On The Complexity of Joint Demodulation and Convolutional Decoding

This paper investigates the combined computational complexity of demodulation and decoding of QAM signals. Four combinations of demodulation and decoding techniques are compared in terms of bit error rate (BER) vs. signal-to-noise (SNR) behavior, finite word length effects, and hardware complexity. It is found that joint demodulation and decoding using a high-radix trellis can be more efficient for higher orders of modulation, while a decoding strategy which produces soft values, followed by a Viterbi decoder is more efficient for lower modulation orders. Complexity formulas that take into account word lengths and modulation order are introduced.