Forums

MAP vs MLSE

Started by bitterbutter July 18, 2007
Hi guys,

I'm currently doing my literature review on both types of equalization.
Based on what I understand, both have comparable performance but use
different approach; MLSE: sequence detection while MAP symbol by symbol
detection. Since both relatively similar, how can we choose which type of
equalizer is suitable to combat ISI?

Responds pointing to publications and articles are very much appreciated,
since I am new in this study.

Cheers


On Jul 18, 7:33 am, "bitterbutter" <bitterbutter2...@yahoo.com> wrote:
> Hi guys, > > I'm currently doing my literature review on both types of equalization. > Based on what I understand, both have comparable performance but use > different approach; MLSE: sequence detection while MAP symbol by symbol > detection. Since both relatively similar, how can we choose which type of > equalizer is suitable to combat ISI? > > Responds pointing to publications and articles are very much appreciated, > since I am new in this study. > > Cheers
I have no idea what you mean by "MAP equalization", but MLSE equalization is the optimal receiver. So there's the short answer. So what is a "MAP equalizer"? What is the AP information given so that as you claim it's a symbol-by-symbol detector? As for references, you can always start with Proakis' text or Biglieri/Benedetto's text on Digital Communication. Julius
Thanks Julius for your reply

I do agree with you, Proakis gave a depth insight of equalization
techniques, will do my own search for the rest other authors. 

From what I understood from Proakis' paper entitled "Adaptive equalization
for TDMA digital mobile radio", published in IEEE Trans on Vehicular Tech,
May 1991, both MLSE and MAP are non-linear equalization techniques,
however he never discussed when and what is the best situation to apply
these equalization techniques
On Jul 20, 5:23 am, "bitterbutter" <bitterbutter2...@yahoo.com> wrote:
> Thanks Julius for your reply > > I do agree with you, Proakis gave a depth insight of equalization > techniques, will do my own search for the rest other authors. > > From what I understood from Proakis' paper entitled "Adaptive equalization > for TDMA digital mobile radio", published in IEEE Trans on Vehicular Tech, > May 1991, both MLSE and MAP are non-linear equalization techniques, > however he never discussed when and what is the best situation to apply > these equalization techniques
I don't know that paper, and you don't describe what "MAP equalization" is at all. But I am going to guess that this whole "MAP equalization" method is a feedback-type method, where you iteratively cancel the effect of the previous symbols on the new symbol as they come. The MAP probably refers to the use of soft-decision instead of hard-decision in the feedback loop. So if this is the case, it probably works well for high SNR, and poorly at low SNR because of the feedback. The problem of using hard- decision in feedback equalizers is that once you get one incorrect symbol, it screws up a bunch of the next ones. With MAP maybe you don't, but it assumes that you know the statistics exactly. Else your MAP computations will be biased. So at best, at low SNR the MAP decision-feedback equalizer will be not much better than using no equalization, since the belief in the decoded symbols will be very low. The ideas are all the same: at high SNR, you can simplify things quite a bit, but these schemes fall apart spectacularly when the SNR turns out to be lower than what you expect. And because they are feedback decision schemes, it is hard to get the exact bound on where they fail. There are tons of paper in this area in the 80s, from Dave Forney (my hero), John Cioffi, John Proakis, Dave Messerschmitt, and many others. I can't claim to be an expert at all in this area, so hopefully you'll either learn it yourself or somebody else here can point you in a better direction. Julius