The MCP2210 SPI module provides the MOSI, MISO and SCK signals to the outside world. The module has the ability to control the GP pins (as Chip Select) only if these pins are configured for Chip Select operation. 1.3.1 SPI MODULE FEATURES The SPI module has the following configurable features: seta rt.Bi.Delays. Java SE 1.3 Downloads. Go to the Oracle Java Archive page. Thank you for downloading this release of the Java TM Platform, Standard Edition Development Kit (JDK TM).The JDK is a development environment for building applications, applets, and components using the Java programming language.

Experience Neo4j 4.3 on your desktop. Get started with the free graph database download today and avoid the costs of self-hosted deployment. 1.3.1.1 ESP32: NodeMCU “Joy-It” This BSB-LAN adapter board is designed for the 30 pin ESP32 NodeMCU board from Joy-It (WROOM32 chip). A user manual is available for the board from the manufacturer. There are both the board-specific pinout scheme and a general guide to using ESP32 boards with the Arduino IDE! 2009 Page 1 of 18 1 OVERVIEW 3 1.1 3 1.2 2 LPC1768 Features RDB1768V2, 3.1 Schematics 7 3.2 Mapping of peripherals to LPC1768 IO pins 14 4 IO PADS FOR, development platform for the development of systems around the NXP LPC1768 Cortex-M3 based Microcontroller.

## MAC layer design for mmWave massive MIMO

G. Lee, Y. Sung, inmmWave Massive MIMO, 2017

### 10.4User Scheduling in Massive MIMO

As seen in the previous section, the sum rate of RBF and SUS scales with the same rate as the capacity with scaling $O\left(MloglogK\right)$ in MU-MISO downlink networks for fixed M under the i.i.d. Rayleigh fading assumption, as K increases [18, 19]. The considered situation with fixed M and large K models MU-MISO networks with small-scale BS antenna arrays of four or eight transmit antennas at the BS and several hundred or thousand active users in a cell, as in a 4G network. In massive MIMO, the number M of transmit antennas at the BS is very large, in the order of hundreds. To analyze the performance of user scheduling in massive MIMO, it is necessary to consider the scenario of large M and the performance behavior as M increases. One of the important findings in this regard is that the scheduling gain obtained from smart user selection is insignificant under rich scattering environments, commonly modeled as i.i.d. Rayleigh fading channel elements (10.8), when M is very large [21–23]. This is mainly due to the “channel hardening effect” [44] and the difficulty of finding a subset of semiorthogonal users in the massive MIMO case [21, 30]:

Channel hardening effect: This refers to the phenomenon that in massive MIMO the norm of each channel vector from the same Rayleigh fading distribution converges to the same constant by the law of large numbers. Assume the channel vector of user k${\mathbf{h}}_{k}\stackrel{i.i.d.}{\sim }\mathcal{C}\mathcal{N}\left(0,\mathbf{I}\right)$. Then, ∥hk2 is a chi-square random variable with 2M degrees-of-freedom and one can show [21]:

for x > 1, using large deviations principle. Then:
(10.13)$\begin{array}{ll}\hfill \mathbb{P}\left(\underset{k}{max}{\parallel }{\mathbf{h}}_{k}{{\parallel }}^{2}\le x\right)& =\mathbb{P}{\left({\parallel }{\mathbf{h}}_{k}{{\parallel }}^{2}\le x\right)}^{K}\hfill \end{array}$
Since $\eta \triangleq x-1-logx>0$ for x > 1, the right-hand side (RHS) of (10.14) converges to:
(10.15)$\begin{array}{l}\hfill {\left(1-{e}^{-\eta M}\right)}^{K}\to 1,\end{array}$
if ${lim}_{M,K\to \mathfrak{\infty }}\frac{logK}{M}=0$. This implies that $\mathbb{P}\left({max}_{k}{\parallel }{\mathbf{h}}_{k}{{\parallel }}^{2}\le x\right)↑1$ for all x > 1, when $logK$ grows slower than M such as K = μM, K = M3, and $K={e}^{{M}^{0.9}}$. It is obvious that $\mathbb{P}\left({max}_{k}{\parallel }{\mathbf{h}}_{k}{{\parallel }}^{2}\le x\right)↓0$ for all x < 1 due to $\mathbb{E}\left[{\parallel }{\mathbf{h}}_{k}{{\parallel }}^{2}\right]=1$ by the law of large numbers. Consequently, ${max}_{k}{\parallel }{\mathbf{h}}_{k}{{\parallel }}^{2}$ converges to the constant 1 in probability. This observation shows that there is no user that has a significantly large channel vector magnitude when ${lim}_{M,K\to \mathfrak{\infty }}\frac{logK}{M}=0$, i.e., $K=o\left({e}^{{c}^{\mathfrak{\prime }}M}\right)$ with some constant

c′ > 0.

