| Abstract: | This article presents a theoretical analysis (based on tight upper bounds on the error probability) of quantization loss with integer metrics used for convolutional decoding in the Rayleigh-fading channel. Optimum configurations with respect to the generalized cutoff rate criterion are established for 2-bit, 3-bit and 4-bit quantizers, and corresponding losses with both de facto industry-standard 1/2-rate and associated punctured 3/4-rate codes are evaluated. Assuming optimized thresholds, 4-bit metrics are shown to incur only a small quantization loss. However, results also indicate that the loss is sensitive to suboptimum threshold spacing. |
