[AusNOG] DSL Line Stats

[Paul Brooks]

Usually its the SNR which is the cause (knob to tweak), and the sync speed is the effect.

Think of ADSL2+ as 512 parallel 56kbps audio modems, each operating at a slightly
different frequency - its effectively DWDM in the audio band.

When the modem and DSLAM first connect and go through the SHOWTIME negotiation, each
one measures the received signal level (and calculates the SNR) in each bin and then
reports these figures to the other end. They then negotiate and tell each other how
many bits-per-second they will try to stuff down each channel - high noise channels
(low SNR) are allocated a low bits-per-second datarate in order to cut through the
noise, low noise channels get more bits-per-second.

The lower the SNR floor is permitted to go by configuration, the more bits-per-second
the modem and DSLAM will try to push down each channel and the higher the achieved
aggregate data-rate will be. Of course this makes the connection more fragile, as a
small increase in noise will stop that high data-rate from being recognisable at the
far end, lots of data-errors occur, and the modem & DSLAM are forced to try to
re-calibrate, re-sync, and re-calculate how many bits-per-second to try to get down
each channel and still be recognisable at the far end.

Set a high SNR target figure, and then each channel is allocated a conservativly low
bits-per-second rate that is easily distinguished by the receiver, and the link should
be rock-solid.




On 1/10/2014 8:12 AM, Ross Wheeler wrote:
>
>> From my understanding - lower SNR targets can also improve sync speed hence why
>> people like iiNet offer it on the 'ThrillSeeker' kind of naming.
>> I may be wrong though.
>
> Cause, or effect though?
>
> If you push for higher sync speeds, the SNR becomes worse. Sad fact of reality.
>
> SNR is an outcome, not a cause, although some might argue that "low SNR is the cause
> of your problems", it's merely a metric to measure the margins between the signal
> (desirable) and the noise (undesirable).
>
> Of course, *I* might be wrong too :)
> _______________________________________________
> AusNOG mailing list
> AusNOG at lists.ausnog.net
> http://lists.ausnog.net/mailman/listinfo/ausnog