The following notation is used throughout this Appendix:
Quantity or Operation | Notation |
---|---|
Size of modulation constellation | $$M$$ |
Number of bits per symbol | $$k={\mathrm{log}}_{2}M$$ |
Energy per bit-to-noise power-spectral-density ratio | $$\frac{{E}_{b}}{{N}_{0}}$$ |
Energy per symbol-to-noise power-spectral-density ratio | $$\frac{{E}_{s}}{{N}_{0}}=k\frac{{E}_{b}}{{N}_{0}}$$ |
Bit error rate (BER) | $${P}_{b}$$ |
Symbol error rate (SER) | $${P}_{s}$$ |
Real part | $$\mathrm{Re}[\cdot ]$$ |
Largest integer smaller than | $$\lfloor \cdot \rfloor $$ |
The following mathematical functions are used:
Function | Mathematical Expression |
---|---|
Q function | $$Q(x)=\frac{1}{\sqrt{2\pi}}{\displaystyle \underset{x}{\overset{\infty}{\int}}\mathrm{exp}(-{t}^{2}/2)dt}$$ |
Marcum Q function | $$Q(a,b)={\displaystyle \underset{b}{\overset{\infty}{\int}}t\mathrm{exp}\left(-\frac{{t}^{2}+{a}^{2}}{2}\right){I}_{0}(at)dt}$$ |
Modified Bessel function of the first kind of order $$\nu $$ | $${I}_{\nu}(z)={\displaystyle \sum _{k=0}^{\infty}\frac{{\left(z/2\right)}^{\upsilon +2k}}{k!\Gamma (\nu +k+1)}}$$ $$\Gamma (x)={\displaystyle \underset{0}{\overset{\infty}{\int}}{e}^{-t}{t}^{x-1}dt}$$ |
Confluent hypergeometric function | $${}_{1}F{}_{1}(a,c;x)={\displaystyle \sum _{k=0}^{\infty}\frac{{(a)}_{k}}{{(c)}_{k}}\frac{{x}^{k}}{k!}}$$ |
The following acronyms are used:
Acronym | Definition |
---|---|
M-PSK | M-ary phase-shift keying |
DE-M-PSK | Differentially encoded M-ary phase-shift keying |
BPSK | Binary phase-shift keying |
DE-BPSK | Differentially encoded binary phase-shift keying |
QPSK | Quaternary phase-shift keying |
DE-QPSK | Differentially encoded quaternary phase-shift keying |
OQPSK | Offset quaternary phase-shift keying |
DE-OQPSK | Differentially encoded offset quaternary phase-shift keying |
M-DPSK | M-ary differential phase-shift keying |
M-PAM | M-ary pulse amplitude modulation |
M-QAM | M-ary quadrature amplitude modulation |
M-FSK | M-ary frequency-shift keying |
MSK | Minimum shift keying |
M-CPFSK | M-ary continuous-phase frequency-shift keying |
M-PSK. From equation 8.22 in [2]
$${P}_{s}=\frac{1}{\pi}{\displaystyle \underset{0}{\overset{(M-1)\pi /M}{\int}}\mathrm{exp}\left(-\frac{k{E}_{b}}{{N}_{0}}\frac{{\mathrm{sin}}^{2}\left[\pi /M\right]}{{\mathrm{sin}}^{2}\theta}\right)d\theta}$$
The following expression is very close, but not strictly equal, to the exact BER (from [4] and equation 8.29 from [2]):
$${P}_{b}=\frac{1}{k}\left({\displaystyle \sum _{i=1}^{M/2}({w}_{i}^{\text{'}}){P}_{i}}\right)$$
where $${w}_{i}^{\text{'}}={w}_{i}+{w}_{M-i}$$, $${w}_{M/2}^{\text{'}}={w}_{M/2}$$, $${w}_{i}$$is the Hamming weight of bits assigned to symbol i, and
$$\begin{array}{c}{P}_{i}=\frac{1}{2\pi}{\displaystyle \underset{0}{\overset{\pi (1-(2i-1)/M)}{\int}}\mathrm{exp}\left(-\frac{k{E}_{b}}{{N}_{0}}\frac{{\mathrm{sin}}^{2}\left[(2i-1)\pi /M\right]}{{\mathrm{sin}}^{2}\theta}\right)d\theta}\\ -\frac{1}{2\pi}{\displaystyle \underset{0}{\overset{\pi (1-(2i+1)/M)}{\int}}\mathrm{exp}\left(-\frac{k{E}_{b}}{{N}_{0}}\frac{{\mathrm{sin}}^{2}\left[(2i+1)\pi /M\right]}{{\mathrm{sin}}^{2}\theta}\right)d\theta}\end{array}$$
Special case of $$M=2$$, e.g., BPSK (equation 5.2-57 from [1]):
$${P}_{s}={P}_{b}=Q\left(\sqrt{\frac{2{E}_{b}}{{N}_{0}}}\right)$$
Special case of $$M=4$$, e.g., QPSK (equations 5.2-59 and 5.2-62 from [1]):
$$\begin{array}{c}{P}_{s}=2Q\left(\sqrt{\frac{2{E}_{b}}{{N}_{0}}}\right)\left[1-\frac{1}{2}Q\left(\sqrt{\frac{2{E}_{b}}{{N}_{0}}}\right)\right]\\ {P}_{b}=Q\left(\sqrt{\frac{2{E}_{b}}{{N}_{0}}}\right)\end{array}$$
DE-M-PSK. $$M=2$$, e.g., DE-BPSK (equation 8.36 from [2]):
$${P}_{s}={P}_{b}=2Q\left(\sqrt{\frac{2{E}_{b}}{{N}_{0}}}\right)-2{Q}^{2}\left(\sqrt{\frac{2{E}_{b}}{{N}_{0}}}\right)$$
$$M=4$$, e.g., DE-QPSK (equation 8.38 from [2]):
$${P}_{s}=4Q\left(\sqrt{\frac{2{E}_{b}}{{N}_{0}}}\right)-8{Q}^{2}\left(\sqrt{\frac{2{E}_{b}}{{N}_{0}}}\right)+8{Q}^{3}\left(\sqrt{\frac{2{E}_{b}}{{N}_{0}}}\right)-4{Q}^{4}\left(\sqrt{\frac{2{E}_{b}}{{N}_{0}}}\right)$$
From equation 5 in [3]:
$${P}_{b}=2Q\left(\sqrt{\frac{2{E}_{b}}{{N}_{0}}}\right)\left[1-Q\left(\sqrt{\frac{2{E}_{b}}{{N}_{0}}}\right)\right]$$
OQPSK. Same BER/SER as QPSK [2].
DE-OQPSK. Same BER/SER as DE-QPSK [3].
M-DPSK. From equation 8.84 in [2]:
$${P}_{s}=\frac{\mathrm{sin}(\pi /M)}{2\pi}{\displaystyle \underset{-\pi /2}{\overset{\pi /2}{\int}}\frac{\mathrm{exp}\left(-(k{E}_{b}/{N}_{0})(1-\mathrm{cos}(\pi /M)\mathrm{cos}\theta )\right)}{1-\mathrm{cos}(\pi /M)\mathrm{cos}\theta}d\theta}$$
The following expression is very close, but not strictly equal, to the exact BER [4]:
$${P}_{b}=\frac{1}{k}\left({\displaystyle \sum _{i=1}^{M/2}({w}_{i}^{\text{'}}){A}_{i}}\right)$$
where $${w}_{i}^{\text{'}}={w}_{i}+{w}_{M-i}$$, $${w}_{M/2}^{\text{'}}={w}_{M/2}$$, $${w}_{i}$$ is the Hamming weight of bits assigned to symbol i, and
$$\begin{array}{l}{A}_{i}=F\left(\left(2i+1\right)\frac{\pi}{M}\right)-F\left(\left(2i-1\right)\frac{\pi}{M}\right)\\ F(\psi )=-\frac{\mathrm{sin}\psi}{4\pi}{\displaystyle \underset{-\pi /2}{\overset{\pi /2}{\int}}\frac{\mathrm{exp}\left(-k{E}_{b}/{N}_{0}(1-\mathrm{cos}\psi \mathrm{cos}t)\right)}{1-\mathrm{cos}\psi \mathrm{cos}t}dt}\end{array}$$
Special case of $$M=2$$ (equation 8.85 from [2]):
$${P}_{b}=\frac{1}{2}\mathrm{exp}\left(-\frac{{E}_{b}}{{N}_{0}}\right)$$
M-PAM. From equations 8.3 and 8.7 in [2], and equation 5.2-46 in [1]:
$${P}_{s}=2\left(\frac{M-1}{M}\right)Q\left(\sqrt{\frac{6}{{M}^{2}-1}\frac{k{E}_{b}}{{N}_{0}}}\right)$$
From [5]:
$$\begin{array}{c}{P}_{b}=\frac{2}{M{\mathrm{log}}_{2}M}\times \\ {\displaystyle \sum _{k=1}^{{\mathrm{log}}_{2}M}\text{\hspace{0.17em}}\text{\hspace{0.17em}}{\displaystyle \sum _{i=0}^{(1-{2}^{-k})M-1}\left\{{(-1)}^{\lfloor \frac{i{2}^{k-1}}{M}\rfloor}\left({2}^{k-1}-\lfloor \frac{i{2}^{k-1}}{M}+\frac{1}{2}\rfloor \right)Q\left((2i+1)\sqrt{\frac{6{\mathrm{log}}_{2}M}{{M}^{2}-1}\frac{{E}_{b}}{{N}_{0}}}\right)\right\}}}\end{array}$$
M-QAM. For square M-QAM, $$k={\mathrm{log}}_{2}M$$ is even (equation 8.10 from [2], and equations 5.2-78 and 5.2-79 from [1]):
$${P}_{s}=4\frac{\sqrt{M}-1}{\sqrt{M}}Q\left(\sqrt{\frac{3}{M-1}\frac{k{E}_{b}}{{N}_{0}}}\right)-4{\left(\frac{\sqrt{M}-1}{\sqrt{M}}\right)}^{2}{Q}^{2}\left(\sqrt{\frac{3}{M-1}\frac{k{E}_{b}}{{N}_{0}}}\right)$$
From [5]:
$$\begin{array}{c}{P}_{b}=\frac{2}{\sqrt{M}{\mathrm{log}}_{2}\sqrt{M}}\\ \times {\displaystyle \sum _{k=1}^{{\mathrm{log}}_{2}\sqrt{M}}\text{\hspace{0.17em}}\text{\hspace{0.17em}}{\displaystyle \sum _{i=0}^{(1-{2}^{-k})\sqrt{M}-1}\left\{{(-1)}^{\lfloor \frac{i{2}^{k-1}}{\sqrt{M}}\rfloor}\left({2}^{k-1}-\lfloor \frac{i{2}^{k-1}}{\sqrt{M}}+\frac{1}{2}\rfloor \right)Q\left((2i+1)\sqrt{\frac{6{\mathrm{log}}_{2}M}{2(M-1)}\frac{{E}_{b}}{{N}_{0}}}\right)\right\}}}\end{array}$$
For rectangular (non-square) M-QAM, $$k={\mathrm{log}}_{2}M$$ is odd, $$M=I\times J$$, $$I={2}^{\frac{k-1}{2}}$$, and $$J={2}^{\frac{k+1}{2}}$$:
$$\begin{array}{c}{P}_{s}=\frac{4IJ-2I-2J}{M}\\ \times Q\left(\sqrt{\frac{6{\mathrm{log}}_{2}(IJ)}{({I}^{2}+{J}^{2}-2)}\frac{{E}_{b}}{{N}_{0}}}\right)-\frac{4}{M}(1+IJ-I-J){Q}^{2}\left(\sqrt{\frac{6{\mathrm{log}}_{2}(IJ)}{({I}^{2}+{J}^{2}-2)}\frac{{E}_{b}}{{N}_{0}}}\right)\end{array}$$
From [5]:
$${P}_{b}=\frac{1}{{\mathrm{log}}_{2}(IJ)}\left({\displaystyle \sum _{k=1}^{{\mathrm{log}}_{2}I}{P}_{I}(k)}+{\displaystyle \sum _{l=1}^{{\mathrm{log}}_{2}J}{P}_{J}(l)}\right)$$
where
$${P}_{I}(k)=\frac{2}{I}{\displaystyle \sum _{i=0}^{(1-{2}^{-k})I-1}\left\{{(-1)}^{\lfloor \frac{i{2}^{k-1}}{I}\rfloor}\left({2}^{k-1}-\lfloor \frac{i{2}^{k-1}}{I}+\frac{1}{2}\rfloor \right)Q\left((2i+1)\sqrt{\frac{6{\mathrm{log}}_{2}(IJ)}{{I}^{2}+{J}^{2}-2}\frac{{E}_{b}}{{N}_{0}}}\right)\right\}}$$
and
$${P}_{J}(k)=\frac{2}{J}{\displaystyle \sum _{j=0}^{(1-{2}^{-l})J-1}\left\{{(-1)}^{\lfloor \frac{j{2}^{l-1}}{J}\rfloor}\left({2}^{l-1}-\lfloor \frac{j{2}^{l-1}}{J}+\frac{1}{2}\rfloor \right)Q\left((2j+1)\sqrt{\frac{6{\mathrm{log}}_{2}(IJ)}{{I}^{2}+{J}^{2}-2}\frac{{E}_{b}}{{N}_{0}}}\right)\right\}}$$
Orthogonal M-FSK with Coherent Detection. From equation 8.40 in [2] and equation 5.2-21 in [1]:
$$\begin{array}{l}{P}_{s}=1-{\displaystyle \underset{-\infty}{\overset{\infty}{\int}}{\left[Q\left(-q-\sqrt{\frac{2k{E}_{b}}{{N}_{0}}}\right)\right]}^{M-1}\frac{1}{\sqrt{2\pi}}\mathrm{exp}\left(-\frac{{q}^{2}}{2}\right)dq}\\ {P}_{b}=\frac{{2}^{k-1}}{{2}^{k}-1}{P}_{s}\end{array}$$
Nonorthogonal 2-FSK with Coherent Detection. For $$M=2$$ (from equation 5.2-21 in [1] and equation 8.44 in [2]):
$${P}_{s}={P}_{b}=Q\left(\sqrt{\frac{{E}_{b}(1-\mathrm{Re}\left[\rho \right])}{{N}_{0}}}\right)$$
$$\rho $$is the complex correlation coefficient:
$$\rho =\frac{1}{2{E}_{b}}{\displaystyle \underset{0}{\overset{{T}_{b}}{\int}}{\tilde{s}}_{1}(t){\tilde{s}}_{2}^{*}(t)dt}$$
where $${\tilde{s}}_{1}(t)$$ and $${\tilde{s}}_{2}(t)$$ are complex lowpass signals, and
$${E}_{b}=\frac{1}{2}{\displaystyle \underset{0}{\overset{{T}_{b}}{\int}}{\left|{\tilde{s}}_{1}(t)\right|}^{2}dt}=\frac{1}{2}{\displaystyle \underset{0}{\overset{{T}_{b}}{\int}}{\left|{\tilde{s}}_{2}(t)\right|}^{2}dt}$$
For example:
$${\tilde{s}}_{1}(t)=\sqrt{\frac{2{E}_{b}}{{T}_{b}}}{e}^{j2\pi {f}_{1}t},\text{}{\tilde{s}}_{2}(t)=\sqrt{\frac{2{E}_{b}}{{T}_{b}}}{e}^{j2\pi {f}_{2}t}$$
$$\begin{array}{c}\rho =\frac{1}{2{E}_{b}}{\displaystyle \underset{0}{\overset{{T}_{b}}{\int}}\sqrt{\frac{2{E}_{b}}{{T}_{b}}}{e}^{j2\pi {f}_{1}t}\sqrt{\frac{2{E}_{b}}{{T}_{b}}}{e}^{-j2\pi {f}_{2}t}dt}=\frac{1}{{T}_{b}}{\displaystyle \underset{0}{\overset{{T}_{b}}{\int}}{e}^{j2\pi ({f}_{1}-{f}_{2})t}dt}\\ =\frac{\mathrm{sin}(\pi \Delta f{T}_{b})}{\pi \Delta f{T}_{b}}{e}^{j\pi \Delta ft}\end{array}$$
where $$\Delta f={f}_{1}-{f}_{2}$$.
