Matched filter decoding of random binary and Gaussian codes in broadband Gaussian channel

In this paper we consider the additive white Gaussian noise channel with an average input power constraint in the power-limited regime. A well-known result in information theory states that the capacity of this channel can be achieved by random Gaussian coding with analog quadrature amplitude modulation (QAM). In practical applications, however, discrete binary channel codes with digital modulation are most often employed. We analyze the matched filter decoding error probability in random binary and Gaussian coding setups in the wide bandwidth regime, and show that the performance in the two cases is surprisingly similar without explicit adaptation of the codeword construction to the modulation. The result also holds for the multiple access and the broadcast Gaussian channels, when signal-to-noise ratio is low. Moreover, the two modulations can be even mixed together in a single codeword resulting in a hybrid modulation with asymptotically close decoding behavior. In this sense the matched filter decoder demonstrates the performance that is largely insensitive to the choice of binary versus Gaussian modulation.