Optimal signaling techniques for crosstalk-dominated communication channels
Gaikwad, Rohit Vijay
Baraniuk, Richard G.
Doctor of Philosophy thesis
In many communication systems, multiple users share a precious resource---transmission bandwidth. Proximity of user channels and suboptimal usage of transmission bandwidth leads to significant multiuser interference or crosstalk. Crosstalk can severely limit achievable bit rates. In this thesis, we design optimal signaling strategies utilizing crosstalk avoidance techniques to maximize data rates. We present a framework for maximizing the capacity (or bit rate) of symmetric and asymmetric bit-rate communication services dominated by crosstalk, in particular Digital Subscriber Line (DSL) services. We assume Gaussian noise and crosstalk characteristics, and hence second order statistics are sufficient to characterize the channel. This implies optimal transmit power spectral density (PSD) design is sufficient to maximize bit rates. Using the channel, noise, and crosstalk transfer functions, we set up optimization problems and solve for transmit spectra that maximize the joint capacity of same-service users. Joint signaling techniques and optimal power distribution yield significant gains in bit rates (or performance margins) over current schemes. Furthermore, by design the optimal transmit PSDs are spectrally compatible with existing services on neighboring lines. Our framework does not depend on the choice of modulation scheme and is extremely simple and of low computational complexity. Besides DSLs, our approach applies to wireless, coaxial cable, power line, and geophysical well-logging telemetry channels.
Engineering, Electronics and Electrical