$$\begin{array}{l}\text{}\mathrm{Re}\left[\rho \right]=\mathrm{Re}\left[\frac{\mathrm{sin}(\pi \Delta f{T}_{b})}{\pi \Delta f{T}_{b}}{e}^{j\pi \Delta ft}\right]=\frac{\mathrm{sin}(\pi \Delta f{T}_{b})}{\pi \Delta f{T}_{b}}\mathrm{cos}(\pi \Delta f{T}_{b})=\frac{\mathrm{sin}(2\pi \Delta f{T}_{b})}{2\pi \Delta f{T}_{b}}\\ \Rightarrow {P}_{b}=Q\left(\sqrt{\frac{{E}_{b}(1-\mathrm{sin}(2\pi \Delta f{T}_{b})/(2\pi \Delta f{T}_{b}))}{{N}_{0}}}\right)\end{array}$$
(from equation 8.44 in [2], where $$h=\Delta f{T}_{b}$$)
Orthogonal M-FSK with Noncoherent Detection. From equation 5.4-46 in [1] and equation 8.66 in [2]:
$$\begin{array}{l}{P}_{s}={\displaystyle \sum _{m=1}^{M-1}{(-1)}^{m+1}\left(\begin{array}{c}M-1\\ m\end{array}\right)\frac{1}{m+1}\mathrm{exp}\left[-\frac{m}{m+1}\frac{k{E}_{b}}{{N}_{0}}\right]}\\ {P}_{b}=\frac{1}{2}\frac{M}{M-1}{P}_{s}\end{array}$$
Nonorthogonal 2-FSK with Noncoherent Detection. For $$M=2$$ (from equation 5.4-53 in [1] and equation 8.69 in [2]):
$${P}_{s}={P}_{b}=Q(\sqrt{a},\sqrt{b})-\frac{1}{2}\mathrm{exp}\left(-\frac{a+b}{2}\right){I}_{0}(\sqrt{ab})$$
where
$$a=\frac{{E}_{b}}{2{N}_{0}}(1-\sqrt{1-{\left|\rho \right|}^{2}}),\text{}b=\frac{{E}_{b}}{2{N}_{0}}(1+\sqrt{1-{\left|\rho \right|}^{2}})\text{}$$
Precoded MSK with Coherent Detection. Same BER/SER as BPSK.
Differentially Encoded MSK with Coherent Detection. Same BER/SER as DE-BPSK.
MSK with Noncoherent Detection (Optimum Block-by-Block). Upper bound (from equations 10.166 and 10.164 in [6]):
$$\begin{array}{c}{P}_{s}={P}_{b}\\ \le \frac{1}{2}\left[1-Q\left(\sqrt{{b}_{1}},\sqrt{{a}_{1}}\right)+Q\left(\sqrt{{a}_{1}},\sqrt{{b}_{1}}\right)\right]+\frac{1}{4}\left[1-Q\left(\sqrt{{b}_{4}},\sqrt{{a}_{4}}\right)+Q\left(\sqrt{{a}_{4}},\sqrt{{b}_{4}}\right)\right]+\frac{1}{2}{e}^{-\frac{{E}_{b}}{{N}_{0}}}\end{array}$$
where
$$\begin{array}{cc}{a}_{1}=\frac{{E}_{b}}{{N}_{0}}\left(1-\sqrt{\frac{3-4/{\pi}^{2}}{4}}\right),& {b}_{1}=\frac{{E}_{b}}{{N}_{0}}\left(1+\sqrt{\frac{3-4/{\pi}^{2}}{4}}\right)\\ {a}_{4}=\frac{{E}_{b}}{{N}_{0}}\left(1-\sqrt{1-4/{\pi}^{2}}\right),& {b}_{4}=\frac{{E}_{b}}{{N}_{0}}\left(1+\sqrt{1-4/{\pi}^{2}}\right)\end{array}$$
CPFSK Coherent Detection (Optimum Block-by-Block). Lower bound (from equation 5.3-17 in [1]):
$${P}_{s}>{K}_{{\delta}_{\mathrm{min}}}Q\left(\sqrt{\frac{{E}_{b}}{{N}_{0}}{\delta}_{\mathrm{min}}^{2}}\right)$$
Upper bound:
$${\delta}_{\mathrm{min}}^{2}>\underset{1\le i\le M-1}{\mathrm{min}}\left\{2i\left(1-\text{sinc}(2ih)\right)\right\}$$
where h is the modulation index, and $${K}_{{\delta}_{\mathrm{min}}}$$ is the number of paths having the minimum distance.
$${P}_{b}\cong \frac{{P}_{s}}{k}$$
Notation. The following notation is used for the expressions found in berfading
.
Value | Notation |
---|---|
Power of the fading amplitude r | $$\Omega =E\left[{r}^{2}\right]$$, where $$E[\cdot ]$$ denotes statistical expectation |
Number of diversity branches | $$L$$ |
SNR per symbol per branch | $${\overline{\gamma}}_{l}=\left({\Omega}_{l}\frac{{E}_{s}}{{N}_{0}}\right)/L=\left({\Omega}_{l}\frac{k{E}_{b}}{{N}_{0}}\right)/L$$ For identically-distributed diversity branches: $$\overline{\gamma}=\left(\Omega \frac{k{E}_{b}}{{N}_{0}}\right)/L$$ |
Moment generating functions for each diversity branch | Rayleigh fading: $${M}_{{\gamma}_{l}}\left(s\right)=\frac{1}{1-s{\overline{\gamma}}_{l}}$$ $${M}_{{\gamma}_{l}}\left(s\right)=\frac{1+K}{1+K-s{\overline{\gamma}}_{l}}{e}^{\left[\frac{Ks{\overline{\gamma}}_{l}}{(1+K)-s{\overline{\gamma}}_{l}}\right]}$$ For identically-distributed diversity branches: $${M}_{{\gamma}_{l}}\left(s\right)={M}_{\gamma}\left(s\right)$$ for all l. |
The following acronyms are used:
Acronym | Definition |
---|---|
MRC | maximal-ratio combining |
EGC | equal-gain combining |
M-PSK with MRC. From equation 9.15 in [2]:
$${P}_{s}=\frac{1}{\pi}{\displaystyle \underset{0}{\overset{(M-1)\pi /M}{\int}}{\displaystyle \prod _{l=1}^{L}{M}_{{\gamma}_{l}}\left(-\frac{{\mathrm{sin}}^{2}(\pi /M)}{{\mathrm{sin}}^{2}\theta}\right)}d\theta}$$
$${P}_{b}=\frac{1}{k}\left({\displaystyle \sum _{i=1}^{M/2}({w}_{i}^{\text{'}}){\overline{P}}_{i}}\right)$$
where $${w}_{i}^{\text{'}}={w}_{i}+{w}_{M-i}$$, $${w}_{M/2}^{\text{'}}={w}_{M/2}$$, $${w}_{i}$$ is the Hamming weight of bits assigned to symbol i, and
$$\begin{array}{l}{\overline{P}}_{i}=\frac{1}{2\pi}{\displaystyle \underset{0}{\overset{\pi (1-(2i-1)/M)}{\int}}{\displaystyle \prod _{l=1}^{L}{M}_{{\gamma}_{l}}\left(-\frac{1}{{\mathrm{sin}}^{2}\theta}{\mathrm{sin}}^{2}\frac{(2i-1)\pi}{M}\right)}d\theta}\\ \text{\hspace{0.17em}}\text{\hspace{0.17em}}\text{\hspace{0.17em}}\text{\hspace{0.17em}}\text{\hspace{0.17em}}\text{\hspace{0.17em}}\text{\hspace{0.17em}}\text{\hspace{0.17em}}-\frac{1}{2\pi}{\displaystyle \underset{0}{\overset{\pi (1-(2i+1)/M)}{\int}}{\displaystyle \prod _{l=1}^{L}{M}_{{\gamma}_{l}}\left(-\frac{1}{{\mathrm{sin}}^{2}\theta}{\mathrm{sin}}^{2}\frac{(2i+1)\pi}{M}\right)}d\theta}\end{array}$$
For the special case of Rayleigh fading with $$M=2$$ (from equations C-18, C-21, and Table C-1 in [6]):
$${P}_{b}=\frac{1}{2}\left[1-\mu {\displaystyle \sum _{i=0}^{L-1}\left(\begin{array}{c}2i\\ i\end{array}\right)}{\left(\frac{1-{\mu}^{2}}{4}\right)}^{i}\right]$$
where
$$\mu =\sqrt{\frac{\overline{\gamma}}{\overline{\gamma}+1}}$$
If $$L=1$$:
$${P}_{b}=\frac{1}{2}\left[1-\sqrt{\frac{\overline{\gamma}}{\overline{\gamma}+1}}\right]$$
DE-M-PSK with MRC. For $$M=2$$ (from equations 8.37 and 9.8-9.11 in [2]):
$${P}_{s}={P}_{b}=\frac{2}{\pi}{\displaystyle \underset{0}{\overset{\pi /2}{\int}}{\displaystyle \prod _{l=1}^{L}{M}_{{\gamma}_{l}}\left(-\frac{1}{{\mathrm{sin}}^{2}\theta}\right)}d\theta}-\frac{2}{\pi}{\displaystyle \underset{0}{\overset{\pi /4}{\int}}{\displaystyle \prod _{l=1}^{L}{M}_{{\gamma}_{l}}\left(-\frac{1}{{\mathrm{sin}}^{2}\theta}\right)}d\theta}$$
M-PAM with MRC. From equation 9.19 in [2]:
$${P}_{s}=\frac{2(M-1)}{M\pi}{\displaystyle \underset{0}{\overset{\pi /2}{\int}}{\displaystyle \prod _{l=1}^{L}{M}_{{\gamma}_{l}}\left(-\frac{3/({M}^{2}-1)}{{\mathrm{sin}}^{2}\theta}\right)}d\theta}$$
$$\begin{array}{c}{P}_{b}=\frac{2}{\pi M{\mathrm{log}}_{2}M}\\ \times {\displaystyle \sum _{k=1}^{{\mathrm{log}}_{2}M}\text{}{\displaystyle \sum _{i=0}^{(1-{2}^{-k})M-1}\left\{{(-1)}^{\lfloor \frac{i{2}^{k-1}}{M}\rfloor}\left({2}^{k-1}-\lfloor \frac{i{2}^{k-1}}{M}+\frac{1}{2}\rfloor \right){\displaystyle \underset{0}{\overset{\pi /2}{\int}}{\displaystyle \prod _{l=1}^{L}{M}_{{\gamma}_{l}}\left(-\frac{{(2i+1)}^{2}3/({M}^{2}-1)}{{\mathrm{sin}}^{2}\theta}\right)}d\theta}\right\}}}\end{array}$$
M-QAM with MRC. For square M-QAM, $$k={\mathrm{log}}_{2}M$$ is even (equation 9.21 in [2]):
$$\begin{array}{c}{P}_{s}=\frac{4}{\pi}\left(1-\frac{1}{\sqrt{M}}\right){\displaystyle \underset{0}{\overset{\pi /2}{\int}}{\displaystyle \prod _{l=1}^{L}{M}_{{\gamma}_{l}}\left(-\frac{3/(2(M-1))}{{\mathrm{sin}}^{2}\theta}\right)}d\theta}\\ -\frac{4}{\pi}{\left(1-\frac{1}{\sqrt{M}}\right)}^{2}{\displaystyle \underset{0}{\overset{\pi /4}{\int}}{\displaystyle \prod _{l=1}^{L}{M}_{{\gamma}_{l}}\left(-\frac{3/(2(M-1))}{{\mathrm{sin}}^{2}\theta}\right)}d\theta}\end{array}$$
$$\begin{array}{c}{P}_{b}=\frac{2}{\pi \sqrt{M}{\mathrm{log}}_{2}\sqrt{M}}\\ \times {\displaystyle \sum _{k=1}^{{\mathrm{log}}_{2}\sqrt{M}}\text{}{\displaystyle \sum _{i=0}^{(1-{2}^{-k})\sqrt{M}-1}\left\{{(-1)}^{\lfloor \frac{i{2}^{k-1}}{\sqrt{M}}\rfloor}\left({2}^{k-1}-\lfloor \frac{i{2}^{k-1}}{\sqrt{M}}+\frac{1}{2}\rfloor \right){\displaystyle \underset{0}{\overset{\pi /2}{\int}}{\displaystyle \prod _{l=1}^{L}{M}_{{\gamma}_{l}}\left(-\frac{{(2i+1)}^{2}3/(2(M-1))}{{\mathrm{sin}}^{2}\theta}\right)}d\theta}\right\}}}\end{array}$$
For rectangular (nonsquare) M-QAM, $$k={\mathrm{log}}_{2}M$$ is odd, $$M=I\times J$$, $$I={2}^{\frac{k-1}{2}}$$, $$J={2}^{\frac{k+1}{2}}$$, $${\overline{\gamma}}_{l}={\Omega}_{l}{\mathrm{log}}_{2}(IJ)\frac{{E}_{b}}{{N}_{0}}$$, and
$$\begin{array}{c}{P}_{s}=\frac{4IJ-2I-2J}{M\pi}{\displaystyle \underset{0}{\overset{\pi /2}{\int}}{\displaystyle \prod _{l=1}^{L}{M}_{{\gamma}_{l}}\left(-\frac{3/({I}^{2}+{J}^{2}-2)}{{\mathrm{sin}}^{2}\theta}\right)}d\theta}\\ -\frac{4}{M\pi}(1+IJ-I-J){\displaystyle \underset{0}{\overset{\pi /4}{\int}}{\displaystyle \prod _{l=1}^{L}{M}_{{\gamma}_{l}}\left(-\frac{3/({I}^{2}+{J}^{2}-2)}{{\mathrm{sin}}^{2}\theta}\right)}d\theta}\end{array}$$
$$\begin{array}{c}{P}_{b}=\frac{1}{{\mathrm{log}}_{2}(IJ)}\left({\displaystyle \sum _{k=1}^{{\mathrm{log}}_{2}I}{P}_{I}(k)}+{\displaystyle \sum _{l=1}^{{\mathrm{log}}_{2}J}{P}_{J}(l)}\right)\\ {P}_{I}(k)=\frac{2}{I\pi}{\displaystyle \sum _{i=0}^{(1-{2}^{-k})I-1}\left\{{(-1)}^{\lfloor \frac{i{2}^{k-1}}{I}\rfloor}\left({2}^{k-1}-\lfloor \frac{i{2}^{k-1}}{I}+\frac{1}{2}\rfloor \right){\displaystyle \underset{0}{\overset{\pi /2}{\int}}{\displaystyle \prod _{l=1}^{L}{M}_{{\gamma}_{l}}\left(-\frac{{(2i+1)}^{2}3/({I}^{2}+{J}^{2}-2)}{{\mathrm{sin}}^{2}\theta}\right)}d\theta}\right\}}\\ {P}_{J}(k)=\frac{2}{J\pi}{\displaystyle \sum _{j=0}^{(1-{2}^{-l})J-1}\left\{{(-1)}^{\lfloor \frac{j{2}^{l-1}}{J}\rfloor}\left({2}^{l-1}-\lfloor \frac{j{2}^{l-1}}{J}+\frac{1}{2}\rfloor \right){\displaystyle \underset{0}{\overset{\pi /2}{\int}}{\displaystyle \prod _{l=1}^{L}{M}_{{\gamma}_{l}}\left(-\frac{{(2j+1)}^{2}3/({I}^{2}+{J}^{2}-2)}{{\mathrm{sin}}^{2}\theta}\right)}d\theta}\right\}}\end{array}$$
M-DPSK with Postdetection EGC. From equation 8.165 in [2]:
$${P}_{s}=\frac{\mathrm{sin}(\pi /M)}{2\pi}{\displaystyle \underset{-\pi /2}{\overset{\pi /2}{\int}}\frac{1}{\left[1-\mathrm{cos}(\pi /M)\mathrm{cos}\theta \right]}{\displaystyle \prod _{l=1}^{L}{M}_{{\gamma}_{l}}\left(-\left[1-\mathrm{cos}(\pi /M)\mathrm{cos}\theta \right]\right)}d\theta}$$
$${P}_{b}=\frac{1}{k}\left({\displaystyle \sum _{i=1}^{M/2}({w}_{i}^{\text{'}}){\overline{A}}_{i}}\right)$$
