Adaptive Mimo Equalizer Algorithms for HSDPA

In recent years, attention has focused on increasing the capacity of wireless data transmission by using multiple transmit and receive antennas, more generally known as multiple-input, multiple-output (MIMO) systems. An extension to the Third Generation Partnership Program (3GPP) standard, called High Speed Downlink Packet Access (HSDPA), combines MIMO techniques with composite code-division multiple access (CDMA) substreams to achieve high throughput. The data rates and capacity demands of HSDPA require MIMO equalizers, which attempt to undo channel coloration and cross-coupling using the pilot signals defined in the 3GPP standard. An LMS-like algorithm (LMScorr) has been proposed which uses an M-chip output correlator to reduce data traffic noise and thereby enhance the pilot signals. Although the LMScorr algorithm is very robust and relatively simple to implement, its convergence is slow, thereby degrading dynamic performance. In this memo, we seek alternative algorithms that provide faster convergence while still being implementable within the afforded computational constraints. We investigate several variations of the correlation filter, a two-step block estimation scheme, and some RLS versions. Of these alternative algorithms, only the RLS with M-chip output correlator (RLScorr) is superior to LMScorr, achieving a dramatic improvement in convergence speed. The added complexity of the RLScorr algorithm is substantially reduced by using conditional updates.