Splitting Receiver with Multiple Antennas

Yanyan Wang, Wanchun Liu and Xiangyun Zhou

arXiv: 2303.04321v1 · PDF · 8 March 2023
Categories: cs.IT, eess.SP, math.IT
Licence: CC BY 4.0 The site shows the text.

Abstract

Recently proposed splitting receivers, utilizing both coherently and non-coherently processed signals for detection, have demonstrated remarkable performance gain compared to conventional receivers in the single-antenna scenario. In this paper, we propose a multi-antenna splitting receiver, where the received signal at each antenna is split into an envelope detection (ED) branch and a coherent detection (CD) branch, and the processed signals from both branches of all antennas are then jointly utilized for recovering the transmitted information. We derive a closed-form approximation of the achievable mutual information (MI) in terms of the key receiver design parameters, including the power splitting ratio at each antenna and the signal combining coefficients from all the ED and CD branches. We further optimize these receiver design parameters and demonstrate important design insights for the proposed multi-antenna ED-CD splitting receiver: 1) the optimal splitting ratio is identical at each antenna, and 2) the optimal combining coefficients for the ED and CD branches are the same, and each coefficient is proportional to the corresponding antenna’s channel power gain. Our numerical results also demonstrate the MI performance improvement of the proposed receiver over conventional non-splitting receivers.

Items 9

Notation

80VI

\tilde{\cdot} and |\cdot| denote a complex number and the absolute-value norm of a complex number, respectively. \mathcal{H} (\cdot), \mathcal{H}(\cdot, \cdot), \mathcal{H}(\cdot|\cdot) represent the differential entropy, joint conditional differential entropy and conditional differential entropy, respectively. \mathcal{I} (\cdot; \cdot) denotes the MI. (\cdot)_r and (\cdot)_i denote the real part and imaginary part of a complex number, respectively. \mathbb{C} and \mathbb{R} are the complex number and real number, respectively. In addition, \mathcal{N}(m,\sigma^2) and \mathcal{CN}(m,\sigma^2) denote the real-valued and complex-valued Gaussian distribution with mean m and variance \sigma^2. \mathbb{E}(\cdot) and \textrm{Var}(\cdot) denote the expectation and variance of a random variable, respectively.

Proposition

Proposition 1

PXEP

In the high SNR regime, the achievable MI of the multi-antenna ED-CD splitting receiver with the power splitting ratios [\rho_1, \rho_2,\cdots,\rho_K] \in [0,1]^K\backslash \{\mathbf{0},\mathbf{1}\}, and the combining coefficients [\alpha_1,\alpha_2,\cdots,\alpha_K] and [\beta_1,\beta_2,\cdots,\beta_K] can be approximated as

\begin{aligned} \mathcal{I}({{{\tilde{X}}}};{{{\tilde{R}}}_1}, {{{R}}_2})\approx &\frac{1}{2}\log_2\left({A^2+1}\right)-\frac{1}{2}\log_2\left( {\sum_{k=1}^{K} B^2_k\sigma_{\textrm{A}}^2+C^2\sigma_{\textrm{cov}}^2}\right)-\\ &\frac{1}{2}\log_2\left({{\sum_{k=1}^{K} \bigg(\frac{B_k}{\sqrt{1+A^2}}+\frac{B'_kA}{\sqrt{1+A^2}} \bigg)^2}\sigma_{\textrm{A}}^2+C^2\sigma_{\textrm{cov}}^2} \right). \end{aligned}

where A\triangleq \gamma \sum_{k=1}^{K} \beta_k, B_k\triangleq\frac{\alpha_k}{A'\sqrt{P}|\tilde{h}_k|}, C\triangleq\frac{\sqrt{\sum_{k=1}^{K} \frac{\alpha^2_k}{\rho_k P |\tilde{h}_k|^2}}}{A'}, A^{'}\triangleq \sum_{k=1}^{K} \alpha_k, B^{'}_k\triangleq \frac{\gamma \beta_k}{\sqrt{P}{|\tilde{h}_k|}}, C^{'}\triangleq \gamma\sqrt{\sum_{k=1}^{K} \frac{\beta^2_k}{(1-\rho_k ) P |\tilde{h}_k|^2}}, and

\begin{aligned} \gamma\triangleq\frac{{\sqrt{\sum_{k=1}^{K} \frac{\alpha^2_k}{\rho_k P |\tilde{h}_k|^2}}}\sigma_{\textrm{cov}}}{\sqrt{2}\sqrt{\sum_{k=1}^{K} \frac{\beta^2_k}{(1-\rho_k ) P |\tilde{h}_k|^2}}A'\sigma_{\textrm{rec}}}. \end{aligned}

Proof: See Appendix A.