where $${w}_{i}^{\text{'}}={w}_{i}+{w}_{M-i}$$, $${w}_{M/2}^{\text{'}}={w}_{M/2}$$, $${w}_{i}$$ is the Hamming weight of bits assigned to symbol i, and
$$\begin{array}{l}{\overline{A}}_{i}=\overline{F}\left(\left(2i+1\right)\frac{\pi}{M}\right)-\overline{F}\left(\left(2i-1\right)\frac{\pi}{M}\right)\\ \overline{F}(\psi )=-\frac{\mathrm{sin}\psi}{4\pi}{\displaystyle \underset{-\pi /2}{\overset{\pi /2}{\int}}\frac{1}{\left(1-\mathrm{cos}\psi \mathrm{cos}t\right)}{\displaystyle \prod _{l=1}^{L}{M}_{{\gamma}_{l}}\left(-\left(1-\mathrm{cos}\psi \mathrm{cos}t\right)\right)}dt}\end{array}$$
For the special case of Rayleigh fading with $$M=2$$, and $$L=1$$ (equation 8.173 from [2]):
$${P}_{b}=\frac{1}{2(1+\overline{\gamma})}$$
Orthogonal 2-FSK, Coherent Detection with MRC. From equation 9.11 in [2]:
$${P}_{s}={P}_{b}=\frac{1}{\pi}{\displaystyle \underset{0}{\overset{\pi /2}{\int}}{\displaystyle \prod _{l=1}^{L}{M}_{{\gamma}_{l}}\left(-\frac{1/2}{{\mathrm{sin}}^{2}\theta}\right)}d\theta}$$
For the special case of Rayleigh fading (equations 14.4-15 and 14.4-21 in [1]):
$${P}_{s}={P}_{b}\text{}=\frac{1}{{2}^{L}}{\left(1-\sqrt{\frac{\overline{\gamma}}{2+\overline{\gamma}}}\right)}^{L}{\displaystyle \sum _{k=0}^{L-1}\left(\begin{array}{c}L-1+k\\ k\end{array}\right)\frac{1}{{2}^{k}}}{\left(1+\sqrt{\frac{\overline{\gamma}}{2+\overline{\gamma}}}\right)}^{k}$$
Nonorthogonal 2-FSK, Coherent Detection with MRC. Equations 9.11 and 8.44 in [2]:
$${P}_{s}={P}_{b}=\frac{1}{\pi}{\displaystyle \underset{0}{\overset{\pi /2}{\int}}{\displaystyle \prod _{l=1}^{L}{M}_{{\gamma}_{l}}\left(-\frac{(1-\mathrm{Re}\left[\rho \right])/2}{{\mathrm{sin}}^{2}\theta}\right)}d\theta}$$
For the special case of Rayleigh fading with $$L=1$$ (equation 20 in [8] and equation 8.130 in [2]):
$${P}_{s}={P}_{b}=\frac{1}{2}\left[1-\sqrt{\frac{\overline{\gamma}(1-\mathrm{Re}[\rho ])}{2+\overline{\gamma}(1-\mathrm{Re}[\rho ])}}\right]$$
Orthogonal M-FSK, Noncoherent Detection with EGC. Rayleigh fading (equation 14.4-47 in [1]):
$$\begin{array}{l}{P}_{s}\text{}=1-{\displaystyle \underset{0}{\overset{\infty}{\int}}\frac{1}{{\left(1+\overline{\gamma}\right)}^{L}\left(L-1\right)!}{U}^{L-1}{e}^{-\frac{U}{1+\overline{\gamma}}}{\left(1-{e}^{-U}{\displaystyle \sum _{k=0}^{L-1}\frac{{U}^{k}}{k!}}\right)}^{M-1}dU}\\ {P}_{b}=\frac{1}{2}\frac{M}{M-1}{P}_{s}\end{array}$$
Rician fading (equation 41 in [8]):
$$\begin{array}{l}{P}_{s}={\displaystyle \sum _{r=1}^{M-1}\frac{{(-1)}^{r+1}{e}^{-LK{\overline{\gamma}}_{r}/(1+{\overline{\gamma}}_{r})}}{{\left(r(1+{\overline{\gamma}}_{r})+1\right)}^{L}}\left(\begin{array}{c}M-1\\ r\end{array}\right)}{\displaystyle \sum _{n=0}^{r(L-1)}{\beta}_{nr}\frac{\Gamma (L+n)}{\Gamma (L)}{\left[\frac{1+{\overline{\gamma}}_{r}}{r+1+r{\overline{\gamma}}_{r}}\right]}^{n}}{}_{1}F{}_{1}\left(L+n,L;\frac{LK{\overline{\gamma}}_{r}/(1+{\overline{\gamma}}_{r})}{r(1+{\overline{\gamma}}_{r})+1}\right)\\ {P}_{b}=\frac{1}{2}\frac{M}{M-1}{P}_{s}\end{array}$$
where
$$\begin{array}{c}{\overline{\gamma}}_{r}=\frac{1}{1+K}\overline{\gamma}\\ {\beta}_{nr}={\displaystyle \sum _{i=n-(L-1)}^{n}\frac{{\beta}_{i(r-1)}}{(n-i)!}{I}_{[0,\text{\hspace{0.17em}}(r-1)(L-1)]}(i)}\\ {\beta}_{00}={\beta}_{0r}=1\\ {\beta}_{n1}=1/n!\\ {\beta}_{1r}=r\end{array}$$
and $${I}_{[a,b]}(i)=1$$ if $$a\le i\le b$$ and 0 otherwise.
Nonorthogonal 2-FSK, Noncoherent Detection with No Diversity. From equation 8.163 in [2]:
$${P}_{s}={P}_{b}=\frac{1}{4\pi}{\displaystyle \underset{-\pi}{\overset{\pi}{\int}}\frac{1-{\varsigma}^{2}}{1+2\varsigma \mathrm{sin}\theta +{\varsigma}^{2}}{M}_{\gamma}\left(-\frac{1}{4}(1+\sqrt{1-{\rho}^{2}})(1+2\varsigma \mathrm{sin}\theta +{\varsigma}^{2})\right)d\theta}$$
where
$$\varsigma =\sqrt{\frac{1-\sqrt{1-{\rho}^{2}}}{1+\sqrt{1-{\rho}^{2}}}}$$
Common Notation for This Section
Description | Notation |
---|---|
Energy-per-information bit-to-noise power-spectral-density ratio | $${\gamma}_{b}=\frac{{E}_{b}}{{N}_{0}}$$ |
Message length | $$K$$ |
Code length | $$N$$ |
Code rate | $${R}_{c}=\frac{K}{N}$$ |
Block Coding. Specific notation for block coding expressions: $${d}_{\mathrm{min}}$$ is the minimum distance of the code.
Soft DecisionBPSK, QPSK, OQPSK, PAM-2, QAM-4, and precoded MSK (equation 8.1-52 in [1]):
$${P}_{b}\le \frac{1}{2}({2}^{K}-1)Q\left(\sqrt{2{\gamma}_{b}{R}_{c}{d}_{\mathrm{min}}}\right)$$
DE-BPSK, DE-QPSK, DE-OQPSK, and DE-MSK:
$${P}_{b}\le \frac{1}{2}({2}^{K}-1)\left[2Q\left(\sqrt{2{\gamma}_{b}{R}_{c}{d}_{\mathrm{min}}}\right)\left[1-Q\left(\sqrt{2{\gamma}_{b}{R}_{c}{d}_{\mathrm{min}}}\right)\right]\right]$$
BFSK, coherent detection (equations 8.1-50 and 8.1-58 in [1]):
$${P}_{b}\le \frac{1}{2}({2}^{K}-1)Q\left(\sqrt{{\gamma}_{b}{R}_{c}{d}_{\mathrm{min}}}\right)$$
BFSK, noncoherent square-law detection (equations 8.1-65 and 8.1-64 in [1]):
$${P}_{b}\le \frac{1}{2}\frac{{2}^{K}-1}{{2}^{2{d}_{\mathrm{min}}-1}}\mathrm{exp}\left(-\frac{1}{2}{\gamma}_{b}{R}_{c}{d}_{\mathrm{min}}\right){\displaystyle \sum _{i=0}^{{d}_{\mathrm{min}}-1}{\left(\frac{1}{2}{\gamma}_{b}{R}_{c}{d}_{\mathrm{min}}\right)}^{i}\frac{1}{i!}{\displaystyle \sum _{r=0}^{{d}_{\mathrm{min}}-1-i}\left(\begin{array}{c}2{d}_{\mathrm{min}}-1\\ r\end{array}\right)}}$$
DPSK:
$${P}_{b}\le \frac{1}{2}\frac{{2}^{K}-1}{{2}^{2{d}_{\mathrm{min}}-1}}\mathrm{exp}\left(-{\gamma}_{b}{R}_{c}{d}_{\mathrm{min}}\right){\displaystyle \sum _{i=0}^{{d}_{\mathrm{min}}-1}{\left({\gamma}_{b}{R}_{c}{d}_{\mathrm{min}}\right)}^{i}\frac{1}{i!}{\displaystyle \sum _{r=0}^{{d}_{\mathrm{min}}-1-i}\left(\begin{array}{c}2{d}_{\mathrm{min}}-1\\ r\end{array}\right)}}$$
General linear block code (equations 4.3, 4.4 in [9], and 12.136 in [6]):
$$\begin{array}{l}{P}_{b}\le \frac{1}{N}{\displaystyle \sum _{m=t+1}^{N}(m+t)\left(\begin{array}{c}N\\ m\end{array}\right){p}^{m}{\left(1-p\right)}^{N-m}}\\ t=\lfloor \frac{1}{2}\left({d}_{\mathrm{min}}-1\right)\rfloor \end{array}$$
Hamming code (equations 4.11, 4.12 in [9], and 6.72, 6.73 in [7]):
$${P}_{b}\approx \frac{1}{N}{\displaystyle \sum _{m=2}^{N}m\left(\begin{array}{c}N\\ m\end{array}\right){p}^{m}{\left(1-p\right)}^{N-m}}=p-p{(1-p)}^{N-1}$$
(24, 12) extended Golay code (equation 4.17 in [9], and 12.139 in [6]):
$${P}_{b}\le \frac{1}{24}{\displaystyle \sum _{m=4}^{24}{\beta}_{m}\left(\begin{array}{c}24\\ m\end{array}\right){p}^{m}{\left(1-p\right)}^{24-m}}$$
where $${\beta}_{m}$$ is the average number of channel symbol errors that remain in corrected N-tuple when the channel caused m symbol errors (table 4.2 in [9]).
Reed-Solomon code with $$N=Q-1={2}^{q}-1$$:
$${P}_{b}\approx \frac{{2}^{q-1}}{{2}^{q}-1}\frac{1}{N}{\displaystyle \sum _{m=t+1}^{N}m\left(\begin{array}{c}N\\ m\end{array}\right){\left({P}_{s}\right)}^{m}{(1-{P}_{s})}^{N-m}}$$
for FSK (equations 4.25, 4.27 in [9], 8.1-115, 8.1-116 in [1], 8.7, 8.8 in [7], and 12.142, 12.143 in [6]), and
$${P}_{b}\approx \frac{1}{q}\frac{1}{N}{\displaystyle \sum _{m=t+1}^{N}m\left(\begin{array}{c}N\\ m\end{array}\right){\left({P}_{s}\right)}^{m}{(1-{P}_{s})}^{N-m}}$$
otherwise.
If $${\mathrm{log}}_{2}Q/{\mathrm{log}}_{2}M=q/k=h$$ where h is an integer (equation 1 in [10]):
$${P}_{s}=1-{(1-s)}^{h}$$
where s is the symbol error rate (SER) in an uncoded AWGN channel.
For example, for BPSK, $$M=2$$ and $${P}_{s}=1-{(1-s)}^{q}$$
Otherwise, $${P}_{s}$$ is given by table 1 and equation 2 in [10].
Convolutional Coding. Specific notation for convolutional coding expressions: $${d}_{free}$$ is the free distance of the code, and $${a}_{d}$$ is the number of paths of distance d from the all-zero path that merge with the all-zero path for the first time.
Soft DecisionFrom equations 8.2-26, 8.2-24, and 8.2-25 in [1], and equations 13.28 and 13.27 in [6]:
$${P}_{b}<{\displaystyle \sum _{d={d}_{free}}^{\infty}{a}_{d}f(d){P}_{2}(d)}$$
with transfer function
$$\begin{array}{l}T(D,N)={\displaystyle \sum _{d={d}_{free}}^{\infty}{a}_{d}{D}^{d}{N}^{f(d)}}\\ {\frac{dT(D,N)}{dN}|}_{N=1}={\displaystyle \sum _{d={d}_{free}}^{\infty}{a}_{d}f(d){D}^{d}}\end{array}$$
where $$f(d)$$ is the exponent of N as a function of d.
Results for BPSK, QPSK, OQPSK, PAM-2, QAM-4, precoded MSK, DE-BPSK, DE-QPSK, DE-OQPSK, DE-MSK, DPSK, and BFSK are obtained as:
$${P}_{2}(d)={{P}_{b}|}_{\frac{{E}_{b}}{{N}_{0}}={\gamma}_{b}{R}_{c}d}$$
where $${P}_{b}$$ is the BER in the corresponding uncoded AWGN channel. For example, for BPSK (equation 8.2-20 in [1]):
$${P}_{2}(d)=Q\left(\sqrt{2{\gamma}_{b}{R}_{c}d}\right)$$
From equations 8.2-33, 8.2-28, and 8.2-29 in [1], and equations 13.28, 13.24, and 13.25 in [6]:
$${P}_{b}<{\displaystyle \sum _{d={d}_{free}}^{\infty}{a}_{d}f(d){P}_{2}(d)}$$
where
$${P}_{2}(d)={\displaystyle \sum _{k=(d+1)/2}^{d}\left(\begin{array}{c}d\\ k\end{array}\right){p}^{k}{(1-p)}^{d-k}}$$
when d is odd, and
$${P}_{2}(d)={\displaystyle \sum _{k=d/2+1}^{d}\left(\begin{array}{c}d\\ k\end{array}\right){p}^{k}{(1-p)}^{d-k}}+\frac{1}{2}\left(\begin{array}{c}d\\ d/2\end{array}\right){p}^{d/2}{(1-p)}^{d/2}$$
when d is even (p is the bit error rate (BER) in an uncoded AWGN channel).
One way to compute the bit error rate or symbol error rate for a communication system is to simulate the transmission of data messages and compare all messages before and after transmission. The simulation of the communication system components using Communications System Toolbox™ is covered in other parts of this guide. This section describes how to compare the data messages that enter and leave the simulation.
Another example of computing performance results via simulation is in Curve Fitting for Error Rate Plots in the discussion of curve fitting.
The biterr
function compares two sets of
data and computes the number of bit errors and the bit error rate.
The symerr
function compares two sets of data
and computes the number of symbol errors and the symbol error rate.
