A high throughput adaptive DFE for HIPERLAN

This paper describes two methods for increasing the throughput of an adaptive decision feedback equaliser (DFE) using the LMS training algorithm. In the first method, a signed power-of-two number representation is used for the equaliser input data. Using this number representation, all multipliers can be replaced with barrel shifters and adders. In the second method, the delayed least mean square algorithm (DLMS) is used to train the equaliser. A delay, equal to the feedforward filter length, is introduced in the filter coefficient update, which allows the DFE to be realised as the cascade of a series of modular sections