The negligible probability of finding a subset of semiorthogonal users: Let us consider the probability of finding a set of M roughly orthogonal users from K users with ${\mathbf{h}}_{k}\stackrel{i.i.d.}{\sim }\mathcal{C}\mathcal{N}\left(0,\mathbf{I}\right)$, k = 1,2,…,K. This probability can be computed by considering a set of M double cones constructed around each axis of the M-dimensional space:

where ei is the ith column of the M × M identity matrix. Note that a set of M channel vectors each of which is contained in one distinct cone of the cone set ${\left\{}{\mathcal{C}}_{1}^{\mathfrak{\prime }},{\mathcal{C}}_{2}^{\mathfrak{\prime }},\dots ,{\mathcal{C}}_{M}^{\mathfrak{\prime }}{\right\}}$ is a set of M roughly orthogonal channel vectors. The probability that the channel vector hk

falls in the cone ${\mathcal{C}}_{i}^{\mathfrak{\prime }}$ is given by [30]:
(10.17)$\begin{array}{ll}\hfill \mathbb{P}{\left\{}{\mathbf{h}}_{k}\mathfrak{\in }{\mathcal{C}}_{i}^{\mathfrak{\prime }}{\right\}}& =\mathbb{P}{\left\{}{}{h}_{k,i}{}\ge {\xi }^{\mathfrak{\prime }}{\parallel }{\mathbf{h}}_{k}{\parallel }{\right\}}\hfill \\ \approx {e}^{-{\left({\xi }^{\mathfrak{\prime }}\right)}^{2}M},\hfill \end{array}$
where the second step results from the fact that ${\parallel }{\mathbf{h}}_{k}{\parallel }\to M$ for large M by the law of large numbers. Therefore, the probability that the cone ${\mathcal{C}}_{i}^{\mathfrak{\prime }}$ contains at least one out of the K

channel vectors is given by:
(10.18)$\begin{array}{l}\mathbb{P}\left\{{\mathcal{C}}_{i}^{\prime }\mathrm{contains}\mathrm{at}\mathrm{least}\mathrm{one}\mathrm{user}\right\}\\ =1-\mathbb{P}{\left\{{\mathbf{h}}_{k}\notin {\mathcal{C}}_{i}^{\prime }\right\}}^{k}\\ \approx 1-{\left(1-{e}^{-{\left({\xi }^{\prime }\right)}^{2}M}\right)}^{k}\to 0\end{array}$
if ${lim}_{M,K\to \mathfrak{\infty }}\frac{logK}{M}=0$. This observation also shows that there are no subset of roughly orthogonal M users for very large M if ${lim}_{M,K\to \mathfrak{\infty }}\frac{logK}{M}=0$, i.e., $K=o\left({e}^{{c}^{\mathfrak{\prime }}M}\right)$ with some constant c′ > 0.

The above two facts show that if K grows slower than the exponential rate with respect to M, there is neither significant MU diversity gain nor a user subset of size M with roughly orthogonal channel vectors. Hence, as the number M of transmit antennas at the BS increases, K should increase exponentially with respect to M to obtain a nontrivial scheduling gain. However, in massive MIMO with M in the order of hundreds, the required number K of active users in the cell is too large, thus the scheduling gain is trivial due to the insufficient number of active users in practical networks under rich scattering environments.

The channel hardening effect leads to simplified resource allocation because fast-fading channel variation in the frequency domain also disappears due to the channel hardening effect. Thus, scheduling over the frequency domain may not be required and the whole spectrum can be assigned simultaneously to all users scheduled in the antenna domain.

Consistent with the above observations, the sum rate of RBF is shown to scale as [18]:

for $M=O\left(logK\right)$ where xy indicates ${lim}_{M,K\to \mathfrak{\infty }}x/y=1$ and c is a positive constant. Furthermore, when ${lim}_{M,K\to \mathfrak{\infty }}\frac{logK}{M}=0$, [18]:

(10.20)$\begin{array}{l}\hfill \underset{M,K\to \mathfrak{\infty }}{lim}\frac{{\mathcal{R}}_{\mathrm{RBF}}}{M}=0.\end{array}$