Companding turbo decoder

The memory requirement for the uniform quantization turbo decoder is very large. The nonuniform quantization provides better or identical performance than the uniform scheme. But it is inconvenient for the latter to implement with the hardware circuits. This paper proposes a companding turbo decoder, which is a combination of uniform and nonuniform quantization. Such companding decoder can be easily implemented with hardware and reduce at most about 30% memory as well as keep almost identical performance to the infinite quantization algorithm.