An error is a discrepancy between corresponding points in the two
sets of data.
Of the two sets of data, typically one represents messages entering a transmitter and the other represents recovered messages leaving a receiver. You might also compare data entering and leaving other parts of your communication system, for example, data entering an encoder and data leaving a decoder.
If your communication system uses several bits to represent one symbol, counting bit errors is different from counting symbol errors. In either the bit- or symbol-counting case, the error rate is the number of errors divided by the total number (of bits or symbols) transmitted.
Note: To ensure an accurate error rate, you should typically simulate enough data to produce at least 100 errors. |
If the error rate is very small (for example, 10^{-6} or smaller), the semianalytic technique might compute the result more quickly than a simulation-only approach. See Performance Results via the Semianalytic Technique for more information on how to use this technique.
The script below uses the symerr
function
to compute the symbol error rates for a noisy linear block code. After
artificially adding noise to the encoded message, it compares the
resulting noisy code to the original code. Then it decodes and compares
the decoded message to the original one.
m = 3; n = 2^m-1; k = n-m; % Prepare to use Hamming code. msg = randi([0 1],k*200,1); % 200 messages of k bits each code = encode(msg,n,k,'hamming'); codenoisy = rem(code+(rand(n*200,1)>.95),2); % Add noise. % Decode and correct some errors. newmsg = decode(codenoisy,n,k,'hamming'); % Compute and display symbol error rates. noisyVec = step(comm.ErrorRate,code,codenoisy); decodedVec = step(comm.ErrorRate,msg,newmsg); disp(['Error rate in the received code: ',num2str(noisyVec(1))]) disp(['Error rate after decoding: ',num2str(decodedVec(1))])
The output is below. The error rate decreases after decoding because the Hamming decoder corrects some of the errors. Your results might vary because this example uses random numbers.
Error rate in the received code: 0.054286 Error rate after decoding: 0.03
In the example above, the symbol errors and bit errors are the same because each symbol is a bit. The commands below illustrate the difference between symbol errors and bit errors in other situations.
a = [1 2 3]'; b = [1 4 4]'; format rat % Display fractions instead of decimals. % Create ErrorRate Calculator System object serVec = step(comm.ErrorRate,a,b); srate = serVec(1) snum = serVec(2) % Convert integers to bits hIntToBit = comm.IntegerToBit(3); a_bit = step(hIntToBit, a); b_bit = step(hIntToBit, b); % Calculate BER berVec = step(comm.ErrorRate,a_bit,b_bit); brate = berVec(1) bnum = berVec(2)
The output is below.
snum = 2 srate = 2/3 bnum = 5 brate = 5/9
bnum
is 5 because the second entries differ
in two bits and the third entries differ in three bits. brate
is
5/9 because the total number of bits is 9. The total number of bits
is, by definition, the number of entries in a
or b
times
the maximum number of bits among all entries of a
and b
.
The technique described in Performance Results via Simulation works well for a large variety of communication systems, but can be prohibitively time-consuming if the system's error rate is very small (for example, 10^{-6} or smaller). This section describes how to use the semianalytic technique as an alternative way to compute error rates. For certain types of systems, the semianalytic technique can produce results much more quickly than a nonanalytic method that uses only simulated data.
The semianalytic technique uses a combination of simulation
and analysis to determine the error rate of a communication system.
The semianalytic
function in Communications System Toolbox™ helps
you implement the semianalytic technique by performing some of the
analysis.
The semianalytic technique works well for certain types of communication systems, but not for others. The semianalytic technique is applicable if a system has all of these characteristics:
Any effects of multipath fading, quantization, and amplifier nonlinearities must precede the effects of noise in the actual channel being modeled.
The receiver is perfectly synchronized with the carrier, and timing jitter is negligible. Because phase noise and timing jitter are slow processes, they reduce the applicability of the semianalytic technique to a communication system.
The noiseless simulation has no errors in the received signal constellation. Distortions from sources other than noise should be mild enough to keep each signal point in its correct decision region. If this is not the case, the calculated BER is too low. For instance, if the modeled system has a phase rotation that places the received signal points outside their proper decision regions, the semianalytic technique is not suitable to predict system performance.
Furthermore, the semianalytic
function
assumes that the noise in the actual channel being modeled is Gaussian.
For details on how to adapt the semianalytic technique for non-Gaussian
noise, see the discussion of generalized exponential distributions
in [11].
The procedure below describes how you would typically implement
the semianalytic technique using the semianalytic
function:
Generate a message signal containing at least M^{L} symbols,
where M is the alphabet size of the modulation and L is the length
of the impulse response of the channel in symbols. A common approach
is to start with an augmented binary pseudonoise (PN) sequence of
total length (log_{2}M)M^{L}
.
An augmented PN sequence is a PN sequence with
an extra zero appended, which makes the distribution of ones and zeros
equal.
Modulate a carrier with the message signal using baseband modulation.
Supported modulation types are listed on the reference page for semianalytic
.
Shape the resultant signal with rectangular pulse shaping, using
the oversampling factor that you will later use to filter the modulated
signal. Store the result of this step as txsig
for
later use.
Filter the modulated signal with a transmit filter. This filter is often a square-root raised cosine filter, but you can also use a Butterworth, Bessel, Chebyshev type 1 or 2, elliptic, or more general FIR or IIR filter. If you use a square-root raised cosine filter, use it on the nonoversampled modulated signal and specify the oversampling factor in the filtering function. If you use another filter type, you can apply it to the rectangularly pulse shaped signal.
Run the filtered signal through a noiseless channel.
This channel can include multipath fading effects, phase shifts,
amplifier nonlinearities, quantization, and additional filtering,
but it must not include noise. Store the result of this step as rxsig
for
later use.
Invoke the semianalytic
function
using the txsig
and rxsig
data
from earlier steps. Specify a receive filter as a pair of input arguments,
unless you want to use the function's default filter. The function
filters rxsig
and then determines the error probability
of each received signal point by analytically applying the Gaussian
noise distribution to each point. The function averages the error
probabilities over the entire received signal to determine the overall
error probability. If the error probability calculated in this way
is a symbol error probability, the function converts it to a bit error
rate, typically by assuming Gray coding. The function returns the
bit error rate (or, in the case of DQPSK modulation, an upper bound
on the bit error rate).
The example below illustrates the procedure described above,
using 16-QAM modulation. It also compares the error rates obtained
from the semianalytic technique with the theoretical error rates obtained
from published formulas and computed using the berawgn
function.
The resulting plot shows that the error rates obtained using the two
methods are nearly identical. The discrepancies between the theoretical
and computed error rates are largely due to the phase offset in this
example's channel model.
% Step 1. Generate message signal of length >= M^L. M = 16; % Alphabet size of modulation L = 1; % Length of impulse response of channel msg = [0:M-1 0]; % M-ary message sequence of length > M^L % Step 2. Modulate the message signal using baseband modulation. hMod = comm.RectangularQAMModulator(M); % Use 16-QAM. modsig = step(hMod,msg'); % Modulate data Nsamp = 16; modsig = rectpulse(modsig,Nsamp); % Use rectangular pulse shaping. % Step 3. Apply a transmit filter. txsig = modsig; % No filter in this example % Step 4. Run txsig through a noiseless channel. rxsig = txsig*exp(1i*pi/180); % Static phase offset of 1 degree % Step 5. Use the semianalytic function. % Specify the receive filter as a pair of input arguments. % In this case, num and den describe an ideal integrator. num = ones(Nsamp,1)/Nsamp; den = 1; EbNo = 0:20; % Range of Eb/No values under study ber = semianalytic(txsig,rxsig,'qam',M,Nsamp,num,den,EbNo); % For comparison, calculate theoretical BER. bertheory = berawgn(EbNo,'qam',M); % Plot computed BER and theoretical BER. figure; semilogy(EbNo,ber,'k*'); hold on; semilogy(EbNo,bertheory,'ro'); title('Semianalytic BER Compared with Theoretical BER'); legend('Semianalytic BER with Phase Offset',... 'Theoretical BER Without Phase Offset','Location','SouthWest'); hold off;
This example creates a figure like the one below.
While the biterr
function discussed above
can help you gather empirical error statistics, you might also compare
those results to theoretical error statistics. Certain types of communication
systems are associated with closed-form expressions for the bit error
rate or a bound on it. The functions listed in the table below compute
the closed-form expressions for some types of communication systems,
where such expressions exist.
Type of Communication System | Function |
---|---|
Uncoded AWGN channel | berawgn |
Coded AWGN channel | bercoding |
Uncoded Rayleigh and Rician fading channel | berfading |
Uncoded AWGN channel with imperfect synchronization | bersync |
Each function's reference page lists one or more books containing the closed-form expressions that the function implements.
The example below uses the bercoding
function
to compute upper bounds on bit error rates for convolutional coding
with a soft-decision decoder. The data used for the generator and
distance spectrum are from [1] and [12], respectively.
coderate = 1/4; % Code rate % Create a structure dspec with information about distance spectrum. dspec.dfree = 10; % Minimum free distance of code dspec.weight = [1 0 4 0 12 0 32 0 80 0 192 0 448 0 1024 ... 0 2304 0 5120 0]; % Distance spectrum of code EbNo = 3:0.5:8; berbound = bercoding(EbNo,'conv','soft',coderate,dspec); semilogy(EbNo,berbound) % Plot the results. xlabel('E_b/N_0 (dB)'); ylabel('Upper Bound on BER'); title('Theoretical Bound on BER for Convolutional Coding'); grid on;
This example produces the following plot.
The example below uses the berawgn
function
to compute symbol error rates for pulse amplitude modulation (PAM)
with a series of E_{b}/N_{0} values. For comparison, the code simulates
8-PAM with an AWGN channel and computes empirical symbol error rates.
The code also plots the theoretical and empirical symbol error rates
on the same set of axes.
% 1. Compute theoretical error rate using BERAWGN. rng('default') % Set random number seed for repeatability % M = 8; EbNo = 0:13; [ber, ser] = berawgn(EbNo,'pam',M); % Plot theoretical results. figure; semilogy(EbNo,ser,'r'); xlabel('E_b/N_0 (dB)'); ylabel('Symbol Error Rate'); grid on; drawnow; % 2. Compute empirical error rate by simulating. % Set up. n = 10000; % Number of symbols to process k = log2(M); % Number of bits per symbol % Convert from EbNo to SNR. % Note: Because No = 2*noiseVariance^2, we must add 3 dB % to get SNR. For details, see Proakis' book listed in % "Selected Bibliography for Performance Evaluation." snr = EbNo+3+10*log10(k); % Preallocate variables to save time. ynoisy = zeros(n,length(snr)); z = zeros(n,length(snr)); berVec = zeros(3,length(EbNo)); % PAM modulation and demodulation system objects h = comm.PAMModulator(M); h2 = comm.PAMDemodulator(M); % AWGNChannel System object hChan = comm.AWGNChannel('NoiseMethod', 'Signal to noise ratio (SNR)'); % ErrorRate calculator System object to compare decoded symbols to the % original transmitted symbols. hErrorCalc = comm.ErrorRate; % Main steps in the simulation x = randi([0 M-1],n,1); % Create message signal. y = step(h,x); % Modulate. hChan.SignalPower = (real(y)' * real(y))/ length(real(y)); % Loop over different SNR values. for jj = 1:length(snr) reset(hErrorCalc) hChan.SNR = snr(jj); % Assign Channel SNR ynoisy(:,jj) = step(hChan,real(y)); % Add AWGN z(:,jj) = step(h2,complex(ynoisy(:,jj))); % Demodulate. % Compute symbol error rate from simulation. berVec(:,jj) = step(hErrorCalc, x, z(:,jj)); end % 3. Plot empirical results, in same figure. hold on; semilogy(EbNo,berVec(1,:),'b.'); legend('Theoretical SER','Empirical SER'); title('Comparing Theoretical and Empirical Error Rates'); hold off;
This example produces a plot like the one in the following figure. Your plot might vary because the simulation uses random numbers.
Error rate plots provide a visual way to examine the performance of a communication system, and they are often included in publications. This section mentions some of the tools you can use to create error rate plots, modify them to suit your needs, and do curve fitting on error rate data. It also provides an example of curve fitting. For more detailed discussions about the more general plotting capabilities in MATLAB^{®}, see the MATLAB documentation set.
semilogy
In many error rate plots, the horizontal axis indicates E_{b}/N_{0} values
in dB and the vertical axis indicates the error rate using a logarithmic
(base 10) scale. To see an example of such a plot, as well as the
code that creates it, see Comparing Theoretical and Empirical Error Rates. The part
of that example that creates the plot uses the semilogy
function
to produce a logarithmic scale on the vertical axis and a linear scale
on the horizontal axis.
Other examples that illustrate the use of semilogy
are
in these sections:
Example: Using the Semianalytic Technique, which also illustrates
Plotting two sets of data on one pair of axes
Adding a title
Adding a legend
Plotting Theoretical Error Rates, which also illustrates
Adding axis labels
Adding grid lines
Curve fitting is useful when you have a small or imperfect data
set but want to plot a smooth curve for presentation purposes. The berfit
function
in Communications System Toolbox offers curve-fitting
capabilities that are well suited to the situation when the empirical
data describes error rates at different E_{b}/N_{0} values. This function
enables you to
Customize various relevant aspects of the curve-fitting process, such as the type of closed-form function (from a list of preset choices) used to generate the fit.
Plot empirical data along with a curve that berfit
fits
to the data.
Interpolate points on the fitted curve between E_{b}/N_{0} values in your empirical data set to make the plot smoother looking.
Collect relevant information about the fit, such as the numerical values of points along the fitted curve and the coefficients of the fit expression.
Note:
The |
For a full list of inputs and outputs for berfit
,
see its reference page.
This example simulates a simple DBPSK (differential binary phase
shift keying) communication system and plots error rate data for a
series of E_{b}/N_{0} values. It uses the berfit
function
to fit a curve to the somewhat rough set of empirical error rates.
Because the example is long, this discussion presents it in multiple
steps:
Setting Up Parameters for the Simulation. The first step in the example sets up the parameters to be used during the simulation. Parameters include the range of E_{b}/N_{0} values to consider and the minimum number of errors that must occur before the simulation computes an error rate for that E_{b}/N_{0} value.
Note:
For most applications, you should base an error rate computation
on a larger number of errors than is used here (for instance, you
might change |
% Set up initial parameters. siglen = 100000; % Number of bits in each trial M = 2; % DBPSK is binary. % DBPSK modulation and demodulation System objects hMod = comm.DBPSKModulator; hDemod = comm.DBPSKDemodulator; % AWGNChannel System object hChan = comm.AWGNChannel('NoiseMethod', 'Signal to noise ratio (SNR)'); % ErrorRate calculator System object to compare decoded symbols to the % original transmitted symbols. hErrorCalc = comm.ErrorRate; EbNomin = 0; EbNomax = 9; % EbNo range, in dB numerrmin = 5; % Compute BER only after 5 errors occur. EbNovec = EbNomin:1:EbNomax; % Vector of EbNo values numEbNos = length(EbNovec); % Number of EbNo values % Preallocate space for certain data. ber = zeros(1,numEbNos); % final BER values berVec = zeros(3,numEbNos); % Updated BER values intv = cell(1,numEbNos); % Cell array of confidence intervals
Simulating the System Using a Loop. The next step in the example is to use a for
loop
to vary the E_{b}/N_{0} value (denoted by EbNo
in the
code) and simulate the communication system for each value. The inner while
loop
ensures that the simulation continues to use a given EbNo
value
until at least the predefined minimum number of errors has occurred.
When the system is very noisy, this requires only one pass through
the while
loop, but in other cases, this requires
multiple passes.
