Summary: Efficient Computation of the DelayOptimized
Naofal AlDhahir \Lambda , Member, IEEE, and John M. Cioffi, Senior Member, IEEE
We present new fast algorithms for computing the optimum settings of a finite--length minimum--
mean--square--error decision feedback equalizer (MMSE--DFE) from channel and noise estimates.
These algorithms are based on displacement structure theory [5, 8] and generalize the algorithms of
[2, 1] by including delay optimization. Both symbol--spaced and fractionally--spaced feedforward filters
The minimum--mean--square--error decision feedback equalizer (MMSE--DFE) is a popular receiver
structure on severe--ISI and noise channels . Recently, there has been increased interest in channel--
estimate--based DFEs; as opposed to the more traditional adaptive DFEs; since the former exhibit
undeniable performance advantages over the latter (see  and the references therein). This perfor
mance improvement arises mainly because of the avoidance of the large eigenvalue spread problem
that slows the convergence of LMS--based adaptive DFEs. Channel--estimate--based equalizers were
also demonstrated in  to have better tracking performance than least--squares methods. Never
theless, they have not yet found wide spread application in practice mainly due to the formidable
computational complexity associated with computing the optimum settings of the DFE filters directly
from channel and noise estimates.