0
$\begingroup$

In a MIMO system, if the receiver has perfect CSI, i.e. knows that channel matrix $\boldsymbol{H}$, then the achievable rate is: $$ R = \log \left|\boldsymbol{I}+\frac{P}{N_RN_T}\boldsymbol{H}\boldsymbol{H}^H\right|\tag1$$ where $P$ is the transmit power and $N_R,N_T$ the number of receive and transmit antennas respectively. What happens if I don't know the channel, but estimate it? Will it simply be $$ R = \log \left|\boldsymbol{I}+\frac{P}{N_RN_T}\boldsymbol{H}\hat{\boldsymbol{H}}^H\right|\tag2$$ where $\hat{\boldsymbol{H}}$ is the channel estimate?

I'm interested specifically in LMMSE estimation (and Gaussian channel).

$\endgroup$
3
  • $\begingroup$ Can't be your equation $(2)$, because in that case, you'd just pick $\hat{\boldsymbol{H}}$ to maximize the $R$, not to estimate the channel as good as possible; for example $\hat{\boldsymbol H}= 10^{20}\cdot \mathbf I$ would probably be a pretty good pick. $\endgroup$ Nov 20 '19 at 13:33
  • $\begingroup$ @MarcusMüller I'll let OP answer but I believe the question was if a channel matrix estimate has been calculated, is it a good approach to just plug that estimate in to equation 1 to get an achievable rate estimate? Or, is there a better way to go about estimating achievable rate $\endgroup$
    – Engineer
    Nov 20 '19 at 13:46
  • $\begingroup$ @MarcusMüller Your estimate of the matrix is pretty good though :) $\endgroup$
    – Engineer
    Nov 20 '19 at 13:47
0
$\begingroup$

I found the following relation which answers my question

$$R=\log\det \left(\boldsymbol{C}_{MMSE}^{-1}\right)$$

where $\boldsymbol{C}_{MMSE}$ is the covariance matrix of the error in the data $\boldsymbol{e}=\boldsymbol{x}-\hat{\boldsymbol{x}}$

I found the answer in the paper "An Iteratively Weighted MMSE Approach to Distributed Sum-Utility Maximization for a MIMO Interfering Broadcast Channel" but it is said to be well known.

Therefore, we need to assume some method for estimation the data using the estimated channel, for example, match-filter or MMSE estimator, and then calculate its error (as a function of the real and estimated channels).

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.