Data transmission error probabilities in the presence of low-frequency removal and noise

Upper bounds on error probability are derived for data transmission systems which are subjected to gaussian noise and to the removal of the low-frequency components of the signal. This error probability can be quite low for random data, even though the eye pattern is closed. Both standard format and partial response signaling are considered, as are binary and multilevel alphabets. Numerical results are given for a high-pass filter containing a single pole and for a cascade of several such identical filters.