FFT in OFDM and 802.11a

Hi,

I have a doubt. May be, someone on this list can help me out here...

In IEEE 802.11a (OFDM system), packet data given at end of the
standard document (for verification, i assume) can be generated if
IFFT-FFT vectors are NOT normalized.
unnormalized IFFT vectors have energy = 1/Nfft (Used in transmitter).
unnormalized FFT vectors have energy = Nfft (used in receiver).
(Nfft->FFT length = 64)

At the OFDM compliant receiver, when AWGN samples (with symbols) are
processed by unnormalized FFT, noise variance will increase. And that
is bad for symbol detection (higher BER).

I am sure I am wrong somewhere or missing something in the standard.
Can anybody offer an explanation?

Thanks,
indie

OFDM is designed to battle with the ISI instead of AWGN noise. So I
think the assumptions are different.

freeindie2000@yahoo.com (indie) wrote in message news:<1511f9e0.0407250814.7b64412b@posting.google.com>...
> Hi,
> 
> I have a doubt. May be, someone on this list can help me out here...
> 
> In IEEE 802.11a (OFDM system), packet data given at end of the
> standard document (for verification, i assume) can be generated if
> IFFT-FFT vectors are NOT normalized.
> unnormalized IFFT vectors have energy = 1/Nfft (Used in transmitter).
> unnormalized FFT vectors have energy = Nfft (used in receiver).
> (Nfft->FFT length = 64)
> 
> At the OFDM compliant receiver, when AWGN samples (with symbols) are
> processed by unnormalized FFT, noise variance will increase. And that
> is bad for symbol detection (higher BER).
> 
> I am sure I am wrong somewhere or missing something in the standard.
> Can anybody offer an explanation?
> 
> Thanks,
> indie

do you mean to say that normalized fourier transform will make OFDM
less "ISI-resilient"?

steven_hyh@yahoo.com (Steve) wrote in message news:<3029ed06.0407271024.2aee9a74@posting.google.com>...
> OFDM is designed to battle with the ISI instead of AWGN noise. So I
> think the assumptions are different.
> 
> freeindie2000@yahoo.com (indie) wrote in message news:<1511f9e0.0407250814.7b64412b@posting.google.com>...
> > Hi,
> > 
> > I have a doubt. May be, someone on this list can help me out here...
> > 
> > In IEEE 802.11a (OFDM system), packet data given at end of the
> > standard document (for verification, i assume) can be generated if
> > IFFT-FFT vectors are NOT normalized.
> > unnormalized IFFT vectors have energy = 1/Nfft (Used in transmitter).
> > unnormalized FFT vectors have energy = Nfft (used in receiver).
> > (Nfft->FFT length = 64)
> > 
> > At the OFDM compliant receiver, when AWGN samples (with symbols) are
> > processed by unnormalized FFT, noise variance will increase. And that
> > is bad for symbol detection (higher BER).
> > 
> > I am sure I am wrong somewhere or missing something in the standard.
> > Can anybody offer an explanation?
> > 
> > Thanks,
> > indie

I have found the same problem when I did a simulation of an OFDM by using
Matlab. Single BPSK user experienced flat Rayleigh fading in my
simulation. The received signal at the receiver is x=fading
*Sending+AWGN.

After X=fft(x) and decision processing, the performance is very strange
and cannot match the theoretical BPSK BER in flat Rayleigh fading channel.
I checked the model. It seems the results of the model X=fft(fading
*sending)+AWGN looks like right. So I investigated the fft(AWGN) by using
the fft operater in Matlab. I found the mean and std are very high even
though I do not know weather the transformed data statify the Gaussian
distribution or not.

One week ago, I conducted a simulation using the model X=HS+AWGN in
frequency domian (after FFT)to estimate the OFDM channel etc, where N is
AWGN. The results match the BER performance given in an IEEE paper. Any
one can tell me how to use the IFFT and the FFT function in Matlab for
OFDM simulation?

Thanks in advance

Li Pingan


>do you mean to say that normalized fourier transform will make OFDM
>less "ISI-resilient"?
>
>steven_hyh@yahoo.com (Steve) wrote in message
news:<3029ed06.0407271024.2aee9a74@posting.google.com>...
>> OFDM is designed to battle with the ISI instead of AWGN noise. So I
>> think the assumptions are different.
>> 
>> freeindie2000@yahoo.com (indie) wrote in message
news:<1511f9e0.0407250814.7b64412b@posting.google.com>...
>> > Hi,
>> > 
>> > I have a doubt. May be, someone on this list can help me out here...
>> > 
>> > In IEEE 802.11a (OFDM system), packet data given at end of the
>> > standard document (for verification, i assume) can be generated if
>> > IFFT-FFT vectors are NOT normalized.
>> > unnormalized IFFT vectors have energy = 1/Nfft (Used in
transmitter).
>> > unnormalized FFT vectors have energy = Nfft (used in receiver).
>> > (Nfft->FFT length = 64)
>> > 
>> > At the OFDM compliant receiver, when AWGN samples (with symbols) are
>> > processed by unnormalized FFT, noise variance will increase. And
that
>> > is bad for symbol detection (higher BER).
>> > 
>> > I am sure I am wrong somewhere or missing something in the standard.
>> > Can anybody offer an explanation?
>> > 
>> > Thanks,
>> > indie
>




		
This message was sent using the Comp.DSP web interface on
www.DSPRelated.com