| 1 | + | \documentclass[11pt]{article} |
| 2 | + | |
| 3 | + | \usepackage{fullpage} |
| 4 | + | \usepackage{amsmath, amssymb, bm, cite, epsfig, psfrag} |
| 5 | + | \usepackage{graphicx} |
| 6 | + | \usepackage{float} |
| 7 | + | \usepackage{amsthm} |
| 8 | + | \usepackage{amsfonts} |
| 9 | + | \usepackage{listings} |
| 10 | + | \usepackage{cite} |
| 11 | + | \usepackage{hyperref} |
| 12 | + | \usepackage{tikz} |
| 13 | + | \usepackage{enumitem} |
| 14 | + | \usetikzlibrary{shapes,arrows} |
| 15 | + | \usepackage{mdframed} |
| 16 | + | \usepackage{mcode} |
| 17 | + | \usepackage{siunitx} |
| 18 | + | %\usetikzlibrary{dsp,chains} |
| 19 | + | |
| 20 | + | %\restylefloat{figure} |
| 21 | + | %\theoremstyle{plain} \newtheorem{theorem}{Theorem} |
| 22 | + | %\theoremstyle{definition} \newtheorem{definition}{Definition} |
| 23 | + | |
| 24 | + | \def\del{\partial} |
| 25 | + | \def\ds{\displaystyle} |
| 26 | + | \def\ts{\textstyle} |
| 27 | + | \def\beq{\begin{equation}} |
| 28 | + | \def\eeq{\end{equation}} |
| 29 | + | \def\beqa{\begin{eqnarray}} |
| 30 | + | \def\eeqa{\end{eqnarray}} |
| 31 | + | \def\beqan{\begin{eqnarray*}} |
| 32 | + | \def\eeqan{\end{eqnarray*}} |
| 33 | + | \def\nn{\nonumber} |
| 34 | + | \def\binomial{\mathop{\mathrm{binomial}}} |
| 35 | + | \def\half{{\ts\frac{1}{2}}} |
| 36 | + | \def\Half{{\frac{1}{2}}} |
| 37 | + | \def\N{{\mathbb{N}}} |
| 38 | + | \def\Z{{\mathbb{Z}}} |
| 39 | + | \def\Q{{\mathbb{Q}}} |
| 40 | + | \def\R{{\mathbb{R}}} |
| 41 | + | \def\C{{\mathbb{C}}} |
| 42 | + | \def\argmin{\mathop{\mathrm{arg\,min}}} |
| 43 | + | \def\argmax{\mathop{\mathrm{arg\,max}}} |
| 44 | + | %\def\span{\mathop{\mathrm{span}}} |
| 45 | + | \def\diag{\mathop{\mathrm{diag}}} |
| 46 | + | \def\x{\times} |
| 47 | + | \def\limn{\lim_{n \rightarrow \infty}} |
| 48 | + | \def\liminfn{\liminf_{n \rightarrow \infty}} |
| 49 | + | \def\limsupn{\limsup_{n \rightarrow \infty}} |
| 50 | + | \def\MID{\,|\,} |
| 51 | + | \def\MIDD{\,;\,} |
| 52 | + | |
| 53 | + | \newtheorem{proposition}{Proposition} |
| 54 | + | \newtheorem{definition}{Definition} |
| 55 | + | \newtheorem{theorem}{Theorem} |
| 56 | + | \newtheorem{lemma}{Lemma} |
| 57 | + | \newtheorem{corollary}{Corollary} |
| 58 | + | \newtheorem{assumption}{Assumption} |
| 59 | + | \newtheorem{claim}{Claim} |
| 60 | + | \def\qed{\mbox{} \hfill $\Box$} |
| 61 | + | \setlength{\unitlength}{1mm} |
| 62 | + | |
| 63 | + | \def\bhat{\widehat{b}} |
| 64 | + | \def\ehat{\widehat{e}} |
| 65 | + | \def\phat{\widehat{p}} |
| 66 | + | \def\qhat{\widehat{q}} |
| 67 | + | \def\rhat{\widehat{r}} |
| 68 | + | \def\shat{\widehat{s}} |
| 69 | + | \def\uhat{\widehat{u}} |
| 70 | + | \def\ubar{\overline{u}} |
| 71 | + | \def\vhat{\widehat{v}} |
| 72 | + | \def\xhat{\widehat{x}} |
| 73 | + | \def\xbar{\overline{x}} |
| 74 | + | \def\zhat{\widehat{z}} |
| 75 | + | \def\zbar{\overline{z}} |
