Puzzle with my SNR margin (ADSL2+)

I've noticed some strange phenomenon on my ADSL2+ connection. It have especially occurred on previous DSLAM (my ISP recently changed DSLAM). I've perhaps written about it some time ago but I probably haven't explained this with sufficient accuracy.

First issue: during 10 minutes period if I was continously connected my SNR margin was continously 6.1 dB. When I was doing many reconnectings, I got either 17800 kbps at SNR margin 6.1 dB or 17100 kbps at SNR margin 6.3 dB. The power output was the same on boths cases. It was happening after my ISP raised cap from about 19800 kbps to 25000 kbps

Second issue: on upstream I had 1277 kbps snr margin 7.0 dB, power output 11.3-12.1 dBm and I rebooted the router and I got 1261 kbps snr margin 6.4 dB and power output 12.4 dBm. It would be not strange if power output was higher at first example.

For upload it would be good explanation: sometimes power cut back causes that starting tone is disabled (when remaining tones are very good bit-loaded) and power output measurement is average across only used tones. Bad for download, however, disabling one tone even with maximal loading 15 bits cause lowering synchronization rate about merely 60 kbps.

I will be grateful if someone give me good explanation. How is it possible that at time I have very stable SNR margin reconnection causes so much changes? Unfortunately I couldn't check SNR on tones and bit-loading at that time.

Perhaps it will be helpful: at normal synchronization rate for download I have always immediately after synchronization establishment 6.1 dB (both on old and new DSLAM), however at strangely lowered synchronization I had always 6.3 dB SNR margin. For upstream I've had always about 6.6 dB at lowered synchronization rate.

Best regards
konrado5

I see no issue in your 2 'issues'. There is no significant difference between any of your measures, NM or speed. ADSL is not a perfect technology. It is expected that the NM will wander about slightly during a connection session and for the speed & NM to vary slightly from 1 connection to the next. All due to unpredictable ambient noise.

Thank you very much for reply. I'm afraid that I probably haven't explained this plainly. I know it is normal on ADSL that SNR margin changes. However I know also SNR margin helps to predict new synchronization rate after reconnecting unless power output, used tones won't be changed. However I can have one synchronization 1261 kbps with SNR margin not higher than 6.6 dB and other 1277 kbps with SNR margin not lower than 7.0 dB. SNR (not only SNR margin) changes more between different synchronizations than during keeping connection.

Too bad when that was case I didn't know how to check SNR on particular tones and bit loading. However I know on download my SNR margin was stable 6.1 dB during 5 minutes period unless I don't reconnect. It implicates there wasn't much changes to disable many tones after establishing new synchronization, Disabling one tone with maximal loading 15 bits cause lowering synchronization rate about merely 60 kbps.

Best regards
konrado5

Edited by konrado5 (Tue 04-Feb-14 11:38:55)

"1261 kbps with SNR margin not higher than 6.6 dB and other 1277 kbps with SNR margin not lower than 7.0 dB."

Don't take this personally, but while I see the difference and I would be worried if it was my bank account, in terms of ADSL2+ those figures are so close as to be considered the same.

Put it this way - you could sit down and do the maths and analysis by hand on all 512 individual frequency bins used by ADSL2+ but it would be wrong again as the noise level will have varied ever so slightly.

In reply to a post by MrSaffron:

as the noise level will have varied ever so slightly.

In the time it takes to do the calculation !

We are almost into Quantum Physics here; the observer perturbs the measures .

Note I did say BY HAND - and with 512 bins to calculate I don't believe my head can calculate that quickly.

yes at the speed the modem operates it is fast enough to usually get it right.

What do you say about issue: I have about 17800 kbps, continously SNR margin 6.1 dB, I reconnect: I get 17200 kbps (SNR margin 6.3 dB, not 6.1 dB), I reconnect once again, I get about 17800 kbps? If at the time period SNR margin is the same, synchronization rate differences are typically about 100 kbps, not 600 kbps. If I had haven before reconnection SNR margin 5.0 dB it was normal: 17200 kbps with SNR margin 6.3 dB (but not if I have 6.1 dB). The power output is the same on both connections.

I don't complain about my connection. I have only crux. I'm very curious about ADSL2+ technology details.

Best regards
konrado5

Edited by konrado5 (Tue 04-Feb-14 13:28:45)

Within the realms of reasonable variation.

Without knowing what the bin/bit loading was for every tone it is impossible to know if there is something like a once every 5 second pulse giving you this nice sort of random variation.

If wanting to worry about small variations, then it is analysis of very bin that needs to be done to try and identify if there is a pattern or not.

Thank you very much for reply. I suspect also this is related with particular tones. However it would require disabling pretty many tones and it would be strange when SNR margin was the same at so much changes of SNR on some tones. Disabling one tone even with full bit-loading causes merely 60 kbps lower synchronization rate.

Recently my ISP rebooted DSLAM. I had haven synchronization rate about 15400 kbps/1221 kbps and I've got 16248 kbps/1245 kbps with SNR margi for download 3.3-4.0 dB and SNR margin for upload 5.8-6.3 dB. When I had 1221 kbps I had 7th tone enabled. Now I have disabled this tone despire of higher synchronization rate (probably all people was disconnected at one). Is it possible that power cut back works thus starting tone is disabled if further tones have high bit-loading (if there are very good conditions).

Following bit-allocation for 1221 kbps.

