Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[MikeS]

I thought I understood, SNR, SNR margin but now I’m not so sure.  I treated SNR as an absolute number at a given point in time for my line, and SNR margin would be a fixed (6dB in my case) amount above the SNR figure to give some headroom for noise fluctuations which inevitably occur during the day.

Because of noise issues on my line I have been using routerstats to record SNR margin variations during the day with several different routers.  I have been trying to rationalise, (with not a lot of success) the results.  Looking back at the plots it is clear that for a given sync speed different routers come up with different noise margins. At roughly the same sync speed the Speedtouch and SAR 600E would report SNR margins 4 or 5dB apart during the noisy period but the same margin during the (relatively) noise free period.  So it seemed to me that SNR/SNR margins reported were not absolute numbers they were router dependent. 

I understand that the downstream ADSL stream is split into about 250 channels with a bandwidth of about 4kHz each.  When the router boots up all these channels are scanned for noise, if present the noisy channels are discarded and the (relatively) noise free ones are used for data transmission.  I guess buried in the router firmware is some algorithm which defines the choice of channels.  I’m not clear whether this channel choice is fixed until a resync event occurs or whether it is a dynamic process, nor whether the channel choice is responsible for the different value of SNR reported during the noisy period.

One of the oddities I noticed particularly with the SAR 600E, which would drop sync at -3 or -4 dB, was that the resyncs were often at only a slightly lower sync rate, whereas I was expecting a much bigger hit on sync rate as it was trying to get back from -4dB to the target margin of +6dB.  The only rationale I could come up with was that the ‘total noise’ had not increased that much but it’s frequency had shifted, so that previously noisy channels were now clear and previously clear channels were now noisy, with the net effect that usable channels were not changed that much.  This would only make sense if the channel choice was fixed at boot and held until sync was dropped. 

So the conclusion I am coming to is that the SNR reported by the router is a function of how it chooses to measure it rather than any absolute measure. Am I talking a load of rubbish, I would appreciate your comments.

[kitz]

SNR (Signal to Noise RAtio) fluctuates throughout the day depending on how much "background" Noise there on the adsl Signal.

>> and SNR margin would be a fixed (6dB in my case)

That would be the TARGET SNR - which is usually fixed and set by BTs DLM.

When you sync up the DLM goes "ok this line has a target SNR of 6dB...  therefore I can use whatever is left over to try and get more speed."
So depending upon your SNR at that moment in time all the rest of the frequencies are allocated.. the more channels that are allocated then the higher your speed can be.

>> different routers come up with different noise margins

Noise Margin is the "safety buffer"...
So if your target snr is 6dB..  when you first sync up the snr margin will record 6dB in line with the target.  For this example say you sync at 4000

But as the day goes on and the line fluctuates then the SNR changes.. so say for eg in the evening when more electrical stuff gets turned on then the margin will start to drop.
As it goes down 6,5,4,3, ..  now the signal is weak and you may start getting errors on the line as your router strugggles to hear the signal and has to say "pardon, I didnt quite catch that can you say it again".  Then it goes down further 3,2,1  and now the router is really struggling to "hear" the signal because theres too much background noise... 0.. and bang the router cant pick up the signal so it looses sync.

So the router resyncs - only this time because the SNR is worse, when it tries to sync at a 6dB target SNR, theres not as many available channel.. therefore it cant sync at the previous high speed.  So it say syncs at 3500 which gives the safety zone of 6dB.

The snr margin can still fluctuate.. and as the evening goes by and noise gets less then you can find that the SNR Margin now starts to go up 6,7,8, etc.

Some routers are better at handling lower SNRs than others and may say drop out at 4db..  whilst theres a few that will go to -ve figures.

>> The only rationale I could come up with was that the ‘total noise’ had not increased that much but it’s frequency had shifted, so that previously noisy channels were now clear and previously clear channels were now noisy, with the net effect that usable channels were not changed that much.  This would only make sense if the channel choice was fixed at boot and held until sync was dropped.

It will attempt to use all of the frequency bins allocated at sync.
When it resyncs at a lower speed those previous frequencies will no longer be available for use - hence the slower speed.

.. and the reason why if you sync at a low speed... when conditions are better you still stay connected at the same low speed. 

If you like you could think that anything over 6dB means that upon a resync then more frequency bins will be allocated (speed) at the next resync.

Theres also something called SRA (seemless rate adaption) where the router and DSLAM can allocate additional frequencies "on the fly" if theres spare SNR Margin.  However SRA isn't done on MAX.

---
Ive tried to simplify things - hope it makes sense... if not ask again.

[MikeS]

Hi Kitz,

Thanks for the reply.  I’m probably being thick but I still have some issues, particularly different routers come up with different noise margins.  At a given instant in time my line has a certain noise profile, or spectrum if you prefer it.  If you could connect several routers to that line at that specific point in time, would you expect them to all come up with the same noise margin, I know this is a theoretical question, but it is at the heart of my question.

What I have observed, over a number of different days, is that at a very similar sync speed (4200+/- 100) the routers are showing a 6dB margin in the ‘quiet’ phase of my line.  However once I’m into my noisy period different routers give quite different SNR measurements (4 or 5 dB apart). As I think it likely that the actual line noise spectrum encountered by both routers is similar, I am proposing that the actual line noise spectrum is interpreted differently by different routers, leading to different reported SNR margin measurements.

Hope this makes some sense to you

On the topic of ‘total noise and frequency shift (noise spectrum shift)’  If I give you an actual example it may be clearer.

This is with the SAR 600E.  Booted up at 9 am syncd at 3766 at the target 6dB margin, noisy period starts at 15.00 and SNR starts to drop.  At 16.30 it is down to -4dB and resyncs to 3744.  (don’t forget my target margin is 6dB).  It does briefly go back to 6dB but by 20.30 is back down to -3dB and resyncs to 3648 back up to to 6dB briefly.  Another resync at 23.30 from -4dB to 3616.  At 2.00am it resyncs from -3dB to 3808 (upwards !).  I’m now going into the less noisy period and sync is held at this level  with 6dB margin.  I don’t see any other way to explain why the resyncs are so trivial when you bear in mind that  clawing back somewhere in the region of 10dB you might well expect a resync some 1500kbps lower.

I have a feeling I’m not making myself very clear, and for that I apologise

[kitz]

>> I am proposing that the actual line noise spectrum is interpreted differently by different routers, leading to different reported SNR margin measurements.

Oh I think I get you you now -  I get different line stats depending which router I use too.  Its something I noticed years ago before maxdsl even came in.  Ive even noticed that updating the firmware on one of my routers made a difference too.

Why I dont honestly know ...  but it is something that several of us noticed .. and why on some lines some routers perform much better than others.

>> so trivial when you bear in mind that  clawing back somewhere in the region of 10dB you might well expect a resync some 1500kbps lower

Indeed it does seem strange.. I'd love to be able to give you an answer but I wont even attempt to explain something I dont understand myself and why some routers do this.. 

I dunno if anyone else has an idea that they may like to chip in?

[roseway]

I don't really know the answer, but I imagine that it's at least partly due to the way that noise margins are calculated by different routers. The figures reported by the router are averages (sort of) over the frequency range - different frequencies will have different levels of attenuation and noise, and each one will fluctuate over time. The reported figure is calculated by summing all these and giving more weighting to some than to others. Each router uses its own method, so the results differ.

[MikeS]

Not wishing to put words into anyones mouth, but we seem to be agreeing that the SNR margin reported is a composite of the 'real' SNR and the method used by the router to measure it.  My guess is that on clean lines there is little if no difference, but as the line gets noisier then differences start to appear.  Wonder whether or not router maufacturers actual test their kit on noisy lines, suspect not.

Interesting, particularly as SNR is such a central part of Broadband Max

[roseway]

I'm quite sure that router manufacturers do test their kit on noisy lines, and in all sorts of other ways too. But different manufacturers make different decisions on what their priorities are (e.g. stability or speed, or a compromise between the two, etc.) and the wide variations in line quality make it certain that the equipment will encounter situations which haven't been specifically tested for.

There's no such thing as the 'real' SNR by the way, except insofar as each frequency bin will have its own SNR (which will vary over time of course). There isn't as far as I know a standard method for calculating the weighted average, and different manufacturers use different methods.

[PhilT]

If you could connect several routers to that line at that specific point in time, would you expect them to all come up with the same noise margin

No. The variation in hardware may mean that the SNR is different. The variation in bits per bin allocation and in coding gain may mean that the SNR margin is different even at the same SNR.

When a G.dmt router syncs to a target margin it finds the fastest speed consistent with that speed by piling as many bits into each frequency bin as possible. If interference then comes along and reduces the SNR margin too far or creates too many errors then it resyncs and performs the same trick - this doesn't mean the speed has to fall for example an interference source that changes frequency may block out a different frequency next time round but have the same overall impact.

The SNR margin is defined as the excess SNR over that required to achieve a bit error rate of one in 10 million. The actual SNR might be 30 or 40 with an SNR margin of 6. The connection would work at an SNR margin of -2 but with more errors than 1 in 10 million, so it retrains.

HTH

Phil

[MikeS]

Thanks for that Phil.  That explains the behaviour of my SAR router.  It also I think implies that the noise frequency on my line is shifting, whilst the total noise is not varying that much.  Another thing is that it underlines that the choice of router for a line which is noisy is rather important.  For those of us on Max with longish, noisy lines there is very little information either from ISP’s or BT to guide us in this area.  It was only visiting this forum that made me aware of the issue, and it has been a long journey to resolve my problems.   

[konrado5]

I suspect overall SNR  is depended not only on SNR on particular tones but also on bit loading. As I get new synchronization I have SNR margin for upload about 4.6 dB but after 20 minutes I have about 6.8 dB. Before I get 6.8 dB I see about 80 bitswaps. Once I had SNR margin for upload either 102.3 dB or 0 dB. After 30 minutes I became reconnected. What is cause of unstable upload only on start of new synchronization?
EDIT: At new synchronization 1209 kbps my SNR margin is only 6.7 dB despire higher power output, lower synchronization rate and the same used tones. Is amount of bit loaded on any tone impacts on SNR margin?

[roseway]

konrado5 - You have been warned before, several times, about repeatedly asking the same question. Attaching your question to a 5-year-old thread will not achieve anything. I'm locking this thread and warning you again about your forum behaviour.