1. We would like to analyze the performance of reverse-link CDMA with limited feedback. The receiver (base station), which is assumed to know channel gains and interfering signatures, can compute the optimal signature (i.e., maximizing the signal-to-interference plus noise ratio (SINR)) for a user. With finite feedback, the receiver needs to quantize the signature. The SINR for the user with quantized signature depends on the quantizer and number of available feedback bits. We proposed a Random Vector Quantization (RVQ) in previous work and showed that RVQ is the optimal quantization scheme in a large system limit (number of interfering users, processing gain, and number of feedback bits tend to infinity with fixed ratios). We presented the approximation for the large system SINR with quantized signature. Here we plan to derive the exact large system SINR with matched filter and MMSE receivers as a function of normalized feedback bits for both ideal and flat fading channels. The large system results are expected to be a good approximation for the performance of finite-size system of interest. The results will tell us how much feedback is approximately needed to achieve a target SINR.
2. RVQ was shown to be optimal in the large system limit, but requires an exhaustive search. We plan to propose a less complex codebook. Using a tree structure, the number of searches needed to locate the optimal signature is reduced dramatically. However, the associated performance is suboptimal since the selected signature from the tree-structured RVQ codebook may not be the optimal signature. We plan to compare the performance of the two codebooks and analyze the performance degradation.
