How does twisted-pair improve DSL?
The extension to that question is does TP improve DSL over long runs? The reason I ask is because when I plug my DSL modem inside my home, I get a 1/3 speed reduction over when I plug it directly into the demarc. It's been recommended that I run Cat-5 from my demarc into my house, into which I'll plug my modem. I thought that this would be because TP ensures that the impedance is consistent, which would allow the signal to travel over large distances, but according to the Wikipedia page TP only protects against RFI. There isn't a lot of RF where I live, and the telephone cable carrying the DSL doesn't have a lot of neighbours, so I'm not sure how this is going to help.
The bottom-line for me is that my demarc is miles from the exchange, so how does the last 40' of my cheap telephone cable degrade the signal so much and how would a replacement cable make those last 40' easier?
I should also point out that I'm running a dry loop so the telephones and their network are completely isolated from the DSL and the demarc. There is absolutely nothing else connected to the DSL line.
There are two separate issues here:
Twisted pair cabling, along with differential signaling, allows RF interference to be cancelled out.
Running a dedicated line of new high-quality copper (even if it weren't twisted pair) from your demarc to your DSL modem, and separating it from your legacy in-home voice telephone wiring by putting a filter right in your demarc, ensures the shortest wiring run and isolates your DSL signal path from all the noise and reflections of your legacy voice wiring.
I doubt that any of this has anything to do with impedance.