The communication system simulation uses these toolbox functions:
randi
to generate a random message
sequence
dpskmod
to perform DBPSK modulation
awgn
to model a channel with
additive white Gaussian noise
dpskdemod
to perform DBPSK demodulation
biterr
to compute the number
of errors for a given pass through the while
loop
berconfint
to compute the final
error rate and confidence interval for a given value of EbNo
As the example progresses through the for
loop,
it collects data for later use in curve fitting and plotting:
ber
, a vector containing the bit
error rates for the series of EbNo
values.
intv
, a cell array containing the
confidence intervals for the series of EbNo
values.
Each entry in intv
is a two-element vector that
gives the endpoints of the interval.
% Loop over the vector of EbNo values. berVec = zeros(3,numEbNos); % Reset for jj = 1:numEbNos EbNo = EbNovec(jj); snr = EbNo; % Because of binary modulation reset(hErrorCalc) hChan.SNR = snr; % Assign Channel SNR % Simulate until numerrmin errors occur. while (berVec(2,jj) < numerrmin) msg = randi([0,M-1], siglen, 1); % Generate message sequence. txsig = step(hMod, msg); % Modulate. hChan.SignalPower = (txsig'*txsig)/length(txsig); % Calculate and % assign signal power rxsig = step(hChan,txsig); % Add noise. decodmsg = step(hDemod, rxsig); % Demodulate. if (berVec(2,jj)==0) % The first symbol of a differentially encoded transmission % is discarded. berVec(:,jj) = step(hErrorCalc, msg(2:end),decodmsg(2:end)); else berVec(:,jj) = step(hErrorCalc, msg, decodmsg); end end % Error rate and 98% confidence interval for this EbNo value [ber(jj), intv1] = berconfint(berVec(2,jj),berVec(3,jj)-1,.98); intv{jj} = intv1; % Store in cell array for later use. disp(['EbNo = ' num2str(EbNo) ' dB, ' num2str(berVec(2,jj)) ... ' errors, BER = ' num2str(ber(jj))]) end
This part of the example displays output in the Command Window
as it progresses through the for
loop. Your exact
output might be different, because this example uses random numbers.
EbNo = 0 dB, 189 errors, BER = 0.18919 EbNo = 1 dB, 139 errors, BER = 0.13914 EbNo = 2 dB, 105 errors, BER = 0.10511 EbNo = 3 dB, 66 errors, BER = 0.066066 EbNo = 4 dB, 40 errors, BER = 0.04004 EbNo = 5 dB, 18 errors, BER = 0.018018 EbNo = 6 dB, 6 errors, BER = 0.006006 EbNo = 7 dB, 11 errors, BER = 0.0055028 EbNo = 8 dB, 5 errors, BER = 0.00071439 EbNo = 9 dB, 5 errors, BER = 0.00022728 EbNo = 10 dB, 5 errors, BER = 1.006e-005
Plotting the Empirical Results and the Fitted Curve
The final part of this example fits a curve to the BER data
collected from the simulation loop. It also plots error bars using
the output from the berconfint
function.
% Use BERFIT to plot the best fitted curve, % interpolating to get a smooth plot. fitEbNo = EbNomin:0.25:EbNomax; % Interpolation values berfit(EbNovec,ber,fitEbNo,[],'exp'); % Also plot confidence intervals. hold on; for jj=1:numEbNos semilogy([EbNovec(jj) EbNovec(jj)],intv{jj},'g-+'); end hold off;
The command bertool
launches
the Bit Error Rate Analysis Tool (BERTool) application.
The application enables you to analyze the bit error rate (BER) performance of communications systems. BERTool computes the BER as a function of signal-to-noise ratio. It analyzes performance either with Monte-Carlo simulations of MATLAB functions and Simulink^{®} models or with theoretical closed-form expressions for selected types of communication systems.
Using BERTool you can:
Generate BER data for a communication system using
Closed-form expressions for theoretical BER performance of selected types of communication systems.
The semianalytic technique.
Simulations contained in MATLAB simulation functions or Simulink models. After you create a function or model that simulates the system, BERTool iterates over your choice of E_{b}/N_{0} values and collects the results.
Plot one or more BER data sets on a single set of axes. For example, you can graphically compare simulation data with theoretical results or simulation data from a series of similar models of a communication system.
Fit a curve to a set of simulation data.
Send BER data to the MATLAB workspace or to a file for any further processing you might want to perform.
Note: BERTool is designed for analyzing bit error rates only, not symbol error rates, word error rates, or other types of error rates. If, for example, your simulation computes a symbol error rate (SER), convert the SER to a BER before using the simulation with BERTool. |
The following sections describe the Bit Error Rate Analysis Tool (BERTool) and provide examples showing how to use its GUI.
To open BERTool, type
bertool
Components of BERTool
A data viewer at the top. It is initially empty.
After you instruct BERTool to generate one or more BER data sets, they appear in the data viewer. An example that shows how data sets look in the data viewer is in Example: Using a MATLAB Simulation with BERTool.
A set of tabs on the bottom. Labeled Theoretical, Semianalytic, and Monte Carlo, the tabs correspond to the different methods by which BERTool can generate BER data.
To learn more about each of the methods, see
A separate BER Figure window, which displays some or all of the BER data sets that are listed in the data viewer. BERTool opens the BER Figure window after it has at least one data set to display, so you do not see the BER Figure window when you first open BERTool. For an example of how the BER Figure window looks, see Example: Using the Theoretical Tab in BERTool.
Interaction Among BERTool Components. The components of BERTool act as one integrated tool. These behaviors reflect their integration:
If you select a data set in the data viewer, BERTool reconfigures the tabs to reflect the parameters associated with that data set and also highlights the corresponding data in the BER Figure window. This is useful if the data viewer displays multiple data sets and you want to recall the meaning and origin of each data set.
If you click data plotted in the BER Figure window, BERTool reconfigures the tabs to reflect the parameters associated with that data and also highlights the corresponding data set in the data viewer.
Note: You cannot click on a data point while BERTool is generating Monte Carlo simulation results. You must wait until the tool generates all data points before clicking for more information. |
If you configure the Semianalytic or Theoretical tab in a way that is already reflected in an existing data set, BERTool highlights that data set in the data viewer. This prevents BERTool from duplicating its computations and its entries in the data viewer, while still showing you the results that you requested.
If you close the BER Figure window, then you can reopen it by choosing BER Figure from the Window menu in BERTool.
If you select options in the data viewer that affect the BER plot, the BER Figure window reflects your selections immediately. Such options relate to data set names, confidence intervals, curve fitting, and the presence or absence of specific data sets in the BER plot.
Note: If you want to observe the integration yourself but do not yet have any data sets in BERTool, then first try the procedure in Example: Using the Theoretical Tab in BERTool. |
Note: If you save the BER Figure window using the window's File menu, the resulting file contains the contents of the window but not the BERTool data that led to the plot. To save an entire BERTool session, see Saving a BERTool Session. |
Section Overview. You can use BERTool to generate and analyze theoretical BER data. Theoretical data is useful for comparison with your simulation results. However, closed-form BER expressions exist only for certain kinds of communication systems.
To access the capabilities of BERTool related to theoretical BER data, use the following procedure:
Open BERTool, and go to the Theoretical tab.
Set the parameters to reflect the system whose performance you want to analyze. Some parameters are visible and active only when other parameters have specific values. See Available Sets of Theoretical BER Data for details.
Click Plot.
For an example that shows how to generate and analyze theoretical BER data via BERTool, see Example: Using the Theoretical Tab in BERTool.
Also, Available Sets of Theoretical BER Data indicates which combinations of parameters are available on the Theoretical tab and which underlying functions perform computations.
Example: Using the Theoretical Tab in BERTool. This example illustrates how to use BERTool to generate and plot theoretical BER data. In particular, the example compares the performance of a communication system that uses an AWGN channel and QAM modulation of different orders.
Running the Theoretical ExampleOpen BERTool, and go to the Theoretical tab.
Set the parameters as shown in the following figure.
Click Plot.
BERTool creates an entry in the data viewer and plots the data in the BER Figure window. Even though the parameters request that E_{b}/N_{0} go up to 18, BERTool plots only those BER values that are at least 10^{-8}. The following figures illustrate this step.
Change the Modulation order parameter
to 16
, and click Plot.
BERTool creates another entry in the data viewer and plots the new data in the same BER Figure window (not pictured).
Change the Modulation order parameter
to 64
, and click Plot.
BERTool creates another entry in the data viewer and plots the new data in the same BER Figure window, as shown in the following figures.
To recall which value of Modulation order corresponds to a given curve, click the curve. BERTool responds by adjusting the parameters in the Theoretical tab to reflect the values that correspond to that curve.
To remove the last curve from the plot (but not from the data viewer), clear the check box in the last entry of the data viewer in the Plot column. To restore the curve to the plot, select the check box again.
Available Sets of Theoretical BER Data. BERTool can generate a large set of theoretical bit-error rates, but not all combinations of parameters are currently supported. The Theoretical tab adjusts itself to your choices, so that the combination of parameters is always valid. You can set the Modulation order parameter by selecting a choice from the menu or by typing a value in the field. The Normalized timing error must be between 0 and 0.5.
BERTool assumes that Gray coding is used for all modulations.
For QAM, when $${\mathrm{log}}_{2}M$$ is odd (M being the modulation order), a rectangular constellation is assumed.
Combinations of Parameters for AWGN Channel SystemsThe following table lists the available sets of theoretical BER data for systems that use an AWGN channel.
Modulation | Modulation Order | Other Choices | |
---|---|---|---|
PSK | 2, 4 | Differential or nondifferential encoding. | |
8, 16, 32, 64, or a higher power of 2 | |||
OQPSK | 4 | Differential or nondifferential encoding. | |
DPSK | 2, 4, 8, 16, 32, 64, or a higher power of 2 | ||
PAM | 2, 4, 8, 16, 32, 64, or a higher power of 2 | ||
QAM | 4, 8, 16, 32, 64, 128, 256, 512, 1024, or a higher power of 2 | ||
FSK | 2 | Orthogonal or nonorthogonal; Coherent or Noncoherent demodulation. | |
4, 8, 16, 32, or a higher power of 2 | Orthogonal; Coherent demodulation. | ||
4, 8, 16, 32, or 64 | Orthogonal; Noncoherent demodulation. | ||
MSK | 2 | Coherent conventional
or precoded MSK; Noncoherent precoded MSK. | |
CPFSK | 2, 4, 8, 16, or a higher power of 2 | Modulation
index > 0. |
BER results are also available for the following:
block and convolutional coding with hard-decision decoding for all modulations except CPFSK
block coding with soft-decision decoding for all binary modulations (including 4-PSK and 4-QAM) except CPFSK, noncoherent non-orthogonal FSK, and noncoherent MSK
convolutional coding with soft-decision decoding for all binary modulations (including 4-PSK and 4-QAM) except CPFSK
uncoded nondifferentially-encoded 2-PSK with synchronization errors
For more information about specific combinations of parameters, including bibliographic references that contain closed-form expressions, see the reference pages for the following functions:
berawgn
—
For systems with no coding and perfect synchronization
bercoding
—
For systems with channel coding
bersync
—
For systems with BPSK modulation, no coding, and imperfect synchronization
The following table lists the available sets of theoretical BER data for systems that use a Rayleigh or Rician channel.
When diversity is used, the SNR on each diversity branch is
derived from the SNR at the input of the channel (EbNo
)
divided by the diversity order.
Modulation | Modulation Order | Other Choices |
---|---|---|
PSK | 2 | Differential or nondifferential encoding
In the case of nondifferential encoding, diversity order being 1, and Rician fading, a value for RMS phase noise (in radians) can be specified. |
4, 8, 16, 32, 64, or a higher power of 2 | Diversity
order ≧1 | |
OQPSK | 4 | Diversity order ≧1 |
DPSK | 2, 4, 8, 16, 32, 64, or a higher power of 2 | Diversity
order ≧1 |
PAM | 2, 4, 8, 16, 32, 64, or a higher power of 2 | Diversity order ≧1 |
QAM | 4, 8, 16, 32, 64, 128, 256, 512, 1024, or a higher power of 2 | Diversity order ≧1 |
FSK | 2 | Correlation coefficient $$\in [-1,1]$$.
In the case of a nonzero correlation coefficient and noncoherent demodulation, the diversity order is 1 only. |
4, 8, 16, 32, or a higher power of 2 | Noncoherent demodulation
only. Diversity order ≧1 |
For more information about specific combinations of parameters,
including bibliographic references that contain closed-form expressions,
see the reference page for the berfading
function.
Section Overview. You can use BERTool to generate and analyze BER data via the semianalytic technique. The semianalytic technique is discussed in Performance Results via the Semianalytic Technique, and When to Use the Semianalytic Technique is particularly relevant as background material.
To access the semianalytic capabilities of BERTool, open the Semianalytic tab.
For further details about how BERTool applies the semianalytic
technique, see the reference page for the semianalytic
function,
which BERTool uses to perform computations.
Example: Using the Semianalytic Tab in BERTool. This example illustrates how BERTool applies the semianalytic
technique, using 16-QAM modulation. This example is a variation on
the example in Example: Using the Semianalytic Technique, but it is tailored
to use BERTool instead of using the semianalytic
function
directly.
To set up the transmitted and received signals, run steps 1 through 4 from the code example in Example: Using the Semianalytic Technique. The code is repeated below.
% Step 1. Generate message signal of length >= M^L. M = 16; % Alphabet size of modulation L = 1; % Length of impulse response of channel msg = [0:M-1 0]; % M-ary message sequence of length > M^L % Step 2. Modulate the message signal using baseband modulation. hMod = comm.RectangularQAMModulator(M); % Use 16-QAM. modsig = step(hMod,msg'); % Modulate data Nsamp = 16; modsig = rectpulse(modsig,Nsamp); % Use rectangular pulse shaping. % Step 3. Apply a transmit filter. txsig = modsig; % No filter in this example % Step 4. Run txsig through a noiseless channel. rxsig = txsig*exp(1i*pi/180); % Static phase offset of 1 degree
Open BERTool and go to the Semianalytic tab.
Set parameters as shown in the following figure.
Click Plot.
After you click Plot, BERTool creates a listing for the resulting data in the data viewer.
BERTool plots the data in the BER Figure window.
Procedure for Using the Semianalytic Tab in BERTool. The procedure below describes how you typically implement the semianalytic technique using BERTool:
Generate a message signal containing at least M^{L} symbols,
where M is the alphabet size of the modulation and L is the length
of the impulse response of the channel in symbols. A common approach
is to start with an augmented binary pseudonoise (PN) sequence of
total length (log_{2}M)M^{L}
.
An augmented PN sequence is a PN sequence with
an extra zero appended, which makes the distribution of ones and zeros
equal.
Modulate a carrier with the message signal using baseband modulation.
Supported modulation types are listed on the reference page for semianalytic
.
Shape the resultant signal with rectangular pulse shaping, using
the oversampling factor that you will later use to filter the modulated
signal. Store the result of this step as txsig
for
later use.
Filter the modulated signal with a transmit filter. This filter is often a square-root raised cosine filter, but you can also use a Butterworth, Bessel, Chebyshev type 1 or 2, elliptic, or more general FIR or IIR filter. If you use a square-root raised cosine filter, use it on the nonoversampled modulated signal and specify the oversampling factor in the filtering function. If you use another filter type, you can apply it to the rectangularly pulse shaped signal.
Run the filtered signal through a noiseless channel.
This channel can include multipath fading effects, phase shifts,
amplifier nonlinearities, quantization, and additional filtering,
but it must not include noise. Store the result of this step as rxsig
for
later use.
On the Semianalytic tab of BERTool, enter parameters as in the table below.
