normrnd(0,sqrt(0.5),M,N)+i*normrnd(0,sqrt(0.5),M,N)

Sorry what do you mean by CSI? channel state information?

If you are concerning about he CSI by channel state information, you can adopt the conventional ZFBF, MMSE, MRC techniques.

for matlab, you can generate one random complex matrix to stand for the channel information as the answer listed before. then you should do the matrix transpose according to which detectors you want to adopt, ZFBF, MRC or MMSE. that is used a lot in MIMO study.

Maybe you can also refer to the works from the Ngo, Emil in Linkoping University.

Hope that will be of help to you.

Thank you!

CSI is for channel state information..

What will be the values of CSI for perfect channel & imperfect channel?

From my side, there is no criteria for the problem. you should adopt different methods while considering different environments. As the studies by prof. Thomas L. Marzetta from Bell lab and also the works by Prof. Emil and Erik G. Larsson in Linkoping, it is said that with perfect CSI, the sum rate can increase infinitely with antenna numbers growing. For imperfect CSI, I know less, but you can still refer to the work from linkoping and the team of Prof. Shi JIn and Prof. Xiqi Gao from Southeast Univesity, China.

Actually the channel estimation itself is a study area that attracts scholars on this, such as Prof. Cheng Xiang Wang and his team,

Nowadays, the blind channel estimation is proposed, also please refer to Ngo's work from Linkoping and also the works of Prof. Shi Jin and Prof. Xiqi Gao's team from southeast university, China.

Imperfect CSI is certainly very important when computing the achievable performance of MU-MIMO, because otherwise one can get overoptimistic results. The best way to obtain CSI is by sending pilot/training signals and estimate the channels. Many researchers, including myself, are now looking to "massive" MU-MIMO systems where there are many more antennas than users, because this gives robustness to imperfect CSI.

I have made a lot of Matlab code available online to enable reproduction of the simulation results in some of my research papers. There are two code packages that might be of interest to you since these consider MU-MIMO with imperfect CSI:

“Receive Combining vs. Multi-Stream Multiplexing in Downlink Systems with Multi-Antenna Users” (https://github.com/emilbjornson/one-or-multiple-streams)

and

"Massive MIMO with Non-Ideal Arbitrary Arrays: Hardware Scaling Laws and Circuit-Aware Design" (https://github.com/emilbjornson/hardware-scaling-laws)

I recommend you to have a look on the first one, which can be easily reused to test MU-MIMO with different number of antennas. The second one contains a lot of code for hardware impairments that are not related to what you want to do.

Thanks a lot @ Emil Sir for your valuable guidance...

Thank you Jagtar sir  and Emil.sir I was also Having a difficulty in realizing channel state for MU MIMO.

Thanks Prabagrane Sir....