Coding, Modulation and Multiuser Decoding for DS-CDMA Systems
Doctoral thesis, 1997
The general trend both in wireless and fixed network communication systems is flexibility in services, data rates, desired delay and performance. Future systems should not be designed for specific services but be more general and allow for a multitude of services, both now existing and those not yet invented. Considered in this thesis is the design of a flexible direct-sequence code-division (DS-CDMA) system, supporting many services, and using advanced receiver structures and coding techniques.
The first part focus on modulation methods to achieve multiple data rates in a DS-CDMA system. Considered are the multi-processing-gain, the multi-modulation, and the multicode schemes. It is shown that the multicode and multi-processing-gain schemes have about the same performance, while the multi-modulation scheme performs significantly worse for high-rate users. Disadvantages are: for the multi-processing-gain the low suppression of external interference for the high-rate users; and for the multicode scheme the high variations in the envelope for high-rate users which results in decreased spectral efficiency, increased bit error rate, and less power efficient amplifiers. The envelope variations can, however, be significantly reduced by the introduction of a high-rate non-linear block code --- a precoder.
To further increase the the number of possible source data rates, a rate matching scheme using a family of rate-compatible convolutional codes is presented. The codes are variable rate codes all decodeable by a single decoder, and almost as good as the best known fixed rate codes. Other possible applications for the codes are unequal error protection and hybrid ARQ.
Multiuser detection is a vital part in a DS-CDMA system to reduce the multiple-access interference and the near-far effect. Extensive simulations, however, show that errors in the estimates of channel parameters such as propagation delays and carrier phases drastically reduce the near-far resistance. Also treated are joint receiver designs. The derived receivers are jointly optimized source, channel, and multiuser decoders, where it is assumed that the source encoders are vector quantizers. Performance evaluations show significant performance gains with respect to separately optimized receivers, using the optimum multiuser detector in tandem with the optimum channel and source decoders. Furthermore the presented decoders are soft-in, soft-out decoders that directly map the matched filter outputs onto the source space. The mapping is minimum mean-squared error (MMSE) optimum, thus not containing any decision devices or other information destroying functions.