cancel
Showing results for 
Search instead for 
Did you mean: 

DSL stats graphs microanalysis

Community Veteran
Posts: 5,059
Thanks: 426
Fixes: 16
Registered: 10-06-2010

DSL stats graphs microanalysis

Or why I should just stop caring and find something else to do.
That DS SNRM won't be going back up until a re-sync, which despite the reading, will most likely achieve a similar speed.
28 REPLIES
Superuser
Superuser
Posts: 11,083
Thanks: 2,430
Fixes: 22
Registered: 22-08-2007

Re: DSL stats graphs microanalysis

Hi ejs,
Stop caring?  Well I guess that's all a matter of what you want to gain - or more to the point prevent.  This could be indicative of a deteriorating line or a failing switched mode power supply.
Looking at the three day's plots, the first shows periods of very flat and stable SNRM and other (non-nighttime) variability of circa 1dB - what is the SNRM resolution of your router please?
The second plot shows a step shift in SNRM at 14:45 and is rather indicative of REIN, which up until this morning has not abated.  The current error rate is somewhat higher than the preceding two graphs.  This is a typical REIN profile.
Personally I'd be checking the phone line for noise on a regular basis and looking for the effects of turning off (not putting into standby) electrical devices around the home and examining the impact on SNRM.
There again, given that the line has not dropped in the presence of interference, you might indeed decide to do nothing.  Don't know what router you have, but it is keeping a good hold of the connection!  Wink
As for the speed you'll see on a reboot, my money would be on it being lower if the interference is still present.
Community Veteran
Posts: 5,059
Thanks: 426
Fixes: 16
Registered: 10-06-2010

Re: DSL stats graphs microanalysis

There isn't any interference that is continuing to knock 2db off the DS SNRM. All that happened is that something happened at 14:36 on Sunday and the graph won't look quite the same until whenever the next re-sync occurs. Probably the telephone being answered but I've given up constantly quizzing people as to what times the telephone was answered. And the damned if I fix it myself, damned if I don't bridge tap (star) wiring.
I've drawn some bits per tone graphs so you can see the frequency the non-existent interference is on. I see there's usually a slight dip there in some historical bits/tone plots. Or perhaps the "interference" was reflected DSL signal bouncing around the wiring.
The SNRM is reported to 0.1db and the reported value flipping between 5.2 and 6 in those hours is pretty common. Although the DS SNRM swimming around down at 3-4 isn't so common.
Community Veteran
Posts: 5,059
Thanks: 426
Fixes: 16
Registered: 10-06-2010

Re: DSL stats graphs microanalysis

Unfortunately for me I was wrong. However I suppose the speed of 3522k is still higher than the expected for the reported attenuation of 58.4. The dips in the SNR lines are almost but not quite evenly spaced. It's annoying knowing that it's not performing as well as it could be or has done in the past, but I suppose the difference in speeds doesn't really make any difference.
Community Veteran
Posts: 5,059
Thanks: 426
Fixes: 16
Registered: 10-06-2010

Re: DSL stats graphs microanalysis

Obviously this is the kind of issue than can be affected by arbitrarily unplugging and plugging things back in, and moving cables about a bit.
04:08 - Get up needlessly early. Look at microfilter at socket with modem/router and telephone connected. Confirmed it was still there.
04:39 - Unplug / plug other microfilter with other telephone attached. Can't remember if I unplugged the telephone from the microfilter or the microfilter from the socket.
05:46 - Unplug / plug telephone from microfilter with modem/router attached and move cables about a little.
I think I'll leave commanding a re-sync until tomorrow, if the telephone being answered today doesn't do that for me.
Superuser
Superuser
Posts: 11,083
Thanks: 2,430
Fixes: 22
Registered: 22-08-2007

Re: DSL stats graphs microanalysis

Hi ejs,
Looks like moving wires - especially the last one - influenced the SNRM.
Would be interesting to know what happens to the tone bins when the wires are moved around.  Specifically does that "hole" change?
Do you have an AM/MW portable radio?
Kevin
Community Veteran
Posts: 5,059
Thanks: 426
Fixes: 16
Registered: 10-06-2010

Re: DSL stats graphs microanalysis

I suspect the unplugging and plugging back in of the telephone had the greater impact. None of the cables were moved any great distances. The large hole in the bitloading graph had mostly been filled in by 06:00. A computer may have been switched on around 08:32 although that could have been unrelated since that computer would be switched on and off every day. I don't know what happened at 13:01.
Yes I do have an AM radio, although I think its only use will be for checking what radio stations are there, although AM radios only start at 531 kHz of course.
Community Veteran
Posts: 19,100
Thanks: 437
Fixes: 21
Registered: 31-08-2007

