Construction of DFE

Hey all,

I am constructing a DFE to serve as a baseline algorithm which I ca n
compare my future work against. 

I run my  DFE on Proakis Channel C(0.227 0.460 0.688 0.460 0.227) and
could not duplicate the BER vs Eb/No result thats in the Digital
Communication book.

It works fine (from the BER graph,again) if I run it through a truncated
version of channel C, [0.688 0.46 0.227]. So I am wondering if theres any
thing I need to take note of when dealing with channel with both precursor
and post cursor ISI.

Any form of help is appreciated.

Thanks!
Kelvin

In his book he has given results for both ZF and MMSE.

I myself did it and I get exactly the same results.

Check the paper by Al-Dahir and implement it.

Hey thanks. 

When dealing with channel like Proakis Channel B where there is a delay,
do we equalize the signal to the first tap? Or should we equalize it to the
dominant tap (which is the 3rd tap in Channel B situation)?

Thanks,
Kelvin

>In his book he has given results for both ZF and MMSE.
>
>I myself did it and I get exactly the same results.
>
>Check the paper by Al-Dahir and implement it.
>

>Hey thanks. 
>
>When dealing with channel like Proakis Channel B where there is a delay,
>do we equalize the signal to the first tap? Or should we equalize it to
the
>dominant tap (which is the 3rd tap in Channel B situation)?
>
>Thanks,
>Kelvin
>

In any channel of Proakis or multipath in general, there will be delay.

And in the channel B, the second tap with value 0.815 is dominant tap. 

There are two ways to equalize: 1. assume u know the channel at receiver
2. assume u do not know the channel at receiver.

I would suggest implement the 1st option for channel B using both ZF and
MMSE criteria.

Chintan

Hey Chintan,

Sorry I was talking about channel C. 

If my understanding of DFE is correct, the first two tap will cause
post-cursor ISI if we are to equalize the channel to the dominant (3rd)
tap, and DFE cant eliminate post-cursor ISI.

I am working on channel-known, MMSE case, not going into adaptive
equalizer...yet :)

Thanks alot.

Kelvin

>In any channel of Proakis or multipath in general, there will be delay.
>
>And in the channel B, the second tap with value 0.815 is dominant tap. 
>
>There are two ways to equalize: 1. assume u know the channel at receiver
>2. assume u do not know the channel at receiver.
>
>I would suggest implement the 1st option for channel B using both ZF and
>MMSE criteria.
>
>Chintan
>

>Hey Chintan,
>
>Sorry I was talking about channel C. 
>
>If my understanding of DFE is correct, the first two tap will cause
>post-cursor ISI if we are to equalize the channel to the dominant (3rd)
>tap, and DFE cant eliminate post-cursor ISI.

The first two taps are pre-cursor and the last two are post cursor. The FF
filter of DFE removes pre-cursor and FB filter removes post-cursor.

>
>I am working on channel-known, MMSE case, not going into adaptive
>equalizer...yet :)
>

http://www.freescale.com/files/dsp/doc/app_note/AN2072.pdf

I had used this document when I implemented DFE for channels B and C using
MMSE. Read it carefully and u should find the ans.

Hope this helps.

Chintan

Hey Chintan,

The page has been removed. Do you still have the softcopy with u? Or u
know of any other sources to access the file?

Thanks,
Kel

>http://www.freescale.com/files/dsp/doc/app_note/AN2072.pdf
>
>I had used this document when I implemented DFE for channels B and C
using
>MMSE. Read it carefully and u should find the ans.
>
>Hope this helps.
>
>Chintan
>
>

I do have the soft copy. Tell me your email add and I will forward that to
u.

Chintan

Thanks! Its piakster@gmail.com

Kelvin
>
>I do have the soft copy. Tell me your email add and I will forward that
to
>u.
>
>Chintan
>

>Thanks! Its piakster@gmail.com
>
>Kelvin
>>
>>I do have the soft copy. Tell me your email add and I will forward that
>to
>>u.
>>
>>Chintan
>>
>

Sent..