Convergence properties of iterative soft onion peeling

The concept of iterative multiuser detection for a multiple access adder AWGN channel is coupled with turbo code used for each user, followed by an interleaver. The iterative decoding exchanges the likelihood information both among the two constituent codes of each user and among the users. The information exchange among users is performed for each chip, using a Gaussian approximation for other user's contribution. The described iterative decoding procedure, when applied to unequal user power case, implicitly implements the "onion peeling" concept-strong users converge first, their contribution to the residual noise decreases, and then the weaker users converge. This property, in focus here, was applied in previous work to illustrate applications to multiple access communication and to broadcast channels. The difference from the classical "onion peeling" is that rather than decoding a user completely and then subtracting its contribution, the convergence here occurs over several layers simultaneously. The convergence properties of the proposed method are studied, with special emphasis on the case of equal rate users with exponentially related powers.