Method and apparatus for implementing the training phase of adaptive channel equalization on a digital communications path

The invention relates to a method and apparatus for implementing the training phase of adaptive channel equalization on a digital communications path passing signals modulated using QAM or CAP in a system configuration where the transmitter or the receiver or both of them include bandstop filters that form at least one stop band on the signal spectrum. In the method, the training of the adaptive equalizers is performed utilizing a sparse symbol constellation, wherein the number of symbol states (S) is reduced from that of the symbol constellation used in the normal data transmission state. The benefit of the sparse symbol constellation is that the strong intersymbol interference caused by a stop band falling on the signal spectrum can thus be reduced to such a low level in regard to the signal power that a linear equalizer may be used to reach a situation, wherein the proportion of correct symbol recovery decisions becomes sufficiently high for the training of a decision-feedback equalizer.