Internet Drops Out - Opinions on my options please

[#1]

Quote from: EnglishMohican

@ejs. Against Trained Mode 3 it says Selected Mode 1 which /proc/avalanche/avsar_dsl_modulation_schemes translates to Multi_Mode.  Have I understood what you intended me to do correctly and can I take this further?

Not quite - "Selected Mode" is what you asked for and "Trained Mode" is what you've ended up with. Plus avsar_dsl_modulation_schemes values are in hexadecimal but avsar_modem_stats gives the modes in decimal.
Selected Mode: 1 = Multi Mode, the auto setting
Trained Mode: 3 = ADSL_G.dmt, ADSL1
The change in behaviour and small speed increase is probably due to the newer equipment from a different vendor in the exchange. But as has already been said, regardless of being ADSL1 or ADSL2, your speed should be much higher.
I think someone did say it's possible to have uncapped upstream on 21CN ADSL1.

[#2]

Hi ejs, you are probably right, but as it doesn't (at this time anyway) seem relevant I didn't bother looking any more at it, and yes you are right I understand you can have uncapped Upstream of 21CN ADSL1. He's got a complex setup there, and we'll need to eliminate any possible sources of noise at his end before considering other things, especially as (from his posts) the general noise problem has been going quite a while.

[#3]

Hi EnglishMohican,
You are on ADSL1 on 20CN 21CN, Your current SNR is 15dB which will be keeping your speed down. Did you notice any issues on your phone line? I could reset the SNR to 6dB but if you have a problem on your phone line this may lift again and it wouldn't be able to be reset for another 10 days. Let me know how you want to proceed with it.
Jojo

[#4]

Can you provide an updated connection chart please Jojo? Also 20CN ADSL1 doesn't seem to tally with the Upstream speed his modem/router is showing?

[#5]

You can have uncapped upstream on ADSL1 on 21CN, just realised the mistake in my post

[#6]

Thanks for the interest everybody.
I attach another plot, this one showing a spike at about 14:46. It is not  the same as last nights as it is shorter (2 points only I think) and both readings are 0. My guess is that it is false data rather than a real spike but lets share that decision. There were a couple more this morning. Neither were associated with a loss of connection.
@Jojo: Lets leave things as they are for the moment please. I would like to get a better understanding of what is going on before changing anything and I fear that might take several weeks of data collecting before we can see a pattern.
Responding to a few other points:- I cannot think of anyway that Chris's plot could be absolutely correct about Sunday. I am quite surprised that Plusnet can link my community name to my account and I certainly do not have two accounts anywhere. (must be careful what I say about Plusnet in future!). My best explanation is that one day recently I lost connection and my normal disconnect/reconnect would not reconnect - so I turned the modem off for an hour or two and then it did reconnect. That could have been Sunday and the earlier session could have been hard to trace (not sure how these plots are logged).
I am currently on a laptop. I could put routerstats onto another machine but it can run on this one alright unless you want two sets of results. Putting it on the NAS box would have saved a few pennies on electricity and given me a "for ever" monitoring solution but I agree I would need to compile from source to get a suitable executable and that feels to be a job for another occasion.
I will continue monitoring and hope it is not too long before something shows up.

[#7]

The main reason for putting Routerstats Lite on another machine (Laptop would be the most convenient - it can be moved easily) is so that the current one can be turned off at some point as we try and eliminate things. The Lite version will suffice at this point on an alternative machine.
What I'll suggest at this point, when it's convenient, is connect your current laptop to the DG834v1 via ethernet and confirm the pattern you are seeing. Then unplug any other ethernet connections from the DG834v1, note any changes to the Noise Margin. If no change, the leave them unplugged, and turn off your wireless router.
If still no change, turn off everything in the house, except the DG834v1 and your laptop, fridges, freezers, central heating, TV's, lights, the lot - nothing in standby - all off. See if there is any change. If the Noise Margin doesn't increase, switch everything back on.
If there is a change and the Noise Margin gets bigger (ie. noise levels decrease), switch things back on, one at a time I'd suggest start with the freezer and the fridge  until you spot something that reduces the margin. Carry on until everything is back on.

[#8]

@Anotherone:    OK, I understand the suggestion but remember that the noise at the moment is not bad (not good either). Therefore it is unlikely to be any of the kit that normally runs  that caused last nights high levels of noise and the spike. I really need to wait for the high noise levels to reoccur and then try the experiment but you can bet it will be the most inconvenient moment possible.
I am away tomorrow and Thursday but will leave routerstats running and see if a pattern develops. Then on Friday, I can try your experiment when I can give it a sensible amount of time.
Thanks again for all the advice and explanations.

