, make me want to think it was something that was switched on at 1142 hours and was then switched off at 1410 hours. But appearances can be deceptive.
Title: Re: A very bad day...
Post by: 8062282 on July 12, 2019, 09:33:11 PM
Hi - OK, thanks. Could that have happened anywhere between me and the exchange or do you think the culprit could be internal?
Title: Re: A very bad day...
Post by: kitz on July 13, 2019, 08:51:48 PM
Seriously? I have no idea.

The fact that it showed as a "elevated block", with virtually vertical sides, make me want to think it was something that was switched on at 1142 hours and was then switched off at 1410 hours. But appearances can be deceptive.

Just mulling some thoughts with you - notice this

Quote
It only stopped at 14.00 when I noticed it & re-booted the router.

Also that the pattern in the graph looks exactly like mine does after a power surge or remote power outage and why I have to watch it like a hawk so it doesn't trigger DLM.      It's not uncommon for most modems to go into a prolonged error state after a remote disconnect -  Max's line did it just last week.   As you know we normally recommend letting the port rest, but sometimes I'm being lazy or in a hurry OR if I'm out and get an alarm from DSLstats... then a simple power down and resync will suffice.

I think there has been mention of brief SNRm spikes in the past week.    Also pondering SHINE - again as an example, how my own line practically every morning will have a brief burst of CRCs which can also on the odd occasion trigger the ErrSecs to go mental until I reboot the modem.   Mentioning it because as you know, this is precisely the reason why I have DSLstats set up to send an alarm to my phone so I can remote reboot if need be.   

From the limited graphs from the OP in the other thread, I'm sure there were a couple of what could be SHINE type spikes.  Thoughts?

____

@8062282     

You mention the errors, did the SNRM stay down at 1.2dB.   It's not clear from your description if it was a short spike or whether the SNRM stayed that low.
 
Title: Re: A very bad day...
Post by: burakkucat on July 13, 2019, 10:12:18 PM
Just mulling some thoughts with you - notice this

Quote
It only stopped at 14.00 when I noticed it & re-booted the router.

Ah, yes. Silly old, grumpy, black cat. I had a niggling thought in the back of my mind after I had typed that reply . . . and then promptly got distracted by something else.  :-[

Quote
I think there has been mention of brief SNRm spikes in the past week.    Also pondering SHINE - again as an example, how my own line practically every morning will have a brief burst of CRCs which can also on the odd occasion trigger the ErrSecs to go mental until I reboot the modem.   Mentioning it because as you know, this is precisely the reason why I have DSLstats set up to send an alarm to my phone so I can remote reboot if need be.   

From the limited graphs from the OP in the other thread, I'm sure there were a couple of what could be SHINE type spikes.  Thoughts?

There are certainly quite a few similarities with the behaviour of your own circuit when it has a crazy fit.

SHINE. Hmm . . . Not the easiest of events to diagnose but having re-read the OP's other thread I guess that one or more SHINE events may be responsible for triggering the CRC crazies.

For the OP: If you haven't already seen it there is a description of SHINE on this page (https://kitz.co.uk/adsl/rein.htm#SHINE).
Title: Re: A very bad day...
Post by: g3uiss on July 13, 2019, 10:49:36 PM
Just mulling some thoughts with you

From the limited graphs from the OP in the other thread, I'm sure there were a couple of what could be SHINE type spikes. 
 

Looks exactly like my line also.  My spike ( like @kitz ) is at a similar time each day and sometimes it resyncs, then the CRC,s are all over the place with continued errors until I force a resync. It would appear to be worst at weekends and certainly not generated in my property, I don’t think there is any way to track it down, or @kitz would have found it before now. I guess perhaps there is another spike during the retrain that trips up the modem.
Title: Re: A very bad day...
Post by: 8062282 on July 14, 2019, 09:02:35 AM

Just mulling some thoughts with you - notice this

Also that the pattern in the graph looks exactly like mine does after a power surge or remote power outage and why I have to watch it like a hawk so it doesn't trigger DLM.      It's not uncommon for most modems to go into a prolonged error state after a remote disconnect -  Max's line did it just last week.   As you know we normally recommend letting the port rest, but sometimes I'm being lazy or in a hurry OR if I'm out and get an alarm from DSLstats... then a simple power down and resync will suffice.


I think there has been mention of brief SNRm spikes in the past week.    Also pondering SHINE - again as an example, how my own line practically every morning will have a brief burst of CRCs which can also on the odd occasion trigger the ErrSecs to go mental until I reboot the modem.   Mentioning it because as you know, this is precisely the reason why I have DSLstats set up to send an alarm to my phone so I can remote reboot if need be.   

From the limited graphs from the OP in the other thread, I'm sure there were a couple of what could be SHINE type spikes.  Thoughts?
____

@8062282     

You mention the errors, did the SNRM stay down at 1.2dB.   It's not clear from your description if it was a short spike or whether the SNRM stayed that low.

