A coherent initialisation method of adaptive MIMO equalizers

The convergence features of adaptive training algorithms are significantly degraded with increasing number of antennas in a MIMO system. Here a technique is presented to initialise the coefficients of a decision feedback equaliser (DFE) filter in order to provide a fast start-up for training. The proposed technique refers to the MMSE criteria of Wiener Filter theory for a basic of initialisation, however the essential intention still remains on an adaptive training algorithm, i.e. LMS or RLS, for the ISI cancellation. When the channel state information (CSI) presents at the receiver, which may well be obtained during synchronisation, the equaliser filter is initialised using the CSI, therefore the training of equaliser would be advanced at the beginning, so equaliser would have no important instability, and convergence speed problem. Although the proposed method has been evaluated and analysed for the SISO and MIMO equalisers, the technique can easily be extended to all other adaptive training applications, providing the CSI is present priori to the technique.