| 76 | + | \def\la{\leftarrow} |
| 77 | + | \def\ra{\rightarrow} |
| 78 | + | \def\MSE{\mbox{\small \sffamily MSE}} |
| 79 | + | \def\SNR{\mbox{\small \sffamily SNR}} |
| 80 | + | \def\SINR{\mbox{\small \sffamily SINR}} |
| 81 | + | \def\arr{\rightarrow} |
| 82 | + | \def\Exp{\mathbb{E}} |
| 83 | + | \def\var{\mbox{var}} |
| 84 | + | \def\Tr{\mbox{Tr}} |
| 85 | + | \def\tm1{t\! - \! 1} |
| 86 | + | \def\tp1{t\! + \! 1} |
| 87 | + | |
| 88 | + | \def\Xset{{\cal X}} |
| 89 | + | |
| 90 | + | \newcommand{\one}{\mathbf{1}} |
| 91 | + | \newcommand{\abf}{\mathbf{a}} |
| 92 | + | \newcommand{\bbf}{\mathbf{b}} |
| 93 | + | \newcommand{\dbf}{\mathbf{d}} |
| 94 | + | \newcommand{\ebf}{\mathbf{e}} |
| 95 | + | \newcommand{\gbf}{\mathbf{g}} |
| 96 | + | \newcommand{\hbf}{\mathbf{h}} |
| 97 | + | \newcommand{\pbf}{\mathbf{p}} |
| 98 | + | \newcommand{\pbfhat}{\widehat{\mathbf{p}}} |
| 99 | + | \newcommand{\qbf}{\mathbf{q}} |
| 100 | + | \newcommand{\qbfhat}{\widehat{\mathbf{q}}} |
| 101 | + | \newcommand{\rbf}{\mathbf{r}} |
| 102 | + | \newcommand{\rbfhat}{\widehat{\mathbf{r}}} |
| 103 | + | \newcommand{\sbf}{\mathbf{s}} |
| 104 | + | \newcommand{\sbfhat}{\widehat{\mathbf{s}}} |
| 105 | + | \newcommand{\ubf}{\mathbf{u}} |
| 106 | + | \newcommand{\ubfhat}{\widehat{\mathbf{u}}} |
| 107 | + | \newcommand{\utildebf}{\tilde{\mathbf{u}}} |
| 108 | + | \newcommand{\vbf}{\mathbf{v}} |
| 109 | + | \newcommand{\vbfhat}{\widehat{\mathbf{v}}} |
| 110 | + | \newcommand{\wbf}{\mathbf{w}} |
| 111 | + | \newcommand{\wbfhat}{\widehat{\mathbf{w}}} |
| 112 | + | \newcommand{\xbf}{\mathbf{x}} |
| 113 | + | \newcommand{\xbfhat}{\widehat{\mathbf{x}}} |
| 114 | + | \newcommand{\xbfbar}{\overline{\mathbf{x}}} |
| 115 | + | \newcommand{\ybf}{\mathbf{y}} |
| 116 | + | \newcommand{\zbf}{\mathbf{z}} |
| 117 | + | \newcommand{\zbfbar}{\overline{\mathbf{z}}} |
| 118 | + | \newcommand{\zbfhat}{\widehat{\mathbf{z}}} |
| 119 | + | \newcommand{\Ahat}{\widehat{A}} |
| 120 | + | \newcommand{\Abf}{\mathbf{A}} |
| 121 | + | \newcommand{\Bbf}{\mathbf{B}} |
| 122 | + | \newcommand{\Cbf}{\mathbf{C}} |
| 123 | + | \newcommand{\Bbfhat}{\widehat{\mathbf{B}}} |
| 124 | + | \newcommand{\Dbf}{\mathbf{D}} |
| 125 | + | \newcommand{\Ebf}{\mathbf{E}} |
| 126 | + | \newcommand{\Gbf}{\mathbf{G}} |
| 127 | + | \newcommand{\Hbf}{\mathbf{H}} |
| 128 | + | \newcommand{\Kbf}{\mathbf{K}} |
| 129 | + | \newcommand{\Pbf}{\mathbf{P}} |
| 130 | + | \newcommand{\Phat}{\widehat{P}} |
| 131 | + | \newcommand{\Qbf}{\mathbf{Q}} |
| 132 | + | \newcommand{\Rbf}{\mathbf{R}} |
| 133 | + | \newcommand{\Rhat}{\widehat{R}} |
| 134 | + | \newcommand{\Sbf}{\mathbf{S}} |
| 135 | + | \newcommand{\Ubf}{\mathbf{U}} |
| 136 | + | \newcommand{\Vbf}{\mathbf{V}} |
| 137 | + | \newcommand{\Wbf}{\mathbf{W}} |
| 138 | + | \newcommand{\Xhat}{\widehat{X}} |
| 139 | + | \newcommand{\Xbf}{\mathbf{X}} |
