A technique for improving the convergence probability of blind equalization
Meng-Lin Yu, P. Larsson · 2002
We present a method that can solve three common cases of misconvergence in blind two-dimensional, 2D, equalizers. The technique is based on measuring a few statistical 'goodness' parameters that describe properties of the constellation at the output of the equalizer. After misconvergence is detected, the tap weights in either the I or Q part of the equalizer are shifted an appropriate number of positions. The positional shift of tap weights moves the equalizer state close to the correct solution and entails a rapid and reliable convergence.