Parameter Name | Meaning |
---|---|
Eb/No range | A vector that lists the values of E_{b}/N_{0} for which you want to collect BER data. The value in this field can be a MATLAB expression or the name of a variable in the MATLAB workspace. |
Modulation type | These parameters describe the modulation scheme you used earlier in this procedure. |
Modulation order | |
Differential encoding | This check box, which is visible and active for MSK and PSK modulation, enables you to choose between differential and nondifferential encoding. |
Samples per symbol | The number of samples per symbol in the transmitted signal. This value is also the sampling rate of the transmitted and received signals, in Hz. |
Transmitted signal | The txsig signal
that you generated earlier in this procedure |
Received signal | The rxsig signal
that you generated earlier in this procedure |
Numerator | Coefficients of the receiver filter that BERTool applies to the received signal |
Denominator |
Note:
Consistency among the values in the GUI is important. For example,
if the signal referenced in the Transmitted signal field
was generated using DPSK and you set Modulation type to |
Click Plot.
After you click Plot, BERTool performs these tasks:
Filters rxsig
and then determines
the error probability of each received signal point by analytically
applying the Gaussian noise distribution to each point. BERTool averages
the error probabilities over the entire received signal to determine
the overall error probability. If the error probability calculated
in this way is a symbol error probability, BERTool converts it to
a bit error rate, typically by assuming Gray coding. (If the modulation
type is DQPSK or cross QAM, the result is an upper bound on the bit
error rate rather than the bit error rate itself.)
Enters the resulting BER data in the data viewer of the BERTool window.
Plots the resulting BER data in the BER Figure window.
Section Overview. You can use BERTool in conjunction with your own MATLAB simulation functions to generate and analyze BER data. The MATLAB function simulates the communication system whose performance you want to study. BERTool invokes the simulation for E_{b}/N_{0} values that you specify, collects the BER data from the simulation, and creates a plot. BERTool also enables you to easily change the E_{b}/N_{0} range and stopping criteria for the simulation.
To learn how to make your own simulation functions compatible with BERTool, see Use Simulation Functions with BERTool.
Example: Using a MATLAB Simulation with BERTool. This example illustrates how BERTool can run a MATLAB simulation
function. The function is viterbisim
, one of the
demonstration files included with Communications System Toolbox software.
To run this example, follow these steps:
Open BERTool and go to the Monte Carlo tab. (The default parameters depend on whether you have Communications System Toolbox software installed. Also note that the BER variable name field applies only to Simulink models.)
Set parameters as shown in the following figure.
Click Run.
BERTool runs the simulation function once for each specified value of E_{b}/N_{0} and gathers BER data. (While BERTool is busy with this task, it cannot process certain other tasks, including plotting data from the other tabs of the GUI.)
Then BERTool creates a listing in the data viewer.
BERTool plots the data in the BER Figure window.
To change the range of E_{b}/N_{0} while reducing the number
of bits processed in each case, type [5 5.2 5.3]
in
the Eb/No range field, type 1e5
in
the Number of bits field, and click Run.
BERTool runs the simulation function again for each new value of E_{b}/N_{0} and gathers new BER data. Then BERTool creates another listing in the data viewer.
BERTool plots the data in the BER Figure window, adjusting the horizontal axis to accommodate the new data.
The two points corresponding to 5 dB from the two data sets are different because the smaller value of Number of bits in the second simulation caused the simulation to end before observing many errors. To learn more about the criteria that BERTool uses for ending simulations, see Varying the Stopping Criteria.
For another example that uses BERTool to run a MATLAB simulation function, see Example: Prepare a Simulation Function for Use with BERTool.
Varying the Stopping Criteria. When you create a MATLAB simulation function for use with BERTool, you must control the flow so that the simulation ends when it either detects a target number of errors or processes a maximum number of bits, whichever occurs first. To learn more about this requirement, see Requirements for Functions; for an example, see Example: Prepare a Simulation Function for Use with BERTool.
After creating your function, set the target number of errors and the maximum number of bits in the Monte Carlo tab of BERTool.
Typically, a Number of errors value of
at least 100
produces an accurate error rate. The Number
of bits value prevents the simulation from running too
long, especially at large values of E_{b}/N_{0}. However, if the Number
of bits value is so small that the simulation collects
very few errors, the error rate might not be accurate. You can use
confidence intervals to gauge the accuracy of the error rates that
your simulation produces; the larger the confidence interval, the
less accurate the computed error rate.
As an example, follow the procedure described in Example: Using a MATLAB Simulation with BERTool and set Confidence
Level to 95
for each of the
two data sets. The confidence intervals for the second data set are
larger than those for the first data set. This is because the second
data set uses a small value for Number of bits relative
to the communication system properties and the values in Eb/No
range, resulting in BER values based on only a small number
of observed errors.
Note: You can also use the Stop button in BERTool to stop a series of simulations prematurely, as long as your function is set up to detect and react to the button press. |
Plotting Confidence Intervals. After you run a simulation with BERTool, the resulting data
set in the data viewer has an active menu in the Confidence
Level column. The default value is off
,
so that the simulation data in the BER Figure window does not show confidence
intervals.
To show confidence intervals in the BER Figure window, set Confidence
Level to a numerical value: 90%
, 95%
,
or 99%
.
The plot in the BER Figure window responds immediately to your choice. A sample plot is below.
For an example that plots confidence intervals for a Simulink simulation, see Example: Using a Simulink Model with BERTool.
To find confidence intervals for levels not listed in the Confidence
Level menu, use the berconfint
function.
Fitting BER Points to a Curve. After you run a simulation with BERTool, the BER Figure window plots individual BER data points. To fit a curve to a data set that contains at least four points, select the box in the Fit column of the data viewer.
The plot in the BER Figure window responds immediately to your choice. A sample plot is below.
For an example that performs curve fitting for data from a Simulink simulation and generates the plot shown above, see Example: Using a Simulink Model with BERTool. For an example that performs curve fitting for data from a MATLAB simulation function, see Example: Prepare a Simulation Function for Use with BERTool.
For greater flexibility in the process of fitting a curve to
BER data, use the berfit
function.
Requirements for Functions. When you create a MATLAB function for use with BERTool, ensure the function interacts properly with the GUI. This section describes the inputs, outputs, and basic operation of a BERTool-compatible function.
Input ArgumentsBERTool evaluates your entries in fields of the GUI and passes data to the function as these input arguments, in sequence:
One value from the Eb/No range vector each time BERTool invokes the simulation function
The Number of errors value
The Number of bits value
Your simulation function must compute and return these output arguments, in sequence:
Bit error rate of the simulation
Number of bits processed when computing the BER
BERTool uses these output arguments when reporting and plotting results.
Simulation OperationYour simulation function must perform these tasks:
Simulate the communication system for the E_{b}/N_{0} value specified in the first input argument.
Stop simulating when the number of errors or the number of processed bits equals or exceeds the corresponding threshold specified in the second or third input argument, respectively.
Detect whether you click Stop in BERTool and abort the simulation in that case.
Template for a Simulation Function. Use the following template when adapting your code to work with BERTool.
You can open it in an editor by entering edit bertooltemplate
in
the MATLAB Command Window. The description in Understanding the Template explains
the template's key sections, while Using the Template indicates how to use the template with
your own simulation code. Alternatively, you can develop your simulation
function without using the template, but be sure it satisfies the
requirements described in Requirements for Functions.
Note:
The template is not yet ready for use with BERTool. You must
insert your own simulation code in the places marked |
function [ber, numBits] = bertooltemplate(EbNo, maxNumErrs, maxNumBits) % Import Java class for BERTool. import com.mathworks.toolbox.comm.BERTool; % Initialize variables related to exit criteria. berVec = zeros(3,1); % Updated BER values % --- Set up parameters. --- % --- INSERT YOUR CODE HERE. % Simulate until number of errors exceeds maxNumErrs % or number of bits processed exceeds maxNumBits. while((berVec(2) < maxNumErrs) && (berVec(3) < maxNumBits)) % Check if the user clicked the Stop button of BERTool. if (BERTool.getSimulationStop) break; end % --- Proceed with simulation. % --- Be sure to update totErr and numBits. % --- INSERT YOUR CODE HERE. end % End of loop % Assign values to the output variables. ber = berVec(1); numBits = berVec(3);
From studying the code in the function template, observe how the function either satisfies the requirements listed in Requirements for Functions or indicates where your own insertions of code should do so. In particular,
The function has appropriate input and output arguments.
The function includes a placeholder for code that simulates a system for the given E_{b}/N_{0} value.
The function uses a loop structure to stop simulating
when the number of errors exceeds maxNumErrs
or
the number of bits exceeds maxNumBits
, whichever
occurs first.
Note:
Although the |
In each iteration of the loop, the function detects when the user clicks Stop in BERTool.
Here is a procedure for using the template with your own simulation code:
Determine the setup tasks you must perform. For example,
you might want to initialize variables containing the modulation alphabet
size, filter coefficients, a convolutional coding trellis, or the
states of a convolutional interleaver. Place the code for these setup
tasks in the template section marked Set up parameters
.
Determine the core simulation tasks, assuming that
all setup work has already been performed. For example, these tasks
might include error-control coding, modulation/demodulation, and channel
modeling. Place the code for these core simulation tasks in the template
section marked Proceed with simulation
.
Also in the template section marked Proceed
with simulation
, include code that updates the values of totErr
and numBits
.
The quantity totErr
represents the number of errors
observed so far. The quantity numBits
represents
the number of bits processed so far. The computations to update these
variables depend on how your core simulation tasks work.
Note: Updating the numbers of errors and bits is important for ensuring that the loop terminates. However, if you accidentally create an infinite loop early in your development work using the function template, click Stop in BERTool to abort the simulation. |
Omit any setup code that initializes EbNo
, maxNumErrs
,
or maxNumBits
, because BERTool passes these quantities
to the function as input arguments after evaluating the data entered
in the GUI.
Adjust your code or the template's code as necessary
to use consistent variable names and meanings. For example, if your
original code uses a variable called ebn0
and the
template's function declaration (first line) uses the variable name EbNo
,
you must change one of the names so they match. As another example,
if your original code uses SNR instead of E_{b}/N_{0}, you must convert
quantities appropriately.
Example: Prepare a Simulation Function for Use with BERTool. This section adapts the function template given in Template for a Simulation Function.
Preparing the FunctionTo prepare the function for use with BERTool, follow these steps:
Copy the template from Template for a Simulation Function into a new MATLAB file
in the MATLAB Editor. Save it in a folder on your MATLAB path
using the file name bertool_simfcn
.
From the original example, the following lines are
setup tasks. They are modified from the original example to rely on
the input arguments that BERTool provides to the function, instead
of defining variables such as EbNovec
and numerrmin
directly.
% Set up initial parameters. siglen = 1000; % Number of bits in each trial M = 2; % DBPSK is binary. % DBPSK modulation and demodulation System objects hMod = comm.DBPSKModulator; hDemod = comm.DBPSKDemodulator; % AWGNChannel System object hChan = comm.AWGNChannel('NoiseMethod', 'Signal to noise ratio (SNR)'); % ErrorRate calculator System object to compare decoded symbols to the % original transmitted symbols. hErrorCalc = comm.ErrorRate; snr = EbNo; % Because of binary modulation hChan.SNR = snr; %Assign Channel SNR
Place these lines of code in the template section marked Set
up parameters
.
From the original example, the following lines are the core simulation tasks, after all setup work has been performed.
msg = randi([0,M-1], siglen, 1); % Generate message sequence. txsig = step(hMod, msg); % Modulate. hChan.SignalPower = (txsig'*txsig)/length(txsig); % Calculate and % assign signal power rxsig = step(hChan,txsig); % Add noise. decodmsg = step(hDemod, rxsig); % Demodulate. berVec = step(hErrorCalc, msg, decodmsg); % Calculate BER
Place the code for these core simulation tasks in the template
section marked Proceed with simulation
.
The bertool_simfcn
function is now compatible
with BERTool. Note that unlike the original example, the function
here does not initialize EbNovec
,
define EbNo
as a scalar, or use numerrmin
as
the target number of errors; this is because BERTool provides input
arguments for similar quantities. The bertool_simfcn
function
also excludes code related to plotting, curve fitting, and confidence
intervals in the original example because BERTool enables you to
do similar tasks interactively without writing code.
To use bertool_simfcn
in conjunction with BERTool,
continue the example by following these steps:
Open BERTool and go to the Monte Carlo tab.
Set parameters on the Monte Carlo tab as shown in the following figure.
Click Run.
BERTool spends some time computing results and then plots
them. They do not appear to fall along a smooth curve because the
simulation required only five errors for each value in EbNo
.
To fit a curve to the series of points in the BER Figure window, select the box next to Fit in the data viewer.
BERTool plots the curve, as shown in the following figure.
Section Overview. You can use BERTool in conjunction with Simulink models to generate and analyze BER data. The Simulink model simulates the communication system whose performance you want to study, while BERTool manages a series of simulations using the model and collects the BER data.
Note: To use Simulink models within BERTool, you must have a Simulink license. Communications System Toolbox software is highly recommended. The rest of this section assumes you have a license for both Simulink and Communications System Toolbox applications. |
To access the capabilities of BERTool related to Simulink models, open the Monte Carlo tab.
For further details about confidence intervals and curve fitting for simulation data, see Plotting Confidence Intervals and Fitting BER Points to a Curve, respectively.
Example: Using a Simulink Model with BERTool. This example illustrates how BERTool can manage a series of
simulations of a Simulink model, and how you can vary the plot.
The model is commgraycode
, one of the demonstration
models included with Communications System Toolbox software. The example
assumes that you have Communications System Toolbox software installed.
To run this example, follow these steps:
Open BERTool and go to the Monte Carlo tab.
The model's file name, commgraycode.mdl
, appears
as the Simulation MATLAB file or Simulink model parameter.
(If viterbisim.m
appears there, select to indicate
that Communications System Toolbox software is installed.)
Click Run.
BERTool loads the model into memory (which in turn initializes several variables in the MATLAB workspace), runs the simulation once for each value of E_{b}/N_{0}, and gathers BER data. BERTool creates a listing in the data viewer.
BERTool plots the data in the BER Figure window.
To fit a curve to the series of points in the BER Figure window, select the box next to Fit in the data viewer.
BERTool plots the curve, as below.
To indicate the 99% confidence interval around each
point in the simulation data, set Confidence Level to 99%
in
the data viewer.
BERTool displays error bars to represent the confidence intervals, as below.
Another example that uses BERTool to manage a series of Simulink simulations is in Example: Prepare a Model for Use with BERTool.
Varying the Stopping Criteria. When you create a Simulink model for use with BERTool, you must set it up so that the simulation ends when it either detects a target number of errors or processes a maximum number of bits, whichever occurs first. To learn more about this requirement, see Requirements for Models; for an example, see Example: Prepare a Model for Use with BERTool.
After creating your Simulink model, set the target number of errors and the maximum number of bits in the Monte Carlo tab of BERTool.
Typically, a Number of errors value of
at least 100
produces an accurate error rate. The Number
of bits value prevents the simulation from running too
long, especially at large values of E_{b}/N_{0}. However, if the Number
of bits value is so small that the simulation collects
very few errors, the error rate might not be accurate. You can use
confidence intervals to gauge the accuracy of the error rates that
your simulation produces; the larger the confidence interval, the
less accurate the computed error rate.
You can also click Stop in BERTool to stop a series of simulations prematurely.
Requirements for Models. A Simulink model must satisfy these requirements before you can use it with BERTool, where the case-sensitive variable names must be exactly as shown below:
The channel block must use the variable EbNo
rather
than a hard-coded value for E_{b}/N_{0}.
The simulation must stop when the error count reaches
the value of the variable maxNumErrs
or when the
number of processed bits reaches the value of the variable maxNumBits
,
whichever occurs first.
You can configure the Error Rate Calculation block in Communications System Toolbox software to stop the simulation based on such criteria.
The simulation must send the final error rate data to the MATLAB workspace as a variable whose name you enter in the BER variable name field in BERTool. The variable must be a three-element vector that lists the BER, the number of bit errors, and the number of processed bits.
This three-element vector format is supported by the Error Rate Calculation block.
