Print

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

[keithop]

Quick question.... possibly a dumb one but still.

Should attainable rate always be an independent, dynamic indication of line quality? ie. if I have a problem I'll see it drop but if I improve things, it should near-immediately show an improvement? or, as the ISP has suggested, does DLM affect it?

The situation for me is that I've had a stable 80/20 sync for a couple of years, attainable at around 95/30. Graphs to show it etc. Then a couple of weeks ago, I had a resync on a sunday down to 74/20 then 69/20 and finally 59/20 which it's been stuck at for 10 days or so now.

At the time, I wasn't home, nothing new plugged in or done, but when I found it, I started trying different router/different cable etc. which probably didn't please DLM overly. DLM put interleaving on at 10 and INP to 5. Then after a couple of days it reduce Interleaving/INP to 8/3, then finally 0/0.

So now it's been "stable" with int/inp 0/0, hardly any errored seconds, sync rate 59999/20000 and Attainable Rate:        62776 kbits/s/22173 kbits/s

Per Band Status:        D1      D2      D3      U0      U1      U2      U3
Line Attenuation(dB):   8.9     20.2    31.7    0.1     15.6    23.6    N/A
Signal Attenuation(dB): 11.5    20.0    31.7    0.1     15.6    23.6    N/A
Noise Margin(dB):       6.9     6.9     6.9     6.8     6.7     6.6     N/A

So if this was a new connection in its first few weeks, I'd have said well done DLM, figured out a nice stable speed. BUT it's been 80/20 for years so I've lost 20Mbs really quickly and there really has to be a reason for it

ISP has said just leave it, see what DLM does it couldve been a day of noise but DLM will eventually get there (could take weeks to reband higher) etc.

I'm ok with that, but the attainable rate worries me. Having fallen from 95mbs to 62mbs, if that's a near instant reading of "line quality" to some extent, then it implies that something is still bad on the line and I should be looking at the router/cabling/local interference etc trying to improve that figure before DLM will do anything ?   my ISP says it doesn't work like that, since DLM has limited the line to 60/20 that will have reduce the attainable rate as well.

That kinda feels wrong to me, I'd have expected attainable to shoot back up to 95 if it was once again a nice clean line but I don't know for sure?

Any ideas?

Thanks!

[skyeci]

I spose I would be looking at snrm graphs if you have them, perhaps you have been hit by a cross talker which is now affecting you which didn't have before? - seems like a large loss?

Are you on mydslstats? - if so what is your username so can have a look at historical data if poss

Cheers

[keithop]

a new connection crosstalking was my thought but it seems a lot to drop

attachments, SNR, syncrate ds, attaintable ds all over 30 day period, you can see the hit on may 8-9th and DLM's affect to try and stabilise.

Interestingly (to me anyway) attainable stayed high for a few more days before diving which I don't really understand. I'd have expected attainable to fall simultaneously with the sync drop not a few days into it. :/

This is why I wonder if attainable has somehow been "agreed" as a lower value rather than just a free to be whatever the line supports type of value.

[Ragnarok]

INP  ( best to think of this as impulse noise protection ) can affect attainable rate. it depends entirely on how the modem calculates this and if it takes INP into consideration especially the use of FEC.

DLM does not, It only limits the line rate ( the attainable rate stays the same)  or adds/remove INP when it decides it is necessary. Crosstalk is a unique type of noise without vectoring it is handled with regular INP.

[kitz]

Should attainable rate always be an independent, dynamic indication of line quality? ie. if I have a problem I'll see it drop but if I improve things, it should near-immediately show an improvement? or, as the ISP has suggested, does DLM affect it?

Its not so much a straight forward question.  DLM itself doesn't affect attainable rate... but things like Interleaving and Error Correction which can be switched on by DLM does.

Attainable rate is simply an estimation by the modem of what speeds could be acheived based on the current target SNR.  When you switch on Interleaving - or more correctly FEC - this affects overheads which are required to transmist the redundant data.  The affect of this, these overheads skews attainable rate.   It entirely depends on the amount of INP, but a very rough (and I must stress rough) guide is to halve the difference between actual sync and attainable rate to get a better indication.  There will be a formula based on the amount of redundancy (these are things 'R' & 'N' values that you will sometimes see myself and wombat harp on about) but unless you want to get into the realms of some very complicated maths then its easier to say 'about half'.  The higher the R to N value, then the more overheads there will be in use.