The ES errors at night are always about the same time 22.00 - 05.30. The SNR has started to go up to 16db each night, about midnight due to a 'message exchange' (in the modem log). Sometimes, the DLM kicks in at 03.30 & it goes back down to 6dB. Last night it went to 16dB & the DLM hasn't intervened this morning. I'm still at 16dB.

TT haven't really been forthcoming in what they're doing, if anything.  The daytime error which seems to happen around lunchtime is more serious in that it doesn't stop. It just carries on until the modem is re-started. I'm at work when it starts so I can only re-start the modem when I get home. By that time, it's too late as it's already created 1000's of ES/CRC & FEC errors. I'll see if I can set an alarm & send it to my phone. It happened Friday at lunchtime as well.

Average error rates for 12 Jul 2019

CRC erors per hour:  2180 Down,  2.25 Up
FEC erors per hour:  2360824 Down,  5.09 Up
HEC erors per hour:  2574 Down,  2.63 Up
ES per hour:  349 Down,  1.25 Up
SES per hour:  0.67 Down,  0 Up

The SNRM went down to 1.2dB on Friday when the daytime errors started, it went back up to 6dB but only when I re-started the modem. I caught it as it started again on Saturday lunchtime & managed to re-start the modem before it caused too much damage. SNRM only went down to 4.8dB that time. There wasn't anything in the house which had just been switched on that could have triggered it. The night time ES doesn't make the SNRM alter, it stays at 6db. It goes a tad wobbly, but only deviating, maybe at most, 0.2dB.  I've also attached a graph of my bit-loading. That doesn't look right to me.

I've had a read up on REIN & SHINE. Still need to read up more, though. The street lamp that I thought might have been creating the ES at night isn't the culprit as the ES starts about 15 mins before the lamp came on. Unless the lamp is revving up to come on.

I'm at a loss as what to do. I've cancelled my contract with TT as there's space in my cabinet for me to get fibre. I have no idea whether I'm going to take all this with me on VDSL. If I do, I could be struggling to keep hold of a decent speed with all these errors.  Is any ISP going to be interested in helping me solve this?

I've ordered a new modem that was going cheap on eBay. It's a a Zyxel VMG3925-B10B. I can test it on my ADSL line & rule out my 7800DXL being faulty. I can also use it on VDSL with DSLstats & monitor my connection to see if I'm having the same issues at roughly the same times on VDSL.

The good thing about ADSL is, it's quite easy to get your ISP to re-start the DLM. If I have the same problems on VDSL, it's going to be a right pain in the proverbial as I've read that ISP's won't start the DLM without an engineer's visit.
Title: Re: A very bad day...
Post by: burakkucat on July 14, 2019, 05:25:59 PM
That bit-loading per sub-carrier plot is somewhat surprising.  :-\

It might be useful to see the Hlog per sub-carrier plot, as that will tell us about the physical state of the metallic pathway.
Title: Re: A very bad day...
Post by: 8062282 on July 14, 2019, 06:07:26 PM
Hi - Hope this is the right one..
Title: Re: A very bad day...
Post by: burakkucat on July 14, 2019, 06:16:22 PM
Yes, that's the one. Unfortunately it does not disclose any fundamental defect in the physical circuit. It looks quite reasonable for a long line.

So that you have something with which to compare your plot, I've attached a copy of the equivalent for my circuit (harvested on July 1st, 2019), below.
Title: Re: A very bad day...
Post by: 8062282 on July 14, 2019, 06:31:55 PM
How long is considered long? I'm about 2.8km from the exchange. What looked 'surprising' about the bit-loading one?
Title: Re: A very bad day...
Post by: burakkucat on July 14, 2019, 08:20:29 PM
I consider a line to be long when it is unable to make use of the full bandwidth that is available . . . In both our cases, our services are provisioned as ADSL2+ but neither of us can make use of sub-carriers numbered from 256 to 511. In my case I have configured the ATU-R (my modem) to operate in ADSL2 mode, not ADSL2+ mode.

The shape of your bit-loading plot is somewhat similar to that of a VDSL2 service, which will have power-shaping applied to the US0 and DS1 bands. (USPBO and DSPBO, respectively.) Again, to give you some reference, I attach my bit loading per sub-carrier plot.
Title: Re: A very bad day...
Post by: 8062282 on July 14, 2019, 09:39:10 PM
I consider a line to be long when it is unable to make use of the full bandwidth that is available . . . In both our cases, our services are provisioned as ADSL2+ but neither of us can make use of sub-carriers numbered from 256 to 511. In my case I have configured the ATU-R (my modem) to operate in ADSL2 mode, not ADSL2+ mode.

The shape of your bit-loading plot is somewhat similar to that of a VDSL2 service, which will have power-shaping applied to the US0 and DS1 bands. (USPBO and DSPBO, respectively.) Again, to give you some reference, I attach my bit loading per sub-carrier plot.

Hi - OK, thanks for the info. I did read about changing to ADSL2. Tried changing it in the 7800DXL but it wouldn't establish a link. I either didn't wait long enough or there's something else I needed to do. 


Sent from my iPad using Tapatalk