On the blind adaptation of an FSE+DFE combination

This paper shows that a fractionally-spaced equalizer (FSE) and decision feedback equalizer (DFE) combination, initialized from an open-eye condition, jointly adapted by a decision-directed algorithm, and in the absence of noise, is guaranteed to converge to the set of parameters minimizing the mean square recovery error for some overall system delay. This paper observes that when channel noise is present, the algorithm may behave erratically: for sufficiently high signal to noise ratios yielding at most 3% decision errors the algorithm may converge to the global minimum mean square error solution, while for sufficiently low signal to noise ratios the algorithm may converge to a "zero FSE, delay DFE" setting.