Cascaded equalizer technique

Abstract

Equipment-generated and dynamic crosstalk are effectively compensated by a signal equalization technique which involves the cascading of equalizers of respectively different lengths and respectively different loop gains. The first equalizer to which a received data sequence is applied is a long equalizer having a large number of stages, the term long being relative to the time dispersion of the high frequency channel. The gain of this long equalizer is adjusted to a low value that is consistent with the basic acquisition requirements of the transmission system, but is such that it averages through rapid variations of the high frequency channel. The second equalizer is short relative to the first but is sufficiently long to cover the span of the time dispersion of the high frequency channel. The first equalizer includes a data storage device which receives and stores successively received data samples. Each data sample is multiplied by a controllable weighting factor and the weighted samples are summed to provide an estimate of a received data signal. This estimate is then supplied to the second equalizer which may either be of the form of a transversal equalizer, like the first, or a decision-configured equalizer. At the output of the second equalizer there is obtained a second estimate which is compared with a reference value to form an error signal. The error signal is correlated with the data stored in the first and second equalizers in order to update the weighting factors of the respective equalizer. The reference value can be a preset value or one determined from the data. The gain factor for correlating the error signal with the data stored in the second equalizer is considerably greater (at least an order of magnitude and preferably two or more orders of magnitude) than that for the first equalizer.