BB and Noise Margin query

[#1]

I am baffled. After all this time I realise I still don't have a simple understandable model in my head of what exactly is happening with ADSL 2+ BB and the ISP/exchange line speed and noise margin settings.

I used to be with TalkTalk, had a TT DSL-2680 router and Routerstats Lite to monitor the line.
Now I am with Plusnet, use a TG582n router and DSLstats to monitor the line.

So more than one thing has changed, which doesn't help.

From memory (so possibly wrong!) with TT the router booted up and Routerstats showed download speen and noise margin. My memory is that the noise margin was set(?) at 6dB. As it was 'set' to 6dB it's plot was uninteresting - it simply flatlined. Of course, with random noise and errors on the line, something's got to give. What 'gave' was the download speed, so (from memory) the Routerstats plot of speed wiggled up and down around its initial speed established at initial connection. Now, to me, this makes sense. I can understand it.

Now, with Plusnet a different router and using DSLstats (and not from memory) I can't make sense of what I see. It is definitely the other way around. At initial connection a download (and upload) speed are established. From then on it is the speeds that flatline on the DSLstats graph and the SNR (Margin) that wobbles about. Now in one sense this makes sense, but how can it make sense in the context of a fixed exchange noise margin setting determining the speed based on the line condition? What am I missing?

Another thing, I tried this morning for a higher speed by rebooting the router on a Sunday morning. It worked to an extent - just not as much as I was hoping for - but what I always find strange is what happens to the plotted (download) noise margin. It started lower than usual, about 4.5 dB, corresponding to a somewhat higher than previous speed. OK so far. But then, as ever, while the upload margin stayed ~6 dB (with usual wiggles up and down) the download margin did what it always did. Over a period (with accompanying wiggles) it gracefully made its way up to meet the download margin at the 6 dB level. As usual, all speeds flatlining at the initial connection values.

So should I take the noise margin going from 4.5 - 6 dB as an improvement on the line, it isn't as noisy as expected initially so leading to a recorded greater margin? But why then should it home in on 6 dB? Why not 5 dB? Why not 7.7 dB depending on conditions? Is 6 db the download target? But then why not START at 6 dB and set a corresponding (lower) speed? And if it can establish a given speed at 4.5 dB margin why, if it can go to a higher (target) margin, can it not establish an even higher speed initially at 6 dB and stay with that?

I used to have a higher normal speed, with a 3 dB margin, and am really trying to get back to that(!) but, at present, am just trying to get an adequate model of how the system works. Can anyone help?

TIA

[#2]

TalkTalk are an LLU provider i.e they have their own equipment in the exchange and I believe they support Seamless Rate Adaption(SRA), which will explain the differences you are seeing between that and your PlusNet(BT Wholesale supplied) connection which does not.

To explain, a line will initially establish synch at a speed which gives the target noise margin, typically 6db. The inherent background noise will vary during the day and thus will affect the available noise margin . On a line with SRA the line will continuously adjust the speed to maintain the target. Without SRA, the speed will remain constant, and therefore the noise margin will vary, until the noise margin drops too far and the line resynchs at a new speed to get back to the target.

The BTw system MAY also periodically adjust the target margin, if the line is generating few errors it may drop the target down as far as 3db. However if error counts are high it will increase the target (and thus lower the attainable speed) to reduce errors.

Hope that helps

[#3]

Thanks for the reply, MisterW. I knew TalkTalk was a LLU provider and your explanation about the differences in service may just confirm that my memory is not totally faulty! So that seems to make sense.

However, when it comes to "a line will initially establish synch at a speed which gives the target noise margin, typically 6dB" - yes, that is what I would have expected. However, in my current experience, if DSLstats is to be believed, this is exactly what DOESN'T happen. At least on the downlink. Though it is what seems to happen on the uplink. This is what is mystifying me.

For instance, this morning, on starting the router, the download speed started at 16000 kBps, 4.5 dB Noise Margin. After a while, download NM curved smoothly (allowing for wiggles) upwards to join the upload NM at 6 dB. All speeds remaining absolutely fixed at the initial start up values during this time. This is what is puzzling me. And this is AFAIK the normal thing following a router start up - the download noise margin does NOT start off at 6 dB, but thereafter it gradually and continuously makes its way to 6 dB NM, where typically the upload NM always was from the start.

The screenshot included shows this happening.

[#4]

With BT Wholesale DLM the DSL equipment targets a certain signal to noise ration when the router initially connects. After that the speed remains fixed but the SNR can fluctuate with line conditions. As long as the line doesn't disconnect due to too many errors the connection speed will remain constant.

Initially the target SNR is 6db, if the line consistently has low error rates the equipment in the exchange can lower the target SNR, however this will not come into effect until the line is re-synchronized by the router disconnecting and reconnecting to the exchange.

The lower SNR will give you a higher connection speed from the point the router reconnects.

[#5]

So is the SN margin, increasing from an initial apparently arbitrary 4.5 dB to a 6 dB (apparent target?), a sign the line is actually noisier than expected (rather than quieter) when the initial connection was set up, at 4.5 dB? Meaning I ought now to expect a lower speed on a reconnection?

This implies the plot of SN Margin isn't a plot of current line conditions as such, rather a prediction of how, based on current line conditions, what SN Margin the line will try to connect at next time?

But that doesn't seem to tie up with the download SN Margin having been around the usual 6 dB last night when I disconnected, as it had been for some while. Yet it reconnected this morning at an initial 4.5 dB and a somewhat higher speed...

Perhaps being left off overnight messes it all up?

Or should I just ignore the SN Margin plot as pretty meaningless?

[#6]

However, when it comes to "a line will initially establish synch at a speed which gives the target noise margin, typically 6dB" - yes, that is what I would have expected. However, in my current experience, if DSLstats is to be believed, this is exactly what DOESN'T happen. At least on the downlink. Though it is what seems to happen on the uplink. This is what is mystifying me.

For instance, this morning, on starting the router, the download speed started at 16000 kBps, 4.5 dB Noise Margin. After a while, download NM curved smoothly (allowing for wiggles) upwards to join the upload NM at 6 dB. All speeds remaining absolutely fixed at the initial start up values during this time. This is what is puzzling me. And this is AFAIK the normal thing following a router start up - the download noise margin does NOT start off at 6 dB, but thereafter it gradually and continuously makes its way to 6 dB NM, where typically the upload NM always was from the start

Strange!  It MAY be that the initial synch is at 6db but drops to 4.5db before DSLStats starts to records it. Does the router event log show what the SNR was at the initial connection ? 

Perhaps being left off overnight messes it all up?

Definitely not a good idea to turn it off every night!. Leave it on all the time, except perhaps when away from home for more than a couple of days...