Round 1

20/06/2017 – Line 1
Noise	Sync

 

26/06/2017 – Line 2
Sync	SNR
Down
Stable at 16Mb/s, with a…	…margin of around 9db. Very nice.
Up
Very strange! On a line that is (apparently) measured correctly on the downstream, the DLM is assessing the upstream independently, judging the same copper wire as only capable of 500kb/s…	…and requiring a margin of an enormous 31db! I’m wondering if the DLM removal has worked on this line – it’s suspiciously well-behaved, with noise of the level of natural background EM interference.

Meanwhile, on the same bundle, and with exactly the same conditions, independently of the router used:  

26/06/2017 – Line 1
Sync	SNR
Down
Drop outs all weekend. This is Monday morning.	After a drop-out, the DLM selects a low SNR to recover, then raises it (presumably through power or bin selection – don’t know). This is the wrong way around. Baffling.
Up
	No explanation for this. Further evidence that upstream and downstream are managed separately.

It’s worth noting the following graphs of noise margins on the two lines over the same time period, complete with drop-out:

LINE	DOWN	UP
1
2

 It is clear that the remaining difficulties on ‘Line 1’ are not environmental, as has been consistently suggested. These two lines are brought to the property in the same bundle, and Line 1 has been checked by openreach. I also swap the routers once in a while, just to rule that out. Line 2 has stabilised since last week, notwithstanding that it may just be the temporary result of a DLM reset. We will see.

A TT line at a different property, 4 miles away.

These drop-outs are clearly due to reducing the SNR. For casual use, this might not matter too much, but for anything that relies on a consistent connection, such as security cameras, remote access or access to a NAS, it makes the service completely useless.

DLM has finally been removed, after 3 years. This is what a healthy line looks like:
DS Sync	US Sync
DS Noise	US Noise

Very stable for four days – perhaps longer.

 

This certainly seems conclusive – DLM has either been removed or fixed. Clearly, the decision to push lines to 3db without considering their individual suitability was a mistake.

There still remains some concern about why the DLM behaved so badly in previous months, when the ds noise target was still 6db. Until this week, it was repeatedly pushing the line(s) faster and faster, irrespective of nominal noise target, again and again, through multiple DLM resets over a period of about 40 months.

30-06-17
Revieved call from Talk Talk – DLM has indeed been removed. Confirmation scheduled for next week to confirm the change has stuck.