Proposition

Proposition 2

AAG3

In the high SNR regime, the followings hold:

the maximum achievable MI is
\begin{aligned} \mathcal{I}_{\textrm{max}}({{{\tilde{X}}}};{{{\tilde{R}}}_1}, {{{R}}_2})& \approx \log_2\bigg(\sum_{k=1}^{K}P|\tilde{h}_k|^2\bigg)\\ &\,\,\,\,-\frac{1}{2}\log_2\bigg(\bigg({{\frac{\sigma_{\textrm{cov}}^2}{\rho^* }}}+\sigma_{\textrm{A}}^2 \bigg) \bigg({\sigma_{\textrm{A}}^2} +\frac{{{\frac{\sigma_{\textrm{cov}}^2}{\rho^*}}}2 \frac{\sigma_{\textrm{rec}}^2}{(1-\rho^* ) }}{{{\frac{\sigma_{\textrm{cov}}^2}{\rho^*}}}+2 \frac{\sigma_{\textrm{rec}}^2}{(1-\rho^* ) }}\bigg)\bigg), \end{aligned}
where \rho^* is given by
\begin{aligned} \rho^*= \begin{cases} \Upsilon, & \sigma_{\textrm{{cov}}}^2> 4 \sigma_{\textrm{{rec}}}^2,\\ 1, & \textrm{else}, \end{cases} \end{aligned}
and \Upsilon is given by
\begin{aligned} \Upsilon=\frac{\sigma_{\textrm{cov}}^2 (\sigma_{\textrm{cov}}^2 - 2 \sigma_{\textrm{rec}}^2) ( \sigma_{\textrm{A}}^2 + 2\sigma_{\textrm{rec}}^2) - \sqrt{2 \Psi}}{\sigma_{\textrm{A}}^2 (\sigma_{\textrm{cov}}^2 - 4 \sigma_{\textrm{rec}}^2) (\sigma_{\textrm{cov}}^2 - 2 \sigma_{\textrm{rec}}^2)}, \end{aligned}
where \Psi=\sigma_{\textrm{cov}}^4 (\sigma_{\textrm{A}}^2 + \sigma_{\textrm{cov}}^2 - 2 \sigma_{\textrm{rec}}^2) (\sigma_{\textrm{cov}}^2 - 2 \sigma_{\textrm{rec}}^2) \sigma_{\textrm{rec}}^2 (\sigma_{\textrm{A}}^2 +2 \sigma_{\textrm{rec}}^2).
the optimal power splitting ratio at each antenna is identical and equals \rho^* in (?).
the optimal combining coefficients satisfy the following properties
\begin{split} & \alpha^*_k=c_{\alpha}|\tilde{h}_k|^2, \\ & \beta^*_k=c_{\beta}|\tilde{h}_k|^2, \end{split}
where k=1,2,\cdots K. c_{\alpha} and c_{\beta} are two arbitrary non-zero constants.

Proof: See Appendix B.

Used by Remark 2

Remark

Remark 1

4HF1

The optimal combining coefficients given in (?) are different from the widely-known equal gain combining (EGC) and MRC used in the conventional multi-antenna receivers[1]It is well-known that the MRC scheme is optimal for multi-antenna CD receivers. Based on the similar analytical steps, it can be readily proved that MRC is also optimal for multi-antenna ED receivers. Note that in this work, we refer to MRC as the well-known optimal combining scheme for the conventional multi-antenna receivers.. If one was to use EGC in the splitting receiver, we would have \alpha_1=\alpha_2=\cdots=\alpha_K and \beta_1=\beta_2=\cdots=\beta_K. If one was to use MRC in the splitting receiver, we would have \alpha_k = c_{\alpha} |\tilde{h}_k| and \beta_k = c_{\beta} |\tilde{h}_k|. In particular, the MRC scheme requires the combining coefficient to be linearly proportional to the magnitude of the corresponding channel. In contrast, the optimal combining scheme in (?) requires the combining coefficient to be linearly proportional to the power (i.e., magnitude squared) of the corresponding channel.

