I'm on a TI DSLAM, and my target SNRM is the usual 6dB, with the DLinks, the SNRM doesn't tend to drop much below that.
I notice the error counters go through the roof with the DLinks if the sync rate is increased slightly by DLM, so it seems that those devices are being pushed as far as they plus the line can go. I think this means that target SNRM tweaking is just right out of the question.
The only way I can see to go is to get a _better_ device, if such is even possible, rather than a foolish one. A device that could run at 4dB even, without error counts off the scale so as to trigger DLM anyway, now that would be interesting.
I can't of course do only 1 bit on a tone, that would require ADSL2. This is an area where I'm missing out, but failing breaking into the exchange and reflashing the DSLAM I'm going nowhere.
I wonder if a device with a better analog front end would be the way to go. An active input filter, that's the only route for possible improvement I can see. Reject more crap by cutting out the tones that are not in use anyway and lose no volts because it's active not passive. So a filter would have to be illegally narrow and tunable.
More gain on the input of course amplifies the noise, but perhaps an amplified _signal_ would give the electronics behind something juicier to get their teeth into.
It seems to me that if someone made a "hifi" quality variable-width active input filter, then this would be a generally useful bit of kit. Perhaps someone needs to make one. (Would need to be a pair poss, so that the US could be dealt with too properly.)