This work addresses the discriminator-based detection of convolutionally coded CPFSK. The discrete channel model encompasses the FM modulator, additive white Gaussian noise (AWGN) and a limiter discriminator detector. The noise samples at the receiver output are not Gaussian, and therefore the classical solution of noise whitening by linear filter does not guarantee statistical independence. This work examines an approach which hinges on finding special IF filters, guaranteeing a memoryless discrete channel. An overall memoryless channel dictates constraints on the IF filters, which can be formulated both in the time domain and in the frequency domain. The minimum normalized bandwidth of the IF filters satisfying the constraints is BT=3. The performance of these special filters is evaluated in terms of bounds on coded signal error probability and the generalized symmetric cutoff rate. The results of some representative cases are calculated and compared to those obtained by invoking an ideal interleaver/deinterleaver, which dictate a large delay. The results are demonstrated for a convolution rate 1/2 code with constraint length 7, where it is shown that the memoryless, interleaving free approach suffers a 3 dB degradation in power efficiency.
展开▼