Re the sudden loss, that is more likely to be crosstalk   Most of us here have lost on average about 20Mbps to crosstalk.  Some a lot more...  unfortunately there isnt anything BT can do about crosstalk which affects all lines (ADSL & VDSL) until vectoring is implemented.

[Moderator edited to correct a spelling mistake in the very last sentence.] 

[ejs]

I don't think it's quite that the overheads "skew" the attainable rate. It's because the attainable net data rate is loosely defined, and that it only has to take into account some of the line parameters, but others can be ignored. The basic attainable net data rate method only has to take into account the target SNRM, the maximum latency, and "all coding gains available (e.g., trellis coding, FEC)" within that maximum latency. So the attainable net data rate could be calculated using interleaving/FEC levels to provide maximum coding gain, not to provide the level of INP that's actually been specified. The actual level of INP does not need to be factored in for the attainable rate calculation.

Essentially, the attainable net data rate can be calculated using different parameters (different levels of FEC/interleaving) than what the line is actually using.

There's also an optional improved attainable net data rate method specified in the VDSL2 G.993.2 document, which has to take into account far more parameters.

[kitz]

>>  not to provide the level of INP that's actually been specified. The actual level of INP does not need to be factored in for the attainable rate calculation.

Perhaps I didnt explain myself properly     By looking at the INP value its sometimes easier to see how much Error Protection has been applied.
You are correct that its FEC overhead etc causing it, but its my understanding that its actually the INP value which controls how much Error Protection is applied.  So for example INP 3 is low interleave and a small amount of redundancy ie recovery from 3 DMT symbols.   INP 4 increases redundancy to be able to recover from the duration of 4 DMT symbols and so on.  Therefore when I look at the INP figure it gives me a quick impression of how much error protection is in use by DLM.   

[ejs]

Yes, that's true, but my point was that's not really why the attainable rate can become so different to the actual rate achieved (besides banding).

The attainable rate calculation does not need to use the INP value (not very many parameters were originally defined for the attainable rate in G.993.2, leaving much of the calculation undefined vendor-specific behaviour). The actual net data rate, of course, will depend on the INP and all the other parameters.

The attainable net data rate, is "net of" the FEC overheads (they've already been subtracted), otherwise it wouldn't be a net data rate. The attainable net data rate could be calculated on the basis of what could be achieved using a different amount of FEC, it does not have to be calculated using the actual FEC level required to provide the INP value.

And that could be why the attainable net data rate becomes so different from the actual net data rate. The actual net data rate might be what's left after much greater FEC overheads have been subtracted, but the attainable net data rate can be calculated using a different FEC level with lower overheads.

[keithop]

Thanks for the replies, that's really interesting reading.

So attainable isn't perhaps as definitive as I once believed, rather than indication of what the MODEM side (my side?) thinks it should be able to achieve if it uses the headroom it has (but very possibly ignores some things).

So in my case, a suddenly much lower attainable probably does point to something new, noise probably, rather than it saying for example "well I'm at 60/20 now and there's only about 6.9 margin left so I could attain 62" but not realising that DLM has capped it at 60/20 anyway.

And I'm right in thinking attainable is moment to moment? So if I were to do something like turn everything off, new faceplate/filter, shorter new shielded dsl cable to modem, then turn it back on. Profiling would still be in place BUT if my work did any good, attainable should pretty instantly be better?

[ejs]

Yes, I was thinking about what I'd written a little more, the DLM can affect the attainable rate, but I think it's more likely to make the attainable rate become much higher than what your line will actually connect at (even if it weren't banded).

From the graphs you posted, the attainable rate stayed fairly high while the SNRM was high, and then fell when the SNRM fell. It does look like that currently, even if your line weren't banded, it wouldn't go much higher than 60Mb anyway. Yes, I would expect the attainable rate to increase if the line quality or environmental conditions improve.

[Chrysalis]

depends what you mean by DLM.

If you mean also the power management, then possibly yes, on my upstream the signal power varies a bit, it doesnt vary whilst synced, but there seems to be a variance to the power level set on a new sync event.  The snrm can be as low as about 9db, or as high as 13db.  Which affects the reported attainable speed by about 1.5mbit.  This may just be a variance of the UPBO also.

In regards to the normal parameters it changes it can affect the "reported" attainable but actual line capability wont change.