With DSL, it depends on the channel frequencies which are allocated specifically for upstream and downstream.
These channels are all predefined as Standards to ensure that upstream and downstream channels are kept separate, otherwise it would cause horrendous problems with cross-talk... which would result in practically non existent adsl for short lines.
SDSL (2Mb) used separate DSLAMs to ensure that their upstream frequencies didnt cause problems with 'normal' adsl users... but became defunct really with the advent of adsl2+ with Annex_M, which provides 'up to' 2.2 Mbps upstream.
Spectrum Masks ensure that Annex_A + Annex_M can work from the same MSAN.. but this does mean that some of the higher upstream frequencies for Annex_M... and some of the lower downstream frequencies on Annex_A will never, ever see their full bit loading capabilities... and its why on some lines you often see the familiar inverted 'U' on bit-loading graphs.
You need a good + short line to be able to get the best out of Annex_M (I sync at around 2,600 kbps upstream)
See also
http://www.kitz.co.uk/adsl/adsl_technology.htm#dmt_modulationFor SME's.... I'd recommend looking at Be Pro (Annex_M)...
BUT it needs to be a good and short line + capable of being able to cope with the reduction in downstream tones. Forget it if the line cant sync at anymore than 8-10Mbps on normal adsl. :/