SNR/Connection speed wobble.

[#1]

@PN staff - sorry to be a pain, but any chance of RADIUS  logs from Friday to date, please?

[#2]

Hi @jab1

RADIUS log here.

Seems to have settled down since Saturday night but still a couple of drops.

Thanks

Dave

[#3]

@RandallFlagg Cheers, Dave. I'm still going to monitor for a couple of days.

[#4]

No problem at all, @jab1

Let us know if there's anything else you need.

Best wishes

Dave

[#5]

Todays' RSL plots.

[#6]

This most certainly looks like SHINE / RAIN. The shapes of the spikes are quite similar. Is xDSL dropping on all of these transient events? Without correlation to the sync speed plots, it is not possible to discern if there’s -10dB in margin for a while or if these plots show an instant loss of sync, followed by a reconnection at 3dB with the noise still present, after which when the noise burst ceases the margin increases to 13dB.

I suspect that if after such an event, the router is restarted, you will see a steady connection at the target SNRM at a higher speed. You need to find some correlation between the spikes and something else. Setting an alarm on the SNRM plotting will alert you to the spike event ... and may be something you are doing in the home.

[#7]

Hi Kevin.

To try and answer your questions:- From the limited data available in the TG582n logs, xDSL certainly appears to be dropping on all these events. I will post the sync speed plots associated with the SNR plots, hopefully some time today.

As these events are happening when I have gone away from the laptop, and usually when I am fast asleep, setting an alarm will not achieve much - I'm not that likely to hear it, and even less likely to want to jump out of bed and run downstairs IF I did.

Again, as these events all occur during the night and I don't see them until the following morning, I am not in a position to restart the router after them. They are nothing to do with anything happening in my home - at the time they occur, I am usually in bed, and the only possible internal culprit is the fridge/freezer, which I think we can safely discount as it runs 24/7/365 and has no visible effect - no 'every 40 minute' interruptions.

FYI, last nights SNR plots are below.

BTW - thanks for your continuing interest.

[#8]

@Townman  Connection speed graphs for the same period as covered by the SNR plots.

[#9]

[#10]

[#11]

Hi John,

Thank you for the plots.  Have you change the sample interval?  If not, these drops look different to the ones we have seen before - they are a straight drop and a clean resync at the same SNRM same sync speed, with the exception of the 05:30 drops which try to establish a SNRM of 6dB at a higher speed.

Do I recall correctly that your line has been fixed to a 12dB SNRM in an attempt to maintain a stable connection?  If yes, then for most of the time when there is no variation (no REIN) you are potentially sacrificing synch speed … but if there are substantial errors, that might remain pragmatic.

Going forwards, only post those plots with something "interesting" showing.  Flat liners can be ignored.

Set the sample interval to its very minimum and graph plot count to give 2 hours.

Try plotting the FECs as "number in interval" not cumulative.

@RandallFlagg @LordFoul @Mads @MasterOfReality 

What is the target SNRM for this line please?

Is it banded?  This would explain most of the reconnections being at 12dB.

Can you tell from the DLM logs the reasons for these drops, particularly those around 05:30 where the reconnections attempted 6dB?

What does the line balance look like?

What has already been tried on this line?

• D-side swap?
• E-side swap?
• Port swap?
• Lift & shift?
• Tied Pair Move?
• TG582n firmware upgrade?

Are there other customers nearby who's line performance might be examined for correlated drops?

Though the TG582n router will now be of some age, it is arguably the best of ADSL routers supplied by PN and is the last one which works with any performance monitoring capability.

[#12]

Morning Kevin.

I am not aware of changing the sample interval (or any other settings, come to that) since installing RSL in accordance with your excellent guide a loooong time ago - set it and forget it is my motto.

You are correct with regards to the 12dB SNRM and I am happy with that, and the resulting small loss of speed - apart from the odd (Very few) big program updates that take a while, nothing I do on this machine needs more speed.

I will adjust the sample interval/graph plot later - got a bit of a headache atm which is disturbing my concentration.

As to 'Try plotting the FECs as "number in interval" not cumulative.' - how?

I will await with interest the reply from the Help Team.

[#13]

@Townman Sample interval now set to every 5 seconds (minimum). How do I set the graph plot to 2 hours, please?

[#14]

Number of plots per graph = X

5 seconds interval = 12 per minute
12 x 60 = 720 per hour

Therefore 1440 delivers 2 hour plots.

Way back when I wrote the original set up instructions 10 seconds was the smallest interval, so 720 gave 2 hours.

I will comment on the previous post later.

[#15]

Thanks, Kevin.

I forgot to say I'd worked it out earlier and adjusted to sort. -sorry.