[#9]

Quote

but you can bet it will be the most inconvenient moment possible.

Ain't that always the way  I'll keep my eye open for your next update 

[#10]

Hi EnglishMohican,
I'll leave you guys to muse on the finer details. Don't hesitate to shout up if you need our help though.
Jojo

[#11]

After all the help offered, I thought it would be rude not to provide an update but I have to admit that nothing looks very promising so far.
I attach 6 plots. The first shows 4 days worth of Routerstats noise data plotted so that each day overlay the others. The data plotted is a 20point moving average of 8 second sample rate data. If I plotted it at full detail it just become a splodge. X axis is hours after midnight and the y axis is Rx noise level. Each day shows phases and jumps but there is always at least one day that contradicts any sort of pattern in the other days.
The second plot shows Rx noise data directly from Routerstats and is included to show the nice flat portion up to about 6am, the sudden switch into noise at 6, the switch to calm roughly half an hour later and the switch back to noise a further half hour later. I think this example confirms that the noise is one source of human controlled noise. It also shows that the noise can go away almost totally. There are other examples in the data but this is the best one.
The third and fourth show examples of the deep, long lasting spikes that I believe are the cause of my drop outs. The 2 or 3 dB of noise on the basic 15db level is a pain but if the deep spikes were not there, I could run with much smaller noise margin. As it is, the deep spikes prevent that. I also think that the deep spikes are new, the 2 or 3dB noise has been with me for ages.
Plot 5 shows a total crash. The router lost contact with the exchange and disconnects and reconnects and LCP up and so on did not work. I eventually rebooted the router and all was well - in fact even better than before as my speed went up.
And the final plot shows another rectangular dip which may be fake - occasionally routerstats does not get the answer it wants and substitutes 0's. However the large dip at the end of the plot is real enough.
I tend to blame the change on the change to 21CN, maybe its a coincidence or maybe 21CN is stretching my line's ability too far.
I have not managed to turn off all the power in the house yet but knowing my house and looking at the noise pattern I shall be very surprised if it is anything under my control causing this. However, I will get round to that shortly.
All comments gratefully received.

[#12]

Hi again,
Hmm, interesting. I suspect we've got a couple of things going on.
Plot 2, I agree, looks as though it due to some noise source under human control, From the timings, my first thought would be central heating but that doesn't quite match with the first 12 minutes or so on, then 15 minutes off, then on again for at least 30 minutes, what time did it actually stop? Would you have had a washing machine on at that time of day?
Plots 3 & 4 the deeper "spikes", I see there appears to be a "stepped" recovery. I'd say this was a symptom of a bad connection somewhere and would be be the time to go listening for noise on the phone - I don't suppose anyone in the house made or received a phone call at the time of those two examples? And yes I agree it's more than likely that it's these sort of "spikes" causing your drops in connection.
Plot 5, well that's a very significant time interval - I don't suppose you checked to see if you had dial tone for that period?
Plot 6, again a good 5 minutes there, did you lose sync at all? That doesn't seem like a data gathering dip - much too long. Yes, when Routerstats can't get the data either because the CPU is working flat out at that instant processing other data or the modem/router is busy handling traffic you get zero dips on all plots (noise, speed and attenuation) but that usually lasts for only one or two samples max. and is confirmed by no change in sync speed or loss of "ppp" session.
The large dip at the end - again was a phone call made or received - not that a phone call will necessarily be the cause of the "spike", but sometimes can cause a bad connection to start playing up, but again I'd be listening to the line for noise on seeing these sort of dips.

[#13]

Plot 2 - Its not the central heating but it could be a washing machine - I will have to try to monitor the wife better to be sure. The evidence against would be that similar changes occur at all sorts of times - but that is only partially against as you would think we were running a laundry from the frequency that the washing machine is on.
I can be confident that the phone was not being used as the kids use mobiles all the time and the wife and I avoid the thing like the plague.
We did have dial tone during that gap on Plot 5. I am not clear why a reboot was necessary to restore the connection. The router was responding  on the internal network and replying happily to routerstats so most of it was happy. Its presumably possible that the modem part had gone into temporary meltdown.
I will try listening for noise and will try the house power down and will report back.
There are some astonishing similarities between mine and  Alex's problems on his thread and I think a few others. You seem to be being kept busy!