>>> wont the amount you get per db (when lowering SNRM) depend on your bit-loading
Yep.. exactly. The average 800kbps comes from BT's mouth which is why you will often find reference to that figure.
You are right though that adsl2+ lines will obviously see more benefit, because of more sub-channels available, but there are way too many variables for anyone ever to be able to quote a definite figure.
The hard facts are:-
~
3dB of SNR is needed to reliably load 1 bit~
data rate in the frame is 4,000 symbols per second - therefore approx 4kbps.The later figure is not quite so straight forward because various things come into play - 4kbps is near enough and commonly used without going into silly decimal places.In the real world adding 3dB of SNRM is highly unlikely to result in an increase in the bit loading of each and every one of the 256/512 sub-channels because:-
~ The lower frequencies may already be fully bit loaded.
~ Certain tones arent designated for data transmission and/or are OOB.
~ PSD masks affect the maximum bit loading in certain tones.
Not all lines are the same - BT use different masks depending on line length.. but the PSD mask prevents full bit loading up to about tone 85 on my own line which is about where I start to see full bit loading. ~ Certain tones may just not have enough SNR to load any bits regardless.
In the first 3 instances no amount of lowering your SNRM will have any difference to the overall data throughput.
Then theres variable things like bit-swapping 'reserving' a certain amount of SNR, whether the line is interleaved.. or rather the depth of error correction... and QAM -
dont ask me about that its beyond anything Ive ever gone into because Im not a mathematician and no good at equations.But you will find that lowering the SNRM by 3dB wont necessarily mean that all sub-channels magically get an extra 4kbps
By lowering the SNRM by 3dB:-
Short lines are most likely to see gains across the mid - higher frequencies.
Even very good and short lines still often see only 14 bits loaded in the lower frequencies for one or more of the reasons mentioned above.
Long lines are most likely to see a gain in the lower frequencies.. and it may extend the sub-channel range, but because longer lines are more susceptible to noise, they may not benefit too much at the later end of the lines range.
Even if in a perfect world a line was able to fully load each and every sub-channel
and none of the variables
nor a PSD mask were in force.. then the maximum would be*
adsl1 sub-channels 33 to 255 = 223 - 2 (pilot & nyquist) = 221 in use * 4kbps = 884 kbps
adsl2+ sub-channels.33 to 511 = 479 - 2 (pilot & nyquist) = 478 in use * 4kbps = 1912 kbps
*If Ive made an error in those calculations please let me know.