Hi,
I am doing MMSE channel estimation in OFDM system.
In the link below, someone has implemented the MMSE channel estimation matlab code, but it seems to have some problems.
https://www.mathworks.com/matlabcentral/fileexchange/41634-channel-estimation-for-ofdm-systems
Here, I wonder how to make the channel autocorrelation matrix.
the autocorrelation matrix in the time domain LMMSE estimation is simply g*g' and g.*conj(g), but my experimental experience shows that g*g'*nFFT and g.*conj(g)*nFFT performs better. Is this the right way??
Because MSE performance of the TD LMMSE is not monotonic decreasing.
Thank you.
Hello,
first, precise in which domain you want to perform (L)MMSE estimator: time or frequency.
The mathwork link you shared is a good link, in the sense that it presents LMMSE estimator in both time and frequency domains.
The only problem I notice is that the frequency response of the channel H is not normalized, indeed, which could explain the difference you obtain.
Best regards,
Vincent
- Badges
- Science topic