Decoding of severely filtered modulation codes using the M-algorithm

The decoding of signals in infinite intersymbol interference caused by severe filtering is considered. A state-variable representation of a linear system is used to define the decoder tree. Tree decoding is then performed by the M-algorithm. It is shown through simulations that performance close to the optimal receiver can be obtained while realizing significant computational savings. Binary phase-shift keying with rectangular baseband pulse shaping and continuous phase/frequency-shift-keyed signals are considered. The channel is modeled as a narrowband filter with two and four-pole Butterworth spectra with symbol-rate normalized 3-dB bandwidths of 0.1 and 0.25.<>