Remark

Remark 2

8VEX

From the proof of Proposition 2, one can readily derive the MI results when the EGC or MRC scheme is used for the splitting receiver. Note that for both EGC and MRC schemes, the combining coefficients satisfy

\begin{aligned} \frac{\alpha_1}{\beta_1}=\frac{\alpha_2}{\beta_2}=\cdots=\frac{\alpha_K}{\beta_K}. \end{aligned}

Following the proof of Proposition 2, one can show that the optimal splitting ratio at each antenna is identical and equals \rho^* in (?) under the condition of (2). Thus, we have the following MI results.

Uses Proposition 2

Corollary

Corollary 1

CVTY

When the EGC scheme is used for signal combining in the multi-antenna ED-CD splitting receiver, the achievable MI in the high SNR regime is given by

\begin{aligned} \mathcal{I}_{\textrm{EGC}}({{{\tilde{X}}}};{{{\tilde{R}}}_1}, {{{R}}_2})&\approx\log_2\bigg(\frac{PK^2}{\sum_{k=1}^{K}\frac{1}{|\tilde{h}_k|^2}}\bigg)\\ &-\frac{1}{2}\log_2\bigg(\bigg({{\frac{\sigma_{\textrm{cov}}^2}{\rho^* }}}+\sigma_{\textrm{A}}^2 \bigg) \bigg({\sigma_{\textrm{A}}^2} +\frac{{{\frac{\sigma_{\textrm{cov}}^2}{\rho^*}}}2 \frac{\sigma_{\textrm{rec}}^2}{(1-\rho^* ) }}{{{\frac{\sigma_{\textrm{cov}}^2}{\rho^*}}}+2 \frac{\sigma_{\textrm{rec}}^2}{(1-\rho^* ) }}\bigg)\bigg). \end{aligned}

Corollary

Corollary 2

9YYR

When the MRC scheme is used for signal combining in the multi-antenna ED-CD splitting receiver, the achievable MI in the high SNR regime is given by

\begin{aligned} \mathcal{I}_{\textrm{MRC}}({{{\tilde{X}}}};{{{\tilde{R}}}_1}, {{{R}}_2})&\approx\log_2\left(\frac{P\bigg(\sum_{k=1}^{K}|\tilde{h}_k|\bigg)^2}{K}\right)\\ &-\frac{1}{2}\log_2\bigg(\bigg({{\frac{\sigma_{\textrm{cov}}^2}{\rho^* }}}+\sigma_{\textrm{A}}^2 \bigg) \bigg({\sigma_{\textrm{A}}^2} +\frac{{{\frac{\sigma_{\textrm{cov}}^2}{\rho^*}}}2 \frac{\sigma_{\textrm{rec}}^2}{(1-\rho^* ) }}{{{\frac{\sigma_{\textrm{cov}}^2}{\rho^*}}}+2 \frac{\sigma_{\textrm{rec}}^2}{(1-\rho^* ) }}\bigg)\bigg). \end{aligned}

Corollary

Corollary 3

MGUU

In the high SNR regime, the achievable MI gain of the multi-antenna ED-CD splitting receiver is given by

\begin{aligned} G_{\textrm{MI}}= \begin{cases} (\ref{equ:MIgain2a}), & \sigma_{\textrm{{cov}}}^2> 4 \sigma_{\textrm{{rec}}}^2,\\ 0, &\textrm{else}. \end{cases} \end{aligned}

Remark

Remark 3

ZXO3

The multi-antenna ED-CD splitting receiver achieves higher MI than the conventional CD and ED receivers. As shown in Corollary 3, the obtained MI gain of the multi-antenna ED-CD splitting receiver approaches to a constant value (which can be shown to be larger than zero) in the high SNR regime when \sigma_{\textrm{{cov}}}^2> 4 \sigma_{\textrm{{rec}}}^2. Also, this asymptotic performance gain is independent of the number of antennas K in high SNR.

Source and licence

The text on this page is from arXiv:2303.04321v1 by Yanyan Wang, Wanchun Liu and Xiangyun Zhou, under the licence CC BY 4.0. The corpus made these changes: it read the items out of the LaTeX source, it expanded the macros of the paper, it put each proof after its statement, and it converted the text to HTML. The formulas are the TeX of the paper, and KaTeX renders them. A formula or a reference can look different from the PDF. The page is published under CC BY 4.0.

The record of each item is in analysis/objects/2303/2303.04321.jsonl.