I/Q uncorrelation method using noise predictor for reduced complexity of MLSD

Abstract

An I/Q uncorrelation method is proposed for efficient equalization of quadrature signals. The I/Q uncorrelator for the Forney receiver which transforms complex valued filters into real valued ones makes I and Q channels uncorrelated. Unfortunately, however, it makes the noise samples colored which brings 1.8 dB loss in 10(-3) BER. By introducing a new distance metric using a noise predictor, we succeeded in recovering a 1 dB gain, thus, reducing the loss to 0.8 dB.