0
$\begingroup$

enter image description hereI am reading a book called Massive MIMO Networks: Spectral, Energy, and Hardware Efficiency, and it implies that $$\frac{\mathbf{h}}{\sqrt{\text{E}\left\{\|\mathbf{h}\|^2\right\}}}$$ is the channel direction, where $\mathbf{h}\in \mathbb{C}^{M}$, where $M$ is a positive integer (see the picture for the snapshot). I don't understand what it means, and how. Could any explain this to me. Thanks in advance

EDIT: It I originally defined the direction as $$\frac{\|\mathbf{h}\|^2}{\sqrt{\text{E}\left\{\|\mathbf{h}\|^2\right\}}}$$ but it should be $$\frac{\mathbf{h}}{\sqrt{\text{E}\left\{\|\mathbf{h}\|^2\right\}}}$$

| improve this question | |
$\endgroup$
  • $\begingroup$ are you sure about the numerator? Could you cite the book and where to find the defintion? $\endgroup$ – AlexTP Jan 1 '19 at 16:26
  • $\begingroup$ @AlexTP I added a snapshot from the book to my original post. $\endgroup$ – BlackMath Jan 2 '19 at 4:31
  • $\begingroup$ Well, these are just definitions. I mean the author defined a term $\mathbf{h}/\sqrt{E |\mathbf{h}|^2}$ which can be interpreted as statistical direction (whose its norm can be different from 1, not like $\mathbf{h}/||\mathbf{h}||$). You can see that the denominator is a constant (by expected value) while the numerator is a random variable. $\endgroup$ – AlexTP Jan 2 '19 at 10:43
  • $\begingroup$ This definition leads to a nice property that when the dimension $M$ approaches infinity, some directions (in your example, $l \to i$ and $j \to k$) can be almost surely orthogonal whatever their channel realizations (impulses) $\mathbf{h}$. In brief, if a model (channel direction) is accepted then its analysis can follow. Just to see how the author use his defintions. $\endgroup$ – AlexTP Jan 2 '19 at 10:44
  • 1
    $\begingroup$ en.wikipedia.org/wiki/Euclidean_vector $\endgroup$ – AlexTP Jan 2 '19 at 11:53
1
$\begingroup$

You should double check the formula.

The classic single input multiple output (SIMO) equation with $N_T$ receive antennas is: $\mathbf{y}=\mathbf{h}x+\mathbf{w}$.

Where $x$ is the transmitted symbol (usually complex valued), $\mathbf{w} \sim CN(0,N_0\mathbf{I})$ is the complex noise, and $\mathbf{y}$ is the received vector, which is a vector of size $N_{R}\times1$.

You are asking about the channel vector $\mathbf{h}$ which contains the channel gains between the transmitter and each of the receive antennas, ie. $\mathbf{h}=[h_1,...,h_{N_R}]^T$. Since $\mathbf{h}$ is just a vector we can talk about its magnitude and direction. To look at a vector's direction we make the vector in question of unit length so we divide by the norm $||\mathbf{h}||$. That is, the direction of $\mathbf{h}$ is given as: $\frac{\mathbf{h}}{||\mathbf{h}||}$.

As far as what does the direction mean? Well, it can be thought of as what is responsible for the rotation of the transmitted symbol. The magnitude of $\mathbf{h}$ scales the transmitted symbol and the direction rotates the symbol in the complex plane.

| improve this answer | |
$\endgroup$
  • $\begingroup$ Thanks. I quoted the book I am reading in my original post. $\endgroup$ – BlackMath Jan 2 '19 at 4:32
  • $\begingroup$ Why the direction of $\mathbf{h}$ is given by $\frac{\mathbf{h}}{\|\mathbf{h}\|}$? It's close to what was mentioned in the book if you take the square root of the denominator in the direction in the original post, with the difference of expectation. $\endgroup$ – BlackMath Jan 2 '19 at 9:46
  • $\begingroup$ There is an expectation just because the channel coefficients are usually thought of as random variables. So taking expectation, norm squared, then square root is same as what I put down although I admit a bit sloppy on my part just because technically $\mathbf{h}$ is a random vector $\endgroup$ – Engineer Sep 30 '19 at 20:03

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.