Re: DSL stats graphs microanalysis

Hi ejs, | want to take a more lengthy look at this lot before commenting, but one query may help with a small puzzle, are you on the east side of the country?
Community Veteran
Posts: 5,059
Thanks: 426
Fixes: 16
Registered: 10-06-2010

Re: DSL stats graphs microanalysis

Yes, I'm in the south east, there are a few faint presumably French radio stations at the low end of the AM range, and also a few faint English stations.
Community Veteran
Posts: 5,059
Thanks: 426
Fixes: 16
Registered: 10-06-2010

Re: DSL stats graphs microanalysis

Sorry, I should have said earlier, that I wouldn't want anyone else to waste much time on this. I did have the netgear modem/router connected at the other socket from 13 July to 30 October, I also double filtered both telephones (both are basic BT branded corded phones) from 3 Oct to 23 Oct. I also tried swapping out the XF-1e microfilters with the ones that came with the 582n on the 30 October. There happened to be more telephone calls received than usual on 30 October (as in four, rather than the usual none or one), and the DSL was dropped three times, which prompted putting the XF-1e filters back and moving the netgear back to the socket where it always used to be.
I think the changes I made didn't really make any difference. I should probably try another modem/router for comparison purposes, although the netgear achieves a fairly good speed under the circumstances and seems to somehow manage to get away with the drops.
Community Veteran
Posts: 19,100
Thanks: 437
Fixes: 21
Registered: 31-08-2007

Re: DSL stats graphs microanalysis

No worries, I will have a look later on.
Community Veteran
Posts: 5,059
Thanks: 426
Fixes: 16
Registered: 10-06-2010

Re: DSL stats graphs microanalysis

For a while I was worried that I might be stuck using the 582n, but I switched it to ADSL2 slightly earlier than planned since the qln command apparently also drops the DSL to do its test. And it didn't increase the speeds to quite as high as the old netgear achieved. Still, switching it from ADSL2+ to ADSL2 increased the upstream speed significantly, increased the upstream noise margin and decreased the FEC error rate on the upstream.
Edit: permanently attach graph
Community Veteran
Posts: 19,100
Thanks: 437
Fixes: 21
Registered: 31-08-2007

Re: DSL stats graphs microanalysis

As I'm sure you appreciate ADSL2/2+ has no direct bearing on the upstream, the resync is more likely to have resulted in what you've seen. I will, as promised, come and look at this lot as soon as I have a moment and we can debate what is going on here. Right now, I have to go.
Community Veteran
Posts: 5,059
Thanks: 426
Fixes: 16
Registered: 10-06-2010

Re: DSL stats graphs microanalysis

The graph was broken after 13:30, it should be fixed now. I'm "pleased" to see that the 582n is no better at holding the DSL connection when a phone is answered than the netgear. Pleased in the sense that there's no particularly compelling reason for me to continue using the 582n.
The difference in downstream frequency range is not the only difference between ADSL2 and ADSL2+. Regarding the upstream, the permitted power level on the frequencies above the upstream tones (above 138 kHz) is much lower on ADSL2+. Which may explain why the bitloading on the higher upstream tones is reduced when using ADSL2+.
How many times would you like me to swap between ADSL2 and ADSL2+ before you believe the difference in speeds is due to switching modes? For obvious reasons I'm not going to do it more than once per day. I also have the first couple of months of DSL stat graphs on ADSL2+ with similar lower speeds compared to the rest at slightly higher upstream and downstream speeds on ADSL2.
Well of course the re-sync resulted in what's on the graph - it kind of goes without saying that the speeds changed due to the re-sync, because speeds don't tend to change without a re-sync, there's no SRA going on here. The first re-sync was me changing the mode to ADSL2.
I expect many people on long ADSL2+ lines might achieve higher speeds from switching to ADSL2. The only reason they need to change that setting manually is because some exchanges (perhaps older 21CN ones?) don't seem to automatically select ADSL2 for those long lines. Others just seem to automatically end up on ADSL2.
Community Veteran
Posts: 19,100
Thanks: 437
Fixes: 21
Registered: 31-08-2007

Re: DSL stats graphs microanalysis

Quote from: ejs
How many times would you like me to swap between ADSL2 and ADSL2+ before you believe the difference in speeds is due to switching modes?

Never said it wasn't, particularly on your connection. Sometimes you read too much into what I do say  Huh  I did say in the particular case of one connection at the moment that it wasn't going to make the slightest bit of difference because of the state of the connection/modem-router, and in any event, I still reckon it falls more into the category of fine tuning, but anyway, we will debate more shortly Smiley