Summary: A ClosedForm Performance UpperBound on the FiniteLength MMSEDFE with
Arbitrary Decision Delay
Naofal AlDhahir \Lambda
, Senior Member IEEE
EDICS Category CL1:5:3 (Equalization)
A simple analytical upper--bound on the decision--point SNR of the unbiased finite--length MMSE--DFE with arbitrary decision
delay is derived. The bound can be easily computed given the channel impulse response, input SNR, length of feedforward filter, and
desired decision delay. The bound is shown to be tighter than the matched--filter bound. A by--product is a computationally--efficient
method for setting the decision delay.
The minimum--mean--square--error decision feedback equalizer (MMSE--DFE) is receiving considerable attention as an
effective receiver structure for broadband packet--based transmission in a multipath wireless environment (see e.g. [6, 5]).
For such high--sampling--rate scenarios, significant reductions in DFE implementation complexity can be achieved by using a
short feedforward filter (FFF), a sufficiently long feedback filter (FBF) with QPSK modulation (since it can be implemented
multiplier--free in this case), and an optimized decision delay [6, 5, 2].
Evaluating the DFE performance for a variety of channel impulse response profiles, number of equalizer taps, and
decision delay settings using Monte Carlo simulations can be a very time--consuming task. Therefore, it is of great interest
to develop analytical upper--bounds on the DFE performance that can be easily evaluated for different scenarios. A widely--
used upper--bound is the matched--filter bound (MFB) for one--shot communication over a multipath noisy channel .