Hi 

I'm trying to compare performance with and without minimum phase 
prefilter 
for 8PSK equaliser with a DDFSE. When we employ a prefilter, we use 
min 
phase all pass prefilter for 2nd half and max phase all pass 
prefilter for 
1st half. Hmin_ap , Hmax_ap would be these all-pass filters for 
filtering 
data, and Hmin and Hmax would be channel estimates to give to DFSE, 
corresponding to this data. Hmax would be time reversed complex 
conjugate 
of Hmin. DFSE works in reverse direction for Hmax (1st half burst) 
My question is towards modelling for "no prefilter scenario". When we 
dont 
want any prefilter but want to use same DFSE (working in fwd, rev 
directions as above), minimum phase works fine with slight 
degradation, as 
expected. Estimated channel is as such treated as Hmin, and no 
filtering is 
applied on min data. For maximum phase part, no filtering of data is 
applied, and Hmax is identified as conj(Hmin(n:-1:1)). But this 
maximum 
phase part gives significant errors compared to minimum phase, which 
is 
totally unexpected (even at high SNR). 
Please clarify whether I'm missing any scaling or other operations 
that 
may be required specific to "maximum phase prefiltering with time 
reversed 
processing". 
Thanks in advance 
Sundar Rajan