Tips for Preparing Models. Here are some tips for preparing a Simulink model for use with BERTool:
To avoid using an undefined variable name in the dialog box for a Simulink block in the steps that follow, set up variables in the MATLAB workspace using a command such as the one below.
EbNo = 0; maxNumErrs = 100; maxNumBits = 1e8;
You might also want to put the same command in the model's preload function callback, to initialize the variables if you reopen the model in a future MATLAB session.
When you use BERTool, it provides the actual values based on what you enter in the GUI, so the initial values above are somewhat arbitrary.
To model the channel, use the AWGN Channel block in Communications System Toolbox software with these parameters:
Mode = Signal to
noise ratio (Eb/No)
Eb/No = EbNo
To compute the error rate, use the Error Rate Calculation block in Communications System Toolbox software with these parameters:
Check Stop simulation.
Target number of errors = maxNumErrs
Maximum number of symbols = maxNumBits
To send data from the Error Rate Calculation block
to the MATLAB workspace, set Output data to Port
,
attach a Signal to Workspace block from DSP System Toolbox™ software,
and set the latter block's Limit data points to last parameter
to 1
. The Variable name parameter
in the Signal to Workspace block must match the value you enter in
the BER variable name field of BERTool.
If your model computes a symbol error rate instead of a bit error rate, use the Integer to Bit Converter block in Communications System Toolbox software to convert symbols to bits.
Frame-based simulations often run faster than sample-based simulations for the same number of bits processed. The number of errors or number of processed bits might exceed the values you enter in BERTool, because the simulation always processes a fixed amount of data in each frame.
If you have an existing model that uses the AWGN Channel
block using a Mode parameter other than Signal
to noise ratio (Eb/No)
, you can adapt the block to use
the Eb/No mode instead. To learn about how the block's different modes
are related to each other, press the AWGN Channel block's Help button
to view the online reference page.
If your model uses a preload function or other callback to initialize variables in the MATLAB workspace upon loading, make sure before you use the Run button in BERTool that one of these conditions is met:
The model is not currently in memory. In this case, BERTool loads the model into memory and runs the callback functions.
The model is in memory (whether in a window or not), and the variables are intact.
If you clear or overwrite the model's variables and want to
restore their values before using the Run button
in BERTool, you can use the bdclose
function
in the MATLAB Command Window to clear the model from memory.
This causes BERTool to reload the model after you click Run.
Similarly, if you refresh your workspace by issuing a clear
all
or clear variables
command, you should
also clear the model from memory by using bdclose all
.
Example: Prepare a Model for Use with BERTool. This example uses a Simulink model that is set up for use with BERTool. The example also illustrates how to compare the BER performance of a Simulink simulation with theoretical BER results. The example assumes that you have Communications System Toolbox software installed.
To prepare the model for use with BERTool, follow these steps, using the exact case-sensitive variable names as shown:
Open the model by entering the following command in the MATLAB Command Window.
doc_bpsk
To initialize parameters in the MATLAB workspace and avoid using undefined variables as block parameters, enter the following command in the MATLAB Command Window.
EbNo = 0; maxNumErrs = 100; maxNumBits = 1e8;
To ensure that BERTool uses the correct amount of
noise each time it runs the simulation, open the dialog box for the AWGN
Channel block by double-clicking the block. Verify that Es/No is
set to EbNo
and click OK.
In this particular model, E_{s}/N_{0} is
equivalent to E_{b}/N_{0} because the modulation type is BPSK.
To ensure that BERTool uses the correct stopping
criteria for each iteration, open the dialog box for the Error
Rate Calculation block. Verify that Target number
of errors is set to maxNumErrs
, and
that Maximum number of symbols is set to maxNumBits
.
Click OK.
To enable BERTool to access the BER results that
the Error Rate Calculation block computes, the Signal to Workspace block, BER
,
is connected to the output of the Error Rate Calculation block.
Note: The Signal to Workspace block is in DSP System Toolbox software and is different from the To Workspace block in Simulink. |
To use the doc_bpsk
model with BERTool,
follow these steps:
Open BERTool and go to the Monte Carlo tab.
Set parameters on the Monte Carlo tab as shown in the following figure.
Click Run.
BERTool spends some time computing results and then plots them.
To compare these simulation results with theoretical results, go to the Theoretical tab in BERTool and set parameters as shown below.
Click Plot.
BERTool plots the theoretical curve in the BER Figure window along with the earlier simulation results.
Exporting Data Sets or BERTool Sessions. BERTool enables you to export individual data sets to the MATLAB workspace or to MAT-files. One option for exporting is convenient for processing the data outside BERTool. For example, to create a highly customized plot using data from BERTool, export the BERTool data set to the MATLAB workspace and use any of the plotting commands in MATLAB. Another option for exporting enables you to reimport the data into BERTool later.
BERTool also enables you to save an entire session, which is useful if your session contains multiple data sets that you want to return to in a later session.
This section describes these capabilities:
Exporting Data SetsTo export an individual data set, follow these steps:
In the data viewer, select the data set you want to export.
Choose File > Export Data.
Set Export to to indicate the format and destination of the data.
If you want to reimport the data into BERTool later,
you must choose either Workspace
structure
or MAT-file structure
to
create a structure in the MATLAB workspace or a MAT-file, respectively.
A new field called Structure name appears. Set it to the name that you want BERTool to use for the structure it creates.
If you selected Workspace structure
and
you want BERTool to use your chosen variable name, even if a variable
by that name already exists in the workspace, select Overwrite
variables.
If you do not need to reimport
the data into BERTool later, a convenient way to access the data
outside BERTool is to have BERTool create a pair of arrays in
the MATLAB workspace. One array contains E_{b}/N_{0} values, while
the other array contains BER values. To choose this option, set Export
to to Workspace arrays
.
Then type two variable names in the fields under Variable names.
If you want BERTool to use your chosen variable names even if variables by those names already exist in the workspace, select Overwrite variables.
Click OK. If you selected MAT-file
structure
, BERTool prompts you for the path to the
MAT-file that you want to create.
To reimport a structure later, see Importing Data Sets.
Examining an Exported StructureThis section briefly describes the contents of the structure
that BERTool exports to the workspace or to a MAT-file. The structure's
fields are indicated in the table below. The fields that are most
relevant for you when you want to manipulate exported data are paramsEvaled
and data
.
Name of Field | Significance |
---|---|
params | The parameter values in the BERTool GUI, some of which might be invisible and hence irrelevant for computations. |
paramsEvaled | The parameter values that BERTool uses when computing the data set. |
data | The E_{b}/N_{0}, BER, and number of bits processed. |
dataView | Information about the appearance in the data viewer. Used by BERTool for data reimport. |
cellEditabilities | Indicates whether the data viewer has an active Confidence Level or Fit entry. Used by BERTool for data reimport. |
The params
and paramsEvaled
fields
are similar to each other, except that params
describes
the exact state of the GUI whereas paramsEvaled
indicates
the values that are actually used for computations. As an example
of the difference, for a theoretical system with an AWGN channel, params
records
but paramsEvaled
omits a diversity order parameter.
The diversity order is not used in the computations because it is
relevant only for systems with Rayleigh channels. As another example,
if you type [0:3]+1
in the GUI as the range of E_{b}/N_{0} values, params
indicates [0:3]+1
while paramsEvaled
indicates 1
2 3 4
.
The length and exact contents of paramsEvaled
depend
on the data set because only relevant information appears. If the
meaning of the contents of paramsEvaled
is not
clear upon inspection, one way to learn more is to reimport the data
set into BERTool and inspect the parameter values that appear in
the GUI. To reimport the structure, follow the instructions in Importing Data Sets or BERTool Sessions.
If your exported workspace variable is called ber0
,
the field ber0.data
is a cell array that contains
the numerical results in these vectors:
ber0.data{1}
lists the E_{b}/N_{0} values.
ber0.data{2}
lists the BER values
corresponding to each of the E_{b}/N_{0} values.
ber0.data{3}
indicates, for simulation
or semianalytic results, how many bits BERTool processed when computing
each of the corresponding BER values.
To save an entire BERTool session, follow these steps:
Choose File > Save Session.
When BERTool prompts you, enter the path to the file that you want to create.
BERTool creates a text file that records all data sets currently in the data viewer, along with the GUI parameters associated with the data sets.
Note:
If your BERTool session requires particular workspace variables
(such as |
Importing Data Sets or BERTool Sessions. BERTool enables you to reimport individual data sets that you previously exported to a structure, or to reload entire sessions that you previously saved. This section describes these capabilities:
To learn more about exporting data sets or saving sessions from BERTool, see Exporting Data Sets or BERTool Sessions.
Importing Data SetsTo import an individual data set that you previously exported from BERTool to a structure, follow these steps:
Choose File > Import Data.
Set Import from to either Workspace
structure
or MAT-file structure
.
If you select Workspace structure
, type
the name of the workspace variable in the Structure name field.
Click OK. If you select MAT-file
, BERTool prompts
you to select the file that contains the structure you want to import.
After you dismiss the Data Import dialog box (and the file selection dialog box, in the case of a MAT-file), the data viewer shows the newly imported data set and the BER Figure window plots it.
Opening a Previous BERTool SessionTo replace the data sets in the data viewer with data sets from a previous BERTool session, follow these steps:
Choose File > Open Session.
Note: If BERTool already contains data sets, it asks you whether you want to save the current session. If you answer no and continue with the loading process, BERTool discards the current session upon opening the new session from the file. |
When BERTool prompts you, enter the path to the file you want to open. It must be a file that you previously created using the Save Session option in BERTool.
After BERTool reads the session file, the data viewer shows the data sets from the file.
If your BERTool session requires particular workspace variables
(such as txsig
or rxsig
for
the Semianalytic tab) that you saved separately
in a MAT-file, you can retrieve them using the load
command
in MATLAB.
Managing Data in the Data Viewer. The data viewer gives you flexibility to rename and delete data sets, and to reorder columns in the data viewer.
To rename a data set in the data viewer, double-click its name in the BER Data Set column and type a new name.
To delete a data set from the data viewer, select it and choose Edit > Delete.
Note: If the data set originated from the Semianalytic or Theoretical tab, BERTool deletes the data without asking for confirmation. You cannot undo this operation. |
To move a column in the data viewer, drag the column's heading to the left or right with the mouse. For example, the image below shows the mouse dragging the BER column to the left of its default position. When you release the mouse button, the columns snap into place.
The Error Rate Test Console is an object capable of running simulations for communications systems to measure error rate performance.
The Error Rate Test Console is compatible with communications systems created with a specific API defined by the testconsole.SystemBasicAPI class. Within this class definition you define the functionality of a communications system.
You attach a system to the Error Rate Test Console to run simulations and obtain error rate data.
You obtain error rate results at different locations in the system under test, by defining unique test points. Each test point contains a pair of probes that the system uses to log data to the test console. The information you register with the test console specifies how each pair of test probes compares data. For example, in a frame based system, the Error Rate Test Console can compare transmitted and received header bits or transmitted and received data bits. Similarly, it can compare CRC error counts to obtain frame error rates at different points in the system. You can also configure the Error Rate Test Console to compare data in multiple pairs of probes, obtaining multiple error rate results.
You can run simulations with as many test parameters as desired, parse the results, and obtain parametric or surface plots by specifying which parameters act as independent variables.
There are two main tasks associated with using the Error Rate Test Console: Creating a System and Attaching a System to the Error Rate Test Console.
When you run a system that is not attached to an Error Rate Test Console, the system is running in debug mode. Debug mode is useful when evaluating or debugging the code for the system you are designing.
To see a full-scale example on creating a system and running simulations, see Bit Error Rate Simulations For Various Eb/No and Modulation Order Values.
The following sections describe the Error Rate Test Console and its functionality:
Methods Allowing You to Communicate with the Error Rate Test Console at Simulation Run Time
Bit Error Rate Simulations For Various Eb/No and Modulation Order Values
You attach a system to the Error Rate Test Console to run simulations and obtain error rate data. When you attach the system under test, you also register specific information to the test console in order to define the system's test inputs, test parameters, and test probes.
Creating a communications system for use with the Error Rate Test Console, involves the following steps.
Writing a system class, extending the testconsole.SystemBasicAPI class.
Writing a registration method
Registration is test related
Defines items such as test parameters, test probes, and test inputs
Writing a setup method
Writing a reset method
Writing a run method
Methods allows the system to communicate with the test console.
To see the system file, navigate to the following location:
matlab\toolbox\comm\comm\+commtest
Then, enter the following syntax at the MATLAB command line:
edit MPSKSYSTEM.m
Writing A Register Method. Using the register
method, you register test
inputs, test parameters, and test probes to the Error Rate Test Console.
You register these items to the Error Rate Test Console using the registerTestInput
, registerTestParameter
,
and registerTestProbe
methods.
Write a register
method for every
communication system you create.
If you do not implement a register
method
for a system, you can still attach the system to the Error Rate Test
Console. While the test console runs the specified number of iterations
on the system, you cannot control simulation parameters or retrieve
results from the simulation.
In order to run simulations, the system under test requests test inputs from the Error Rate Test Console. These test inputs provide data, driving simulations for the system under test.
A system under test cannot request a specific input type until you attach it to the Error Rate Test Console. Additionally, the specific input type must be registered to the test console.
Inside the register
method, you call the registerTestInput(sys,inputName)
method
to register test inputs.
sys
represents the handle to a
user-defined system object.
inputName
represents the name of
the input that the system registers. This name must coincide with
the name of an available test input in the Error Rate Test Console
or an error occurs.
'NumTransmissions' - calling the getInput
method
returns the frame length. The system itself is responsible for generating
a data frame using a data source.
'RandomIntegerSource' - calling the getInput
method
returns a vector of symbols with a length the Error Rate Test Console FrameLength
property
specifies. If the system registers this source type, then it must
also register a test parameter named 'M' that corresponds to the modulation
order.
Test parameters are the system parameters for which the Error Rate Test Console obtains simulation results. You specify the sweep range of these parameters using the Error Rate Test Console and obtain simulation results for different system conditions.
The system under test registers a system parameter to the Error
Rate Test Console, creating a test parameter. You register a test
parameter to the Error Rate Test Console using the registerTestParameter(sys,name,default,validRange)
method.
sys
represents the handle to the
user-defined system object
name
represents the parameter name
that the system registers to the Error Rate Test Console
default
specifies the default value
of the test parameter – it can be a numeric value or a character
vector
validRange
specifies a range of
input values for the test parameter — it can be a 1x2 vector
of numeric values with upper and lower ranges or a cell array of chars
(an Enum).
A parameter of type char becomes useful when defining system conditions. For example, in a communications system, a Channel parameter may be defined so that it takes values such as ‘Rayleigh', ‘Rician', or ‘AWGN'. Depending on the Channel char value, the system may filter transmitted data through a different channel. This allows the simulation of the system over different channel scenarios.
If the system registers a test parameter named ‘X'
then the system must also contain a readable property named ‘X'.
If not, the registration process issues an error. This process ensures
that calling the getTestParameter
method in debug
mode returns the value held by the corresponding property.
Test probes log the simulation data the Error Rate Test Console uses for computing test metrics, such as: number of errors, number of transmissions, and error rate. To log data into a probe, your communications system must register the probe to the Error Rate Test Console.
You register a test probe to the Error Rate Test Console using
the registerTestProbe(sys,name,description)
method.
sys
represents the handle to the
user-defined system object
name
represents the name of the
test probe
description
contains information
about the test probes; useful for indicating what the probe is used
for. The description input is optional.
You can define an arbitrary number of probes to log test data at several points within the system.
Writing a Setup Method. The Error Rate Test Console calls the setup
method
at the beginning of simulations for each new sweep point. A sweep
point is one of several sets of simulation parameters for which the
system will be simulated. Using the getTestParameter
method
of the system under test, the setup
method requests
the current simulation sweep values from the Error Rate Test Console
and sets the various system components accordingly. The setup
method
sets the system to the conditions the current test parameter sweep
values generate.
