The optimization of receivers constrained to consist of a (linear) filter and level slicer is considered for general data communication signals in additive white Gaussian noise. Special consideration is given to the two limiting cases of arbitrarily low and high signal-to-noise ratios (SNRs). The low-SNR case results in an average matched filter receiver; the high-SNR filter is shown to be the solution to a specific nonlinear mathematical problem with a simple geometric interpretation. An efficient algorithm is developed to solve the problem. The results are illustrated for a duobinary minimum-shift-keyed signal observed over an interval of three-symbol-period duration.>

In this paper we show that many spectrally efficient modified MSK schemes, termed generalized MSK, although not representable as OQPSK, may nevertheless be (suboptimally) demodulated using anI-Qreceiver with a proper choice of carrier-phase offset. Correlatively coded MSK schemes withI-Qreceivers are studied, and it is concluded that duobinary MSK and(1 + 2D + D^{2})/4MSK represent good performance-bandwidth tradeoffs among first- and second-order correlative coding polynomial schemes. The optimal design of these receivers are considered subject to the constraint of a finite duration impulse response, especially for asymptotic cases of arbitrarily small and large SNR. Filter design based on a zero-intersymbol interference constraint for PAM-based approximations of the signals is also considered. The optimized linearI-Qreceivers for(1 + D)/2MSK and(1 + D)^{2}/4MSK are presented. These receivers are only 0.28 and 1.24 dB poorer than the optimal (Viterbi) receivers at high SNR.

It is pointed out that the sufficiency portion of Theorem 3 in the above paper concerning the existence of discrete spectral components is not proved and, in fact, is not true. A counterexample is provided.

Many signals ocurring in data communications can be described by Markovian models. Expressions for the mean power spectral density function of a signal driven by a stationary discrete-parameter finite-state Markov chain are derived and discussed, both for irreducible and reducible chains. Two examples are presented to illustrate the application of the results.