FPGA Central

Hi ,

I am trying to model a system ( in MATLAB ) where i have a OFDM system .s
i start with generating a random set of data , convert the data to
parallel form, then a QPSK modulation ,perform a IFFT and then insert
guard interval . now what i intend to do is pass it through a underwate
channel model and check for the error rate vs SNR . so in parallel i hav
a channel matrix as a result of a underwater channel model . Now , i did
convultion ( conv(x,H)) between the channel matrix and the resultant dat
from the OFDM transmitter , and from the result of the convolution ,
removed the guard interval , passed it through a FFT , demodulated th
data , transformed the signal back into a serial form and then calculate
the BER . the error rate rate drops and increases . therefore the BER v
SNR curve looks something like a flat line ,the error rate just drops an
then increases again .interestingly , if i avoid the underwater channel
then i get a normal BER vs SNR curve , where the BER drops for increasin
values of SNR .so i am not sure if i should be doing just a convolutio
between H and data generated or anything more . Any help or questions t
get me going would be appreciated .

Thanks ,
NonameNoface

On Nov 29, 6:03 pm, "NonameNoface" <manosam...@rediffmail.com> wrote:
> Hi ,
>
> I am trying to model a system ( in MATLAB ) where i have a OFDM system .so
> i start with generating a random set of data , convert the data to a
> parallel form, then a QPSK modulation ,perform a IFFT and then insert a
> guard interval . now what i intend to do is pass it through a underwater
> channel model and check for the error rate vs SNR . so in parallel i have
> a channel matrix as a result of a underwater channel model . Now , i did a
> convultion ( conv(x,H)) between the channel matrix and the resultant data
> from the OFDM transmitter , and from the result of the convolution , i
> removed the guard interval , passed it through a FFT , demodulated the
> data , transformed the signal back into a serial form and then calculated
> the BER . the error rate rate drops and increases . therefore the BER vs
> SNR curve looks something like a flat line ,the error rate just drops and
> then increases again .interestingly , if i avoid the underwater channel ,
> then i get a normal BER vs SNR curve , where the BER drops for increasing
> values of SNR .so i am not sure if i should be doing just a convolution
> between H and data generated or anything more . Any help or questions to
> get me going would be appreciated .
>
> Thanks ,
> NonameNoface


Quick summary of your issue (as I understand): When you are passing
the data through a multipath channel, you are not getting
decreasing BER for increasing SNR.
Some questions:
1. Is your channel delay more than the guard interval duration?
2. When multipath channel is introduced, how are you identifying the
symbol boundary?

Hope pondering over these may help you arrive at the answer.

Regards,
Krishna
www.dsplog.com