Equalization confusion on OFDM

[thavamaran]

Hi guys, I am working on optical ofdm, its not much of a different from normal ofdm. I have a problem at the receiver side where my constellation shifts in phase.

So I did a very simple equalization, where the data before IFFT which will be in so call frequency domain, assume that as

X

. Then the data at receiver after FFT which will be in so call frequency domain, assume that as

Y/QUOTE].

So I did the channel estimation by taking

Y/X

and I will get

H

which is the channel property. Then to get what I trasmitted at the receiver side, to compensate all the distortion, since H=Y/X, so to get back

X

at the receiver, I did

X=Y/H

and I got back my exact constellation at the receiver.

Im not sure whether this is correct or no, please someone advice me on this. Thank you very much.

 

[Tom Duke]

It is correct. Actually this is the LS channel estimator.

 

[thavamaran]

Hi Tom, I hope it is LS channel estimator, cause that is how it explains in theory, but when I include noise it should be gone right, I tried with noise, its still perfectly equalized, and there is some explanation on LS about autocorrelation and interpolation, may I get advice on that as well?

Thank you very much.

 

[Tom Duke]

Autocorellation will help you to get the LMMSE channel estimator, that uses the LS estimator. Interpolation is used when in some subcarriers you do not have pilots and you want their channel estimate. Then you use interpolation with the assistance of the LS estimated subcarriers.

 

[thavamaran]

So what I should do now. Cause the way I did the channel estimation is right. But after adding noise to the system, I realize the equalization or compensation I perform at the receiver based on channel estimation is wrong, cause it even give perfect output with noise where else LS gets totally corrupted when noise is there.

I have fixed my pilot in every subcarrier, since I dont have doppler shift cause fading in my system doesnt happen with respect to time, therefore I fix my pilot according to the frequency portion of my system which corresponds to every subcarrier.

I have fixed pilot at the beginning of each subcarrier and at the end as well. but the changes of this symbol with respect to channel is same for every subcarrier, the level of changes in pilot is same for every subcarrier and its same for the beginning pilot and ending pilot.

I really dont have a clue on what should be my next step, please someone advice me on this!! Thanks very much!

 

[cristinik]

So what I should do now. Cause the way I did the channel estimation is right. But after adding noise to the system, I realize the equalization or compensation I perform at the receiver based on channel estimation is wrong, cause it even give perfect output with noise where else LS gets totally corrupted when noise is there.

I have fixed my pilot in every subcarrier, since I dont have doppler shift cause fading in my system doesnt happen with respect to time, therefore I fix my pilot according to the frequency portion of my system which corresponds to every subcarrier.

I have fixed pilot at the beginning of each subcarrier and at the end as well. but the changes of this symbol with respect to channel is same for every subcarrier, the level of changes in pilot is same for every subcarrier and its same for the beginning pilot and ending pilot.

I really dont have a clue on what should be my next step, please someone advice me on this!! Thanks very much!

Hi, I'm not sure if you have already solved your problem with equalization. Just in case, maybe your channel is not frequency selective, or at least in the bandwidth you are using, that is way the variations remain constant for those frequencies.

I am also working with OFDM now and equalization. If by any chance you had some useful matlab code or simulink and you are willing to share it, just let me know!