Kitz Forum
Chat => Tech Chat => Topic started by: Weaver on May 26, 2018, 05:58:42 AM
-
A really daft and only half-thought through idea, so do not bite. :-) I wonder if a modem could be designed which could listen to an extra input signal taken from a ‘quiet’ copper pair that is in the same bundle/main cable as your existing ‘signal’ DSL copper pair, that is, from the same bundle that is running to a group of properties.
The idea is that the modem could pick up a reference copy of all interference noise and some crosstalk. I am wondering if something very clever could be done regarding noise cancellation then? G.Vector is done properly, I would say, and this isn't. Is it meaningful / theoretically possible?
Regarding crosstalk, the quiet pair would have different neighbour pairs from those heard by your ‘signal’ pair so that could be a problem.
-
I think using a different pair of wires is not going to work because the detected noise would include crosstalk from your own line.
The Broadcom 63138 is supposed to include Reference Noise Cancellation (https://www.prnewswire.com/news-releases/broadcom-adds-advanced-noise-cancellation-technique-to-dsl-gateways-300161290.html). So the general idea of increased noise cancellation processing may have already been done, but without using a spare pair. There doesn't appear to be many people switching to 63138-based devices and telling everyone how much better they are, so perhaps it doesn't achieve very much, and/or relatively few people have compared them perhaps because the devices aren't so common yet.
-
I found this: http://www.cs.cmu.edu/~aarti/pubs/ANC.pdf
It seems to me that one alternative would be to have an antenna instead of using a wire out of another copper pair. That would not be as good because it would not be so realistic, but I wonder if this could be compensated for by a suitable transformation, based on the known characteristics of the antenna. If you know additional information about the phone line then that could be incorporated too to tweak the transformation it seems to me. Most but not necessarily all of this latter info could come from one-off measurements - such as when gathering info during training.
I also wonder if the main copper pair could be used to gather information whenever there is a brief period of total inactivity in both directions?