| 140 | + | \newcommand{\Ybf}{\mathbf{Y}} |
| 141 | + | \newcommand{\Zbf}{\mathbf{Z}} |
| 142 | + | \newcommand{\Zhat}{\widehat{Z}} |
| 143 | + | \newcommand{\Zbfhat}{\widehat{\mathbf{Z}}} |
| 144 | + | \def\alphabf{{\boldsymbol \alpha}} |
| 145 | + | \def\betabf{{\boldsymbol \beta}} |
| 146 | + | \def\mubf{{\boldsymbol \mu}} |
| 147 | + | \def\lambdabf{{\boldsymbol \lambda}} |
| 148 | + | \def\etabf{{\boldsymbol \eta}} |
| 149 | + | \def\xibf{{\boldsymbol \xi}} |
| 150 | + | \def\taubf{{\boldsymbol \tau}} |
| 151 | + | \def\sigmahat{{\widehat{\sigma}}} |
| 152 | + | \def\thetabf{{\bm{\theta}}} |
| 153 | + | \def\thetabfhat{{\widehat{\bm{\theta}}}} |
| 154 | + | \def\thetahat{{\widehat{\theta}}} |
| 155 | + | \def\mubar{\overline{\mu}} |
| 156 | + | \def\muavg{\mu} |
| 157 | + | \def\sigbf{\bm{\sigma}} |
| 158 | + | \def\etal{\emph{et al.}} |
| 159 | + | \def\Ggothic{\mathfrak{G}} |
| 160 | + | \def\Pset{{\mathcal P}} |
| 161 | + | \newcommand{\bigCond}[2]{\bigl({#1} \!\bigm\vert\! {#2} \bigr)} |
| 162 | + | \newcommand{\BigCond}[2]{\Bigl({#1} \!\Bigm\vert\! {#2} \Bigr)} |
| 163 | + | |
| 164 | + | \def\Rect{\mathop{Rect}} |
| 165 | + | \def\sinc{\mathop{sinc}} |
| 166 | + | \def\NF{\mathrm{NF}} |
| 167 | + | \def\Real{\mathrm{Re}} |
| 168 | + | \def\Imag{\mathrm{Im}} |
| 169 | + | \newcommand{\tran}{^{\text{\sf T}}} |
| 170 | + | \newcommand{\herm}{^{\text{\sf H}}} |
| 171 | + | |
| 172 | + | |
| 173 | + | % Solution environment |
| 174 | + | \definecolor{lightgray}{gray}{0.95} |
| 175 | + | \newmdenv[linecolor=white,backgroundcolor=lightgray,frametitle=Solution:]{solution} |
| 176 | + | |
| 177 | + | |
| 178 | + | |
| 179 | + | \begin{document} |
| 180 | + | |
| 181 | + | \title{Problems: Adaptive Modulation and Coding\\ |
| 182 | + | ECE-GY 6023. Wireless Communications} |
| 183 | + | \author{Prof.\ Sundeep Rangan} |
| 184 | + | \date{} |
| 185 | + | |
| 186 | + | \maketitle |
| 187 | + | |
| 188 | + | |
| 189 | + | \begin{enumerate} |
| 190 | + | \item \emph{MCS selection:} A system has four MCS selections with minimum required SNRs |
| 191 | + | as shown: |
| 192 | + | \begin{table}[h] |
| 193 | + | \centering |
| 194 | + | \begin{tabular}{|c|c|c|c|c|} |
| 195 | + | \hline |
| 196 | + | \textbf{MCS} & 1 & 2 & 3 & 4 \\ \hline |
| 197 | + | \textbf{Min SNR [dB]} & 0 & 4 & 8 & 12 \\ \hline |
| 198 | + | \end{tabular} |
| 199 | + | \end{table} |
| 200 | + | \begin{enumerate}[label=(\alph*)] |
| 201 | + | \item Suppose that the SNR in \si{dB}, $\gamma$, is unknown and can be modeled as Gaussian |
| 202 | + | variable with mean \SI{6}{dB} and standard deviation of \SI{2}{dB}. |
| 203 | + | What is the probability that $\gamma > 8$\, \si{dB}, the required SNR for MCS 3. |
| 204 | + | |
| 205 | + | \item The TX attempts MCS 3 and it fails, meaning $\gamma \leq 8$\, \si{dB}. |
| 206 | + | If it now attempts MCS 2, what is the probability that it will pass assuming the channel has not changed. |
| 207 | + | \end{enumerate} |
| 208 | + | |
| 209 | + | |
| 210 | + | \item \emph{Markovian errors:} |
| 211 | + | In this problem, we will show how to use Markov processes |
| 212 | + | to model error probabilities on correlated channels. |
| 213 | + | As a simple example, suppose that a channel in time slot $k$ |
| 214 | + | can be modeled as being in one of two states: a good state ($X_k=1$), |
| 215 | + | and a bad state ($X_k = 0$). |
| 216 | + | Assume $X_k$ is Markov with transition probability matrix |
| 217 | + | \[ |
| 218 | + | P = \begin{bmatrix} |
| 219 | + | 0.8 & 0.2 \\ |
| 220 | + | 0.3 & 0.7 |
| 221 | + | \end{bmatrix}. |
| 222 | + | \] |
| 223 | + | A transmitter sends packets in each time slot. |
| 224 | + | Let $Y_k=0$ or $1$ be the indicator that a packet fails or |
| 225 | + | passes in time slot $k$. Assume, |
| 226 | + | \[ |
| 227 | + | P(Y_k=1|X_k=1) = 0.8, \quad P(Y_k=1|X_k=0) = 0.4. |
| 228 | + | \] |
| 229 | + | Assume that, given the $X_k$'s, the $Y_k$'s are independent. |
| 230 | + | \begin{enumerate}[label=(\alph*)] |
| 231 | + | \item Let $\alpha_k(i) = P(X_k=i)$. Find the recursion |
| 232 | + | for the values $\alpha_{k+1}(i)$ in terms of |
| 233 | + | the values $\alpha_k(j)$. |
| 234 | + | |
| 235 | + | \item Let |
| 236 | + | \[ |
| 237 | + | \alpha^0_k(i) = P(X_k=i|Y_0=0,\ldots,Y_{k-1}=0). |
| 238 | + | \] |
| 239 | + | That is, $\alpha^0_k(i)$ is the probability that |
| 240 | + | $X_k=i$ given that the previous $k-1$ transmissions have failed. Find the recursion for $\alpha^0_{k+1}(i)$ in |
| 241 | + | terms of the values $\alpha^0_k(i)$. |
| 242 | + | |
| 243 | + | \item Let $T$ be the time, |
| 244 | + | \[ |
| 245 | + | T = \min \left\{~k~|Y_k = 1~\right\}, |
| 246 | + | \] |
| 247 | + | which is the index of the first slot that the packet passes. |
| 248 | + | Suppose that $X_0=0$. |
| 249 | + | Write a simple MATLAB program to compute $P(T=k)$ for |
| 250 | + | $k=0,1,\ldots,9$ using the above recursions. |
| 251 | + | \end{enumerate} |
| 252 | + | |
| 253 | + | |
| 254 | + | |
| 255 | + | \item \emph{Multi-Process ARQ timeline:} Suppose that a gNB wants to send $N=10$ packet data units |
| 256 | + | (PDUs). |
| 257 | + | The PDUs are indexed $n=0,1,\ldots,N-1$. |
| 258 | + | In each slot, it attempts to send one PDU beginning in slot $k=0$ starting with PDU 0. |
| 259 | + | There are $K=4$ parallel HARQ processes. |
| 260 | + | Suppose that the transmissions fails in slots $k=5, 6$ and 7 and passes in all other slots. |
| 261 | + | \begin{enumerate}[label=(\alph*)] |
| 262 | + | |
| 263 | + | \item For each PDU, indicate the fist slot it is correctly decoded at the receiver. |
| 264 | + | |
| 265 | + | \item Suppose the receiver only releases decoded the PDUs in order to the higher layer. |
| 266 | + | So, for example, it holds PDU 3 back until it receives PDU 2. Also, there is a fixed delay |
| 267 | + | of 3 slots from the time of transmission to the PDU being available at the receiver for higher layers. |
| 268 | + | When do the PDUs arrive at the higher layer? |
| 269 | + | \end{enumerate} |
| 270 | + | |
| 271 | + | |
| 272 | + | |
| 273 | + | \item \emph{TB size:} Suppose that a \SI{64}{kbps} voice over IP (VoIP) |
| 274 | + | system transmits frames once every |
| 275 | + | \SI{20}{ms}. Each voice frame also requires a \SI{20}{B} IP header, \SI{20}{B} UDP header, |
| 276 | + | and 24 bits CRC. |
| 277 | + | \begin{enumerate}[label=(\alph*)] |
| 278 | + | \item How many bits are in each voice frame? |
| 279 | + | |
| 280 | + | \item Suppose the data is transmitted in an NR-like system with 14 OFDM symbols and 12 sub-carriers |
| 281 | + | per RB. In each RB, 14 REs are used for overhead. At a spectral efficiency of 2 bits / RE, |
| 282 | + | how many RBs are needed to transmit the voice packet. |
| 283 | + | |
| 284 | + | \item If the system has 51 RBs in bandwidth with one slot every \SI{0.5}{ms}, what is the fraction |
| 285 | + | of RBs used by the VoIP application? |
| 286 | + | \end{enumerate} |
| 287 | + | |
| 288 | + | |
| 289 | + | \item \emph{HARQ Errors:} For each of the following events, state what will occur: |
| 290 | + | \begin{itemize} |
| 291 | + | \item The PDU can eventually recovered through HARQ, or |
| 292 | + | \item The PDU cannot be recovered through HARQ and will need to be recovered from |
| 293 | + | a high-layer ARQ protocol (e.g.\ at the RLC or TCP layer). |
| 294 | + | \end{itemize} |
| 295 | + | \begin{enumerate}[label=(\alph*)] |
| 296 | + | \item A DL PDCCH for an initial transmission is not seen by the UE, so it does not even |
| 297 | + | know that there is a DL data transmission. |
| 298 | + | |
| 299 | + | \item The UE decodes the DL data and sends an ACK to the gNB. But, the gNB mistakes the |
| 300 | + | ACK for a NACK. |
| 301 | + | |
| 302 | + | \item The UE fails to decode the DL data and sends a NACK to the gNB. But, the gNB mistakes the |
| 303 | + | NACK for an ACK. |
| 304 | + | |
| 305 | + | \end{enumerate} |
| 306 | + | |
| 307 | + | |
| 308 | + | \item \emph{Power and SNR estimation:} Suppose that |
| 309 | + | we have two groups of reference symbols: |
| 310 | + | \begin{itemize} |
| 311 | + | \item Zero-power RS that contain noise only, |
| 312 | + | \[ |
| 313 | + | r_k = w_k, \quad w_k \sim C{\mathcal N}(0,N). |
| 314 | + | \] |
| 315 | + | On these symbols, we compute a noise estimate, |
| 316 | + | \[ |
| 317 | + | \widehat{N} = \frac{1}{K} \sum_{k=1}^K |r_k|^2, |
| 318 | + | \] |
| 319 | + | where $K$ is the number of symbols over which we average. |
| 320 | + | |
| 321 | + | |
| 322 | + | \item Non zero-power RS with signal and noise, |
| 323 | + | \[ |
| 324 | + | r_k = h_kx_k + w_k, \quad |
| 325 | + | h_k \sim C{\mathcal N}(0,E_s), \quad |
| 326 | + | w_k \sim C{\mathcal N}(0,N), |
| 327 | + | \] |
| 328 | + | and $|x_k|=1$ is known to the receiver. |
| 329 | + | On these symbols, we compute a signal power estimate |
| 330 | + | \[ |
| 331 | + | \widehat{S} = \frac{1}{M} \sum_{k=1}^K |r_k|^2, |
| 332 | + | \] |
| 333 | + | where $M$ is the number of symbols over which we average. |
| 334 | + | \end{itemize} |
| 335 | + | |
| 336 | + | \begin{enumerate}[label=(\alph*)] |
| 337 | + | \item Show that $\widehat{N}$ and $\widehat{S}$ |
| 338 | + | can be written as a scaled chi-squared |
| 339 | + | distributions with a certain number of degrees of freedom. |
| 340 | + | You can look up this distribution in any source. |
| 341 | + | |
| 342 | + | \item Show that the ratio $\widehat{S}/\widehat{N}$ can be written as a scaled $F$-distribution |
| 343 | + | distribution with a certain number of degrees of freedom. |
| 344 | + | You can look up this distribution in any source. |
| 345 | + | |
| 346 | + | \item Suppose we use |
| 347 | + | \[ |
| 348 | + | \widehat{\gamma} = \max\left\{ 0, \frac{\widehat{S}}{\widehat{N}} - 1\right\} |
| 349 | + | \] |
| 350 | + | as the estimate of the true SNR $\gamma = E_s/N$. |
| 351 | + | Plot the probability that the SNR is accurate within \SI{0.5}{dB} as a function of $K$ with $K=M$. |
| 352 | + | You can use the MATLAB function \mcode{fcdf}. |
| 353 | + | Assume the true SNR is, $\gamma = $\, \SI{3}{dB}. |
| 354 | + | \end{enumerate} |
| 355 | + | |
| 356 | + | \item \emph{CSI estimation bias:} Suppose that in a group of $K$ symbols, |
| 357 | + | reference symbols $x_k$ are received as |
| 358 | + | \[ |
| 359 | + | r_k = hx_k + w_k, \quad w_k \sim {\mathcal CN}(0,N), |
| 360 | + | \quad |hx_k|^2 = E_s, |
| 361 | + | \] |
| 362 | + | where $h$ is an unknown channel, $N$ is the noise power, and $E_s$ is the received signal energy. |
| 363 | + | We channel and noise estimates via |
| 364 | + | \[ |
| 365 | + | \widehat{h} = \frac{ \sum_{k=1}^K x_k^*r_k } {\sum_{k=1}^K |x_k|^2 }, |
| 366 | + | \quad |
| 367 | + | \widehat{N} = \frac{\alpha }{K} \sum_{k=1}^K |r_k - \widehat{h}x_k|^2. |
| 368 | + | \] |
| 369 | + | \begin{enumerate}[label=(\alph*)] |
| 370 | + | \item Find the constant $\alpha$ such that the noise estimate is unbiased. That is, |
| 371 | + | \[ |
| 372 | + | \Exp\left[ \widehat{N} \right] = N. |
| 373 | + | \] |
| 374 | + | \item Suppose that you obtain an accurate estimate of the noise $\widehat{N}=N$ |
| 375 | + | (for example, by averaging over large numbers of groups). How would you get an unbiased |
| 376 | + | estimate of $E_s$? |
| 377 | + | \end{enumerate} |
| 378 | + | |
| 379 | + | |
| 380 | + | |
| 381 | + | |
| 382 | + | \item \emph{Rate matching:} Suppose you send 200 bits with a rate $1/2$ convolutional code |
| 383 | + | with constraint length $K=7$. |
| 384 | + | \begin{enumerate}[label=(\alph*)] |
| 385 | + | \item How many coded bits are output from the convolutional encoder? Remember to include |
| 386 | + | the tail bits. |
| 387 | + | |
| 388 | + | \item Suppose you want to send the data on 150 QPSK symbols. How many bits should be punctured or |
| 389 | + | repeated? |
| 390 | + | \end{enumerate} |
| 391 | + | |
| 392 | + | \item \emph{Comparing Chase and IR:} |
| 393 | + | Suppose that a TX can create mother codes |
| 394 | + | that, on an AWGN channel, require an SNR $\gamma$ |
| 395 | + | and provide a rate per symbol of |
| 396 | + | \[ |
| 397 | + | R(\gamma) |
| 398 | + | = \min\{ \rho_{\rm max}, \alpha \log_2(1+\gamma) \}, |
| 399 | + | \] |
| 400 | + | where $\rho_{\rm max}$ is the maximum spectral efficiency |
| 401 | + | and $\alpha$ is the fraction that the code achieves within |
| 402 | + | the Shannon rate. |
| 403 | + | Now suppose we use this code for HARQ with $K$ transmissions. |
| 404 | + | Suppose that all the symbols in transmission $k$ experience |
| 405 | + | some SNR $\gamma_k$, $k=1,\ldots,K$. |
| 406 | + | \begin{enumerate}[label=(\alph*)] |
| 407 | + | \item Suppose, we use Chase combining where we |
| 408 | + | create a packet from the mother code and retransmit it |
| 409 | + | in each of the $K$ transmissions. For each target $\gamma$, |
| 410 | + | find the condition on $\gamma_1,\ldots,\gamma_K$ that the |
| 411 | + | packet will pass. Also, find the rate, $R_{\rm chase}(\gamma)$ that will be achieved if |
| 412 | + | the packet passes after $K$ transmissions. |
| 413 | + | |
| 414 | + | \item Next, suppose we use IR where we create a longer packet |
| 415 | + | and transmit a fraction $1/K$ symbols in each transmission. |
| 416 | + | For each target $\gamma$, |
| 417 | + | find the condition on $\gamma_1,\ldots,\gamma_K$ that the |
| 418 | + | packet will pass. Also, find the rate, $R_{\rm IR}(\gamma)$ that will be achieved if |
| 419 | + | the packet passes after $K$ transmissions. |
| 420 | + | |
| 421 | + | \item Set $K=3$ and generate random i.i.d.\ $\gamma_k$ |
| 422 | + | that are exponentially distributed with an average of \SI{3}{dB} (i.e.\ independent Rayleigh fading). |
| 423 | + | Generate $n=1000$ instances of this channel |
| 424 | + | using MATLAB. |
| 425 | + | By varying the target SNR $\gamma$, |
| 426 | + | plot the probability that the packet the packet passes |
| 427 | + | after $K$ transmissions, |
| 428 | + | vs.\ the rates $R_{\rm chase}(\gamma)$ and |
| 429 | + | $R_{\rm IR}(\gamma)$, for both Chase and IR combining. |
| 430 | + | |
| 431 | + | \end{enumerate} |
| 432 | + | |
| 433 | + | |
| 434 | + | |
| 435 | + | \end{enumerate} |
| 436 | + | |
| 437 | + | |
| 438 | + | |
| 439 | + | \end{document} |
| 440 | + | |
| 441 | + | |