Forums

Soft Decision for Viterbi Decoding..

Started by swong_88 October 5, 2008
Hello all,

This is my first post here.  :)  I am currently working on a 802.16d
receiver where I have to combine signals from 4 antennas and then pass on
soft decisions values to the Viterbi decoder for the final output.  I have
managed to figure out how to combine the 4 signals and figure out which
point was transmitted.  I used a method described by a user on this very
forum named James Sungjin Kim.  He recommended that you multiply each of
the received subcarriers by the channel estimate at that subcarrier.  The
method is described in the following thread:

http://www.dsprelated.com/showmessage/48861/1.php 

The only problem I am having now is how to determine how good this point
is.  I have read in papers which says that for a channel with only AWGN,
using the signal power itself is good enough.  But when we combine a
rayleigh fading channel or any kind of frequency selective fading, this
method falls apart.  Since fading only happens at specific frequencies, the
SNR is no longer uniform across all of the sub-carriers.  We don't know if
the power at that specific subcarrier is all noise power or actual signal
power.  Does anyone know of a method to properly estimate the SNR using the
pilot tones??   One paper I read recommended that the SNR can be estimated
as:

imag(ek)^2

Where ek is the equalized signal at the pilot tones.  This does not make
much sense in simulation since the signal at the pilot tone locations are
used as starting points for linear interpolation for the channel estimate. 
 In my simulation, equalizing the pilot tones with the channel estimate
will always result in zero for the imaginary part.   Anyone have any clues?
 I would greatly appreciate any help.  Thanks in advance..

Sam