Writing a setup
method for each communication
system you create is not necessary. The setup
method
is optional.
Writing a Reset Method. Use the reset
method to reset states of various
system components, such as: objects, data buffers, or system flags.
The Error Rate Test Console calls the reset
method
of the system:
at the beginning of simulations for a new sweep point.
(This condition occurs when you set the ResetMode
of
the Error Rate Test Console to "Reset at new simulation point'.)
at each simulation iteration. (This condition occurs
when you set the ResetMode
of the Error Rate Test
Console to 'Reset at every iteration'.)
Writing a reset
method for each communication
system you create is not mandatory. The reset
method
is optional.
Writing a Run Method. Write a run
method for each communication
system you create. The run
method includes the
core functionality of the system under test. At each simulation iteration,
the Error Rate Test Console calls the run
method
of the system under test.
When designing a communication system, ensure at run time that your system sets components to the current simulation test parameter sweep values. Depending on your unique design, at run time, the communication system:
requests test inputs from the test console using the getInput
method
logs test data to its test probes using the setTestProbeData
method
logs user-data to the test console using the setUserData
method
Although it is recommended you do this at setup time,
the system can also request the current simulation sweep values using
the getTestParameter
method.
Getting Test Inputs From the Error Rate Test Console. At simulation time, the communications system you design can request input data to the Error Rate Test Console. To request a particular type of input data, the system under test must register the specific input type to the Error Rate Test Console. The system under test calls getInput(obj,inputName) method to request test inputs to the test console.
obj represents the handle of the Error Rate Test Console
inputName represents the input that the system under test gets from the Error Rate Test Console
For an Error Rate Test Console, 'NumTransmissions' or 'RandomDiscreetSource'
are acceptable selections for inputName
.
The system under test provides the following inputs:
'NumTransmissions' - calling the getInput
method
returns the frame length. The system itself is responsible for generating
a data frame using a data source.
'RandomIntegerSource' - calling the getInput
method
returns a vector of symbols with a length the Error Rate Test Console FrameLength
property
specifies. If the system registers this source type, then it must
also register a test parameter named 'M' that corresponds to the modulation
order.
Getting the Current Simulation Sweep Value of a Registered
Test Parameter. For each simulation iteration, the system under test may require
the current simulation sweep values from the registered test parameters.
To obtain these values from the Error Rate Test console, the system
under test calls the getTestParameter(sys,name)
method.
Logging Test Data to a Registered Test Probe. At simulation time, the system under test may log data to a
registered test probe using the setTestProbeData(sys,name,data)
method.
sys
represents the handle to the
system
name
represents the name of a registered
test probe
data
represents the data the probe
logs to the Error Rate Test Console.
Logging User-Defined Data To The Test Console. At simulation time, the system under test may log user-data
to the Error Rate Test Console by calling the setUserData
method.
This user-data passes directly to the specific user-defined metric
calculator functions. Log user-data to the Error Rate Test Console
as follows:
setUserData(sys,data)
sys
represents the handle to the
system
data
represents the data the probe
logs to the Error Rate Test Console.
When you run a system that is not attached to an Error Rate Test Console, the system is running in debug mode. Debug mode is useful when evaluating or debugging the code for the system you are designing.
A system that extends the testconsole.SystemBasicAPI class can run by itself, without the need to attach it to a test console. This scenario is referred to as debug mode. Debug mode is useful when evaluating or debugging the code for the system you are designing. For example, if you define break points when designing your system, you can run the system in debug mode and confirm that the system runs without errors or warnings.
Implementing A Default Input Generator Function For Debug Mode. If your system registers a test input and calls the getInput
method
at simulation run time then for it to run in debug mode, the system
must implement a default input generator function. This method should
return an input congruent to the test console.
input = generateDefaultInput(obj)
Running simulations with the Error Rate Test Console involves the following tasks:
Creating a test console
Attaching a system
Defining simulation conditions
Specifying stop criterion
Specifying iteration mode
Specifying reset mode
Specifying sweep values
Registering test points
Running simulations
Getting results and plotting
Creating a Test Console. You create a test console in one of the following ways:
h = commtest.ErrorRate
returns
an error rate test console, h. The error rate test console runs simulations
of a system under test to obtain error rates.
h = commtest.ErrorRate(sys)
returns
an error rate test console, h, with an attached system under test,
sys.
h = commtest.ErrorRate(sys,'PropertyName',PropertyValue,...)
returns
an error rate test console, h, with an attached system under test,
sys. Each specified property, 'PropertyName', is set to the specified
value, PropertyValue.
h = commtest.ErrorRate('PropertyName',PropertyValue,...)
returns
an error rate test console, h, with each specified property 'PropertyName',
set to the specified value, PropertyValue.
Attaching a System to the Error Rate Test Console. You attach a system to the Error Rate Test Console to run simulations and obtain error rate data. There are two ways to attach a system to the Error Rate Test Console.
To attach a system to the Error Rate Test Console, type the following at the MATLAB command line:
attachSystem(testConsole, mySystem)
To attach a system at construction time of an Error Rate Test Console, see Creating a Test Console.
mySystem
is the name of the system
under test
If system under test A is currently attached to the Error Rate
Test Console H1, and you call attachSystem(H2,A)
,
then A detaches from H1 and attaches to Error Rate Test Console H2.
This causes system A to display a warning message, stating that it
has detached from H1 and attached to H2.
Defining Simulation Conditions
Registering a Test Point. You obtain error rate results at different points in the system under test, by defining unique test points. Each test point groups a pair of probes that the system under test uses to log data and the Error Rate Test Console uses to obtain data. In order to create a test point for a pair of probes, the probes must be registered to the Error Rate Test Console.
The Error Rate Test Console calculates error rates by comparing the data available in a pair of probes.
Test points hold error and transmission counts for each sweep point simulation.
The info method displays which test points are registered to the test console.
registerTestPoint(h, name, actprobe, expprobe)
registers
a new test point with name, name, to the error rate test console, h
.
The test point must contain a pair of registered test probes actprobe
and expprobe
whose
data will be compared to obtain error rate values. actprobe
contains
actual data, and expprobe
contains expected data.
Error rates will be calculated using a default error rate calculator
function that simply performs one-to-one comparisons of the data
vectors available in the probes.
registerTestPoint(h, name, actprobe, expprobe, fcnhandle)
adds
a function handle, fcnhandle
, that points to a
user-defined error calculator function that will be used instead of
the default function to compare the data in probes actprobe
and exprobe
,
to obtain error rate results.
A user-defined error calculator function must comply with the following syntax:
[ecnt tcnt] = functionName(act, exp, udata)
where ecnt
output
corresponds to the error count, and tcnt
output
is the number of transmissions used to obtain the error count. Inputs act
and
exp
correspond to actual and expected data. The
error rate test console will set these inputs to the data available
in the pair of test point probes actprobe
and expprobe
previously
mentioned. udata
is a user data input that the
system under test may pass to the test console at run time using the setUserData
method. udata
may
contain data necessary to compute errors such as delays, data buffers,
and so on. The error rate test console will pass the same user data
logged by the system under test to the error calculator functions
of all the registered test points. You call the info method to see
the names of the registered test points and the error rate calculator
functions associated with them, and to see the names of the registered
test probes.
Getting Test Information. Returns a report of the current test console settings.
info(h) displays:
Test console name
System under test name
Available test inputs
Registered test inputs
Registered test parameters
Registered test probes
Registered test points
Metric calculator functions
Test metrics
Running a Simulation. You run simulations by calling the run
method
of the Error Rate Test Console.
run(testConsole)
runs a specified number
of iterations of an attached system under test for a specified set
of parameter values. If a Parallel Computing Toolbox™ license is
available and a parpool is open, then you can distribute the iterations
among the available number of workers.
Getting Results and Plotting Data. Call the getResults
method of the error rate
test console to obtain test results.
r = getResults(testConsole)
returns the simulation
results, r, for the test console, testConsole. r is an object of type testconsole.Results
and
contains the simulation data for all the registered test points.
You call the getData
method of results object
r to get simulation results data. You call the plot and semilogy
method
of the results object r to plot results data. See testconsole.Results
for more information.
Parsing and Plotting Results for Multiple Parameter Simulations. The DPSKModulationTester.mat file contains an Error Rate Test Console with a DPSK modulation system. This system defines three test parameters:
The bit energy to noise power spectral density ratio, EbNo
(in
decibels)
The modulation order, M
The maximum Doppler shift, MaxDopplerShift
(in
hertz)
These parameters have the following sweep values:
EbNo
= [-2:4] dB
M
= [2 4 8 16]
MaxDopplerShift
= [0 0.001 0.09]
Hz
Because simulations generally take a long time to run, a simulation was run offline. DPSKModulationTester.mat file contains a saved Error Rate Test Console with the saved results. The simulations were run to obtain at least 2500 errors and 5e6 frame transmissions per simulation point.
Load the simulation results by entering the following at the MATLAB command line:
load DPSKModulationTester.mat
To parse and plot results for multiple parameter simulations, perform the following steps:
Using the getSweepParameterValues
method,
display the sweep parameter values used in the simulation for each
test parameter. For example, you display the sweep values for MaxDopplerShift
by
entering:
getTestParameterSweepValues(testConsole,'MaxDopplerShift')
MATLAB returns the following result:
ans = 0 0.0010 0.0900
Get the results object that parses and plots simulation results by entering the following at the command line:
DPSKResults = getResults(testConsole)
MATLAB returns the following result:
DPSKResults = TestConsoleName: 'commtest.ErrorRate' SystemUnderTestName: 'commexample.DPSKModulation' IterationMode: 'Combinatorial' TestPoint: 'BitErrors' Metric: 'ErrorRate' TestParameter1: 'EbNo' TestParameter2: 'None'
Use the setParsingValues
method to enable
the plotting of error rate results versus Eb/No for a modulation order
of 4 and maximum Doppler shift of 0.001 Hz. To do so, enter the following:.
setParsingValues(DPSKResults,'M',4,'MaxDopplerShift',0.001)
Use the getParsingValues
method to verify
the current parsing values settings:
getParsingValues(DPSKResults)
MATLAB returns the following:
ans = EbNo: -2 M: 4 MaxDopplerShift: 1.0000e-003
If not specified, the parsing value for a test parameter defaults
to its first sweep value. In this example, the first sweep value for
EbNo equals -2 dB. However, in this example, TestParameter1
is
set to EbNo; therefore, the Error Rate Test Console plots results
for all EbNo sweep values, not just for the value listed by the getParsingValues
method.
Obtain a log-scale plot of bit error rate versus Eb/No for a modulation order of 4 and a maximum Doppler shift of 0.001 Hz:
semilogy(DPSKResults)
MATLAB generates the following figure.
Set the TestParameter2
property of the
results object to 'MaxDopplerShift'. This setting enables the plotting
of multiple error rate curves versus Eb/No for each sweep value of
the maximum Doppler shift.
DPSKResults.TestParameter2 = 'MaxDopplerShift';
Obtain log-scale plots of bit error rate versus Eb/No for a modulation order of 2 at each of the maximum Doppler shift sweep values.
setParsingValues(DPSKResults,'M',2)
semilogy(DPSKResults)
MATLAB generates the following figure.
Obtain the same type of curves as in the previous step, but now for a modulation order of 16.
setParsingValues(DPSKResults,'M',16)
semilogy(DPSKResults)
MATLAB generates the following figure.
Obtain error rate plots versus the modulation order for
each Eb/No sweep value by setting TestParameter1
equal
to M and TestParameter2
equal to EbNo. You can plot
the results for the case when the maximum Doppler shift is 0 Hz by
using the setParsingValues
method:
DPSKResults.TestParameter1 = 'M';
DPSKResults.TestParameter2 = 'EbNo';
setParsingValues(DPSKResults, 'MaxDopplerShift',0)
semilogy(DPSKResults)
MATLAB generates the following figure.
Obtain a data matrix with the bit error rate values previously plotted by entering the following:
BERMatrix = getData(DPSKResults)
MATLAB returns the following result:
BERMatrix = Columns 1 through 7 0.2660 0.2467 0.2258 0.2049 0.1837 0.1628 0.1418 0.3076 0.2889 0.2702 0.2504 0.2296 0.2082 0.1871 0.3510 0.3384 0.3258 0.3120 0.2983 0.2837 0.2685 0.3715 0.3631 0.3535 0.3442 0.3350 0.3246 0.3147 Columns 8 through 13 0.1217 0.1022 0.0844 0.0677 0.0534 0.0406 0.1658 0.1451 0.1254 0.1065 0.0890 0.0728 0.2531 0.2369 0.2204 0.2042 0.1874 0.1704 0.3044 0.2945 0.2839 0.2735 0.2626 0.2512
The rows of the matrix correspond to the values of the test
parameter defined by the TestParameter1
property,
M. The columns correspond to the values of the test parameter defined
by the TestParameter2
property, EbNo.
Plot the results as a 3-D data plot by entering the following:
surf(DPSKResults)
MATLAB generates the following plot:
In this case, the parameter defined by the TestParameter1
property,
M, controls the x-axis and the parameter defined by the TestParameter2
property,
EbNo, controls the y-axis.
Tasks for running bit error rate simulations for various En/No and modulation order values.
Load the Error Rate Test Console. The Error Rate Test Console is a simulation tool for obtaining error rate results. The MATLAB software includes a data file for use with the Error Rate Test Console. You will use the data file while performing the steps of this tutorial. The data file contains an Error Rate Test Console object with an attached Gray coded modulation system. This example Error Rate Test Console is configured to run bit error rate simulations for various EbNo and modulation order, or M, values.
Load the file containing the Error Rate Test Console and attached Gray coded modulation system. At the MATLAB command line, enter:
load GrayCodedModTester_EbNo_M
Examine the test console by displaying its properties. At the MATLAB command line, enter:
testConsole
MATLAB returns the following output:
testConsole = Description: 'Error Rate Test Console' SystemUnderTestName: 'commexample.GrayCodedMod_EbNo_M' IterationMode: 'Combinatorial' SystemResetMode: 'Reset at new simulation point' SimulationLimitOption: 'Number of errors or transmissions' TransmissionCountTestPoint: 'DemodBitErrors' MaxNumTransmissions: 100000000 ErrorCountTestPoint: 'DemodBitErrors' MinNumErrors: 100
Notice that SystemUnderTest is a Gray coded modulation system. Because the SimulationLimitOption is 'Number of error or transmission', the simulation runs until reaching 100 errors or 1e8 bits.
Run the Simulation and Obtain Results. In this example, you use tic
and toc
to
compare simulation run time.
Run the simulation, using the tic
and toc
commands
to measure simulation time. At the MATLAB command line, enter:
tic; run(testConsole); toc
MATLAB returns output similar to the following:
Running simulations... Elapsed time is 174.671632 seconds.
Obtain the results of the simulation using the getResults
method
by typing the following at the MATLAB command line:
grayResults = getResults(testConsole)
MATLAB returns the following output:
grayResults = TestConsoleName: 'commtest.ErrorRate' SystemUnderTestName: 'commexample.GrayCodedMod_EbNo_M' IterationMode: 'Combinatorial' TestPoint: 'DemodBitErrors' Metric: 'ErrorRate' TestParameter1: 'EbNo' TestParameter2: 'None'
In the next section, you use the results object to obtain error values and plot error rate curves.
Generate an Error Rate Results Figure Window. The semilogy
method generates a figure containing
error rate curves for the demodulator bit error test point (DemodBitErrors)
of the Gray coded modulation system. The next figure shows an Error
Rate and E_{b} over N_{o} curve
for the demodulator bit errors test point. This test point collects
bit errors by comparing the bits the system transmits with the bits
it receives. The x-axis displays the TestParameter1
property
of grayResults
, which contains EbNo values.
Generate the figure by entering the following at the MATLAB command line:
semilogy(grayResults)
This script generates the following figