Upload
jane-washington
View
226
Download
0
Embed Size (px)
Citation preview
BROADBAND BEAMFORMING
Presented by:Kalpana Seshadrinathan
INTRODUCTIONEqualization Distortion caused by the channel results in
Inter-Symbol Interference (ISI) which has to be compensated for to reduce the probability of error
Such a compensator is called an equalizerProblem in Broadband Communications High Data Rates are transmitted and
multipath distortion is more severe Low complexity equalizers required without
sacrificing too much in ISI mitigation
TYPES OF EQUALIZERS
Maximum Likelihood Sequence Estimation (MLSE) Optimal in terms of probability of error Disadvantage: Exponential increase in computation
requirements with increase in channel memory
Linear equalizers Sub-optimal approach Used when computational complexity of MLSE is
prohibitive Commonly used optimality criteria for choosing
filter coefficients include mean-squared error and peak distortion
BROADBAND BEAMFORMERS
MLSE is not implementable in the broadband case due to computational complexityA broadband beamformer reduces the ISI to narrowband levels in the receiverThe space-time receiver is made up of an antenna array followed by an FIR filter bankOptimal MAP equalization is then performed on the beamformer output
Koca and Levy, 2002
POWER COMPLEMENTARITY
Noise at the filter output is colored and cannot be applied to a trellis-based equalizer
Filter coefficients are hence chosen to satisfy the power complementarity property to ensure white noise at receiver output [Koca and Levy, 2002]
Power Complementarity property of an N-channel filterbank is defined by [Vaidyanathan, 1993]
1 ) ( ) (~
1
z Wz Wi
N
ii
EFFECT OF OVERSAMPLING
Oversampling in the receive filter colors the noise
BEAMFORMER DESIGN
The channel is modeled by an L-ray complex baseband impulse response with Additive White Gaussian Noise
Filter coefficients are designed using Lagrangian optimizationObjective function is the mean-squared error at the beamformer outputThe power complementarity constraint is imposed with suitable modifications to account for oversampling
)()(1
l
L
ll tatg
RESULTS
CONCLUSIONS AND FUTURE WORK
The impulse response of the channel is seen to be shortened effectively by the beamformer ISI is reduced to about 3 baud intervals and MAP equalization can be performed on the beamformer outputFuture work may incorporate co-channel interference suppression in the beamformer design