Channel Estimation and Adaptive M-QAM in Cognitive Radio Links

Cognitive radios have the ability to sense their RF environment and adapt their transmission parameters to perform optimally in any situation. Part of this involves selecting the best modulation type for a particular channel. In this paper we consider a variable-rate, variable-power, adaptive, m-ary quadrature amplitude modulation (M-QAM) scheme in a single-user communication scenario. The channel between the transmitter and receiver is assumed to be a Rayleigh block-fading channel. Each block is divided into training and data phases. During the training phase, the receiver estimates the channel and feeds the estimate back to the transmitter. During the data phase, the transmitter sends its message by adapting the size of the M- QAM constellation. We first find a closed-form expression that relates the bit error rate (BER) to the constellation size of the M-QAM, and therefore to the data rate of our system. Then, for a given target BER, we maximize the data rate over the training parameters, which are the training signal, the training duration, and the training power. When these optimum parameters are used in a MATLAB implementation, we find that the target BER is matched to within an order of magnitude, and the resulting data rate is close to the theoretical limit.