Error performance of the blind decision feedback equalizer using decorrelation

The decision feedback equalizer has the advantage of compensating for channel amplitude distortion without enhancing noise. Conventional decision feedback equalizers require a training sequence to help them adapt to unknown channel characteristics. Training sequences are not always appropriate. In such situations blind equalizers are more suitable. The blind decision feedback equalizer inherits from the conventional the advantage of not enhancing noise, while compensating for severe ISI. It is thus superior over the blind linear equalizer. On the other hand the blind decision feedback equalizer also suffers from error propagation. In this paper we study the transient behavior of the blind DFE. This is done by deriving an upper bound on the probability of error due to residual ISI and show that it vanishes for the noiseless case. We also determine lower and upper bounds on the probability of error for the additive white Gaussian noise case. The upper bound obtained is the same as that of Duttweiler, et al. (1974). Using a different approach from that of Duttweiler, et al., we derive a lower bound. This bound is tighter than tile “no ISI” lower bound