Tone number Bit Allocation
0 0
1 0
2 0
3 0
4 0
5 0
6 0
7 7
8 9
9 12
10 12
11 13
12 14
13 14
14 14
15 14
16 14
17 14
18 15
19 15
20 15
21 15
22 15
23 14
24 14
25 14
26 14
27 13
28 13
29 12
30 10

Following bit allocation for 1245 kbps.

Tone number Bit Allocation
0 0
1 0
2 0
3 0
4 0
5 0
6 0
7 0
8 10
9 13
10 13
11 14
12 14
13 15
14 15
15 15
16 15
17 15
18 15
19 15
20 15
21 15
22 15
23 15
24 14
25 14
26 13
27 13
28 13
29 12
30 11
31 11

Best regards
konrado5

Edited by konrado5 (Tue 04-Feb-14 16:04:23)

And?

I am lost, you appear to know all about being able to disable tones, what you are asking for it appears is how does the modem calculate to allocate the bits per bin, and then how does it arrive at a single SNR and power figure, when this has actually being done individually across lots of bins.

In effect this is a day long lecture from an advanced RF/Digital electronics course.

Best bet if curious, find some open source modem code that will show the maths involved.

Too bad when that was case I didn't know how to check SNR on particular tones and bit loading. Perhaps I would have answer. However, I've noticed once also strange change on upload despire of the same bit-loading and the same power output. I've found other people have similar issues:
http://forum.kitz.co.uk/index.php?topic=13083.0

Best regards
konrado5

I guess you are not in UK. If your country's ISPs are anything like ours then the sync-time Target NM is almost invariably a multiple of 3dB, e.g. 6dB it looks like in your case. The sync speed for that connection session is determined by that sync-time Target NM. What you seem to be forgetting is that by the time the modem has disconnected and reconnected, and you have had time to login to its GUI and view its stats the NM has moved on due to noise variation. In other words, all the NMs you are quoting are not the sync-time Target NMs that gave rise to the sync-speeds you are seeing and so you can't relate them in the way you are attempting to.

You just can't be that precise!

All you say about bins, tones and powers is beyond me and I believe what you are asking calls for a deep theoretical thesis on DSL technology and I don't believe that this is the forum to supply that.

In reply to a post by XRaySpeX:

I guess you are not in UK. If your country's ISPs are anything like ours then the sync-time Target NM is almost invariably a multiple of 3dB, e.g. 6dB it looks like in your case.

It's a case. Target SNR margin in my ISP is 6 dB.

The sync speed for that connection session is determined by that sync-time Target NM.

I know it.

What you seem to be forgetting is that by the time the modem has disconnected and reconnected, and you have had time to login to its GUI and view its stats the NM has moved on due to noise variation.

I did reconections during flat SNR period for example 10 minutes period. Moreover, as I've had 1241 kbps upstream, my SNR margin was continously no more than 6.6 dB during very long time. Suddenly, as I've rebooted router I've got higher synchronization rate and SNR margin despire lower power output measurement. Similar case was for download, as my ISP changed cap from about 18800 kpbs to 25000 kbps. During stable SNR margin period I've had either 17200 kbps with SNR margin 6.3 dB or 17800 kbps with 6.1 dB SNR margin. Now i know how to check SNR on tones, however on new DSLAM I don't see this case. I have always 6.1 dB immediately after synchronization rate and during 10 minutes period differences now higher than 100 kbps.

All you say about bins, tones and powers is beyond me and I believe what you are asking calls for a deep theoretical thesis on DSL technology and I don't believe that this is the forum to supply that.

Could you recommend any other forum where people know about details?

Best regards
konrado5

Edited by konrado5 (Tue 04-Feb-14 17:16:09)

For all anyone knows the figures displayed may have rounding errors and mistakes in the display maths - plenty of instances where firmware says an impossible set of figures but people have a good connection.

But without snr values and attenuation for each bin it is hard to say. Here is another puzzle...

Some routers supply the SNR/attenuation values based on a single tone in either direction.
Some routers average it out across the bins to arrive at a consensus value.

Most of the time the two are very similar, and are irrelevant to how each bin loading is calculated.

Next step beyond here is the people who work on the micro-code and standards e.g.

http://www.broadband-forum.org/

Among router manufacturers I am willing to bet many are just using reference code and do not actually understand the full mathematics and oddities that RF can throw at the DSL chipset.

In reply to a post by MrSaffron:

For all anyone knows the figures displayed may have rounding errors and mistakes in the display maths - plenty of instances where firmware says an impossible set of figures but people have a good connection.

I think it wasn't calculation error. As I've had strangely lowered synchronization rate and SNR margin I've had also more errored seconds. I suspect I've had then many tones disabled but why these tones would be disabled? It would require siginificant SNR drop on many tones and in turn it would be noticeable on SNR margin.

Some routers supply the SNR/attenuation values based on a single tone in either direction.
Some routers average it out across the bins to arrive at a consensus value.

I've found something about bit amount calculation for every tone (page 204).
http://books.google.pl/books?id=m77kZl71gysC&printse...

It seems attainable synchronization rate is not depended on the way router measure SNR margin, it is depended only on SNR on particular tones.

How is power output value calculated? Is it average across all used tones power?

Moreover, I've written other post:
http://forums.thinkbroadband.com/technical/f/4302367...

Best regards
konrado5

Edited by konrado5 (Tue 04-Feb-14 19:40:34)

Yes, that seems the right sort of forum for these sort of deep considerations.