The SMI algorithm, which is also known as direct matrix inversion (DMI) algorithm, has recently been used for 3G systems and beyond, because the fast convergence property makes it suitable for use with high data rate transmissions [143,150]. However the complexity grows three-orders exponentially with the number of the weights (M 3). Recursive equations for the inverse of the correlation matrix thus had been used for the implementation on digital signal processors. | Initiatives in 4Gmobile Design 291 Figure Frame structures a W-CDMA and b EDGE. Least-mean square LMS algorithm Recursive least squares RLS algorithm Sample matrix inversion SMI algorithm The SMI algorithm which is also known as direct matrix inversion DMI algorithm has recently been used for 3G systems and beyond because the fast convergence property makes it suitable for use with high data rate transmissions 143 150 . However the complexity grows three-orders exponentially with the number of the weights M3 . Recursive equations for the inverse of the correlation matrix thus had been used for the implementation on digital signal processors. Space-time Equaliser Using Adaptive Antennas Space-time equalisers using the adaptive antennas with equalisers have also recently investigated a more powerful technique than only using the adaptive antennas 13 36-39 46 63 104 143 144 150 . They are also called smart antennas or intelligent antennas. Nonlinear adaptive equalisers such as decision-feedback equaliser DFE and maximumlikelihood sequence estimator MLSE had been investigated and implemented on commercial systems to compensate for the ISI. However DFE cancels undesired delayed paths by subtracting replica from received signal and thus cannot obtain path diversity gain. MLSE which is well-known optimum equaliser and can be implemented by Viterbi algo 292 Broadband Wireless Mobile 3G and Beyond rithm VA but instead can obtain path diversity gain by exploiting the delayed path information. MLSE thus can have higher efficiency than DFE under the multipath-rich environment. However the longer the span of the multipath the more complicated the hardware implementation of the VA with the exponential behaviour in complexity as a function of the span of the ISI. Adaptive antennas on the other hand can suppress the relatively longer-delayed paths without the hardware overhead though adaptive antennas cannot obtain path diversity gain in the same way as the DFE. The .
