Is this the effect of Crosstalk?

Provider : BT FTTC
Modem : Zyxel VMG3925-B10B
Router : ASUS AC86U with Merlin firmware
Connection : Ethernet

I am on the end of a long rural line and have been noticing my Sync speeds slowly sagging over the past few months with a few modem resyncs, inevitably downwards. About six months ago I routinely saw a sync of around 9500 down. Over time it got worse to the current point where I sync at about 6400 and I am really struggling with even a single YouTube stream.

My TBB Speed Test this morning came in at 4.8Mbps down and 1.0Mbps up which is about the worst it has ever been. The BT Availability Checker for my line gives a range of 3.9 to 8.4 expected with a handback of 2.9 (and as the quality of the line sags, the handback value sags too).

http://www.dcoffey.co.uk/images/misc/BTChecker.jpg

Since I am using the Zyxel I am able to watch the line with tools such as DSLStats and I noticed that the Max Attainable was sagging at evening peak times with a corresponding drop in SNR. The modem was trying to make sure I still maintained a 6dB SNR at even the worst time of the evening which meant that I have been putting up with slow speeds but a very high SRN of 10dB to 11dB at off-peak times.

I watched the Max Attainable and SNR graphs for a full day and spotted something odd...

http://www.dcoffey.co.uk/images/misc/MaxAttainable.jpg
http://www.dcoffey.co.uk/images/misc/SNR.jpg

You will see that the day starts off at about 8500 Max Attainable and 10dB SNR. It remains very steady all throughout the day until about 8PM where it begins to sag. By 1AM it is down to a 7000 Max Attainable and around 7dB SNR (which is the low point so is probably why my Modem won't resync any higher throughout the day) but here is the oddity. It STAYS low until WAY after the peak time is over. It doesn't really start to recover until 4AM.

Is the BT peak period really that long?

Do you think this is simple crosstalk at peak time or is something else happening here?

I took the graph over a Saturday night but the weekly pattern is pretty much the same too.

Those graphs show a sag overnight and propogation of radio waves in the MF and HF bands is improved at night. This means a lot of stray radiation could be picked up by your line and thus cause an increase in noise. It is NOT down to BT peak usage periods.

Can you get some "bits per tone" plots taken at say 18:00, 21:00, 00:00, 03:00 amd 06:00. Putt them in a single image and see how they compare.

Hmm... Interference is not something I had considered.

It is a line that runs underground for the first kilometer or so then pops up and runs along the usual poles for another couple of kilometers. There are large overhead pylons in the vicinity as well as local overhead three-phase lines.

Can DSLStats record the Bits over time or do I have to manually take a screenshot at specific times?

I believe you need to do a screenshot - or DSL stats can do its own capture, I believe although not sure if it can do it timed. You can also add the SNR/tone to Bits/tone in general settings (from memory).

There is an optional Bitswaps per Minute graph and Bitswaps per tone graph as well as the usual Bitloading one. Oh, and I see SRN per tone too.

Which would be useful here?

Thanks for the tip - I have got DSLStats snapshotting at any interval I want. I can certainly do a 3-hourly snapshot without intervention but have I got the graph set up in a way that would be useful?

I ended up with this...

http://www.dcoffey.co.uk/images/misc/Bitloading-2019...

If you think this is useful I will leave it running overnight and we can see what it gets.

Whilst the OH peer stuff can provide noise, it is extremely unlikely to the cause of this, and certainly nothing you might get anything done about.

Do you have an SSFP on your NTE ? They have a built in RF filter.

Our socket is an Openreach Mk4 Master Socket 5C. Is that what you mean?

That is the graph - snapshots would be useful.

There really is something amiss with that already - from tone 290 at around 1.2 Mhz to to 520 at 2.2 Mhz there is a total void.

There are a lot of frequecy blocks arround there that are "free to use and unlicenced" including cordless phones, remote controls, ...

edit to add:

http://static.ofcom.org.uk/static/spectrum/fat.html and put 1 - 3 Mhz in te search

Edited by MHC (Sun 11-Aug-19 13:24:51)

Do you think there is a fault here then that might be remedied with an OR visit or are we stuck with it?

From talking to my neighbours, we all have similar speed test results but none of the others are bothered about improving it.

I will take three-hourly snapshots overnight and list them tomorrow.