FPGA Central

Hi,
I am simulating the performance of turbo codes (log-Map decoding used) o
stationary rayleigh fading channel on MATLAB. When i simulate using AWG
channel i get the perfect curve i am looking for but for Rayleigh channe
(one fading coefficient) I get a very bad curve. At 10dB EbNo the BER i
3x10exp-2. I am generating the channel for a SISO system as follows:

rayl_channel =(randn(1,1)+i*randn(1,1))/sqrt(2);

My question is what am i doing wrong. since the turbo decoder needs th
variance of noise to decode I am inputting the noise variance exactly th
same way I do with AWGN noise channel. Do i need to do something else wit
the variance when rayleigh channel is involved?
Note: I am doing a MMSE detection at the receiver side.
Please provide suggestions. Thanks.

khan

On Tue, 26 Sep 2006 15:47:14 -0500, "Khan" <[email protected]>
wrote:

>Hi,
>I am simulating the performance of turbo codes (log-Map decoding used) on
>stationary rayleigh fading channel on MATLAB. When i simulate using AWGN
>channel i get the perfect curve i am looking for but for Rayleigh channel
>(one fading coefficient) I get a very bad curve. At 10dB EbNo the BER is
>3x10exp-2. I am generating the channel for a SISO system as follows:
>
>rayl_channel =(randn(1,1)+i*randn(1,1))/sqrt(2);
>
>My question is what am i doing wrong. since the turbo decoder needs the
>variance of noise to decode I am inputting the noise variance exactly the
>same way I do with AWGN noise channel. Do i need to do something else with
>the variance when rayleigh channel is involved?
>Note: I am doing a MMSE detection at the receiver side.
>Please provide suggestions. Thanks.
>
>khan

One would expect a peformance degradation in Rayleigh statistics
compared to AWGN. You didn't provide a performance number for the
AWGN case, so I can't tell how much performance you lost in Rayleigh
noise, but it may be working as expected.

Another thing to consider is that often the LLR metric generation is
done assuming Gaussian noise. A Rayleigh-specific LLR generator may
change the situation.


Eric Jacobsen
Minister of Algorithms, Intel Corp.
My opinions may not be Intel's opinions.
http://www.ericjacobsen.org