Comparison between SOS and HOS methods for blind equalization of microwave radio channels

We present a comparison study between three SOS-based methods, namely the least squares, subspace and cross-spectrum, and one recently proposed HOS-based technique, for the blind linear equalization of channels measured in actual microwave radio experiments. The algorithms considered are used from a cold start-up, until the "eye" is sufficiently open, to enable transfer to a decision-directed mode of operation. Our study shows that: (a) with the exception of the cross-spectrum method, the SOS-based methods are very sensitive to the length of the channel-model used, and exhibit significant performance losses if this length is not correctly chosen; (b) the HOS method is quite robust with respect to channel-model lengths, but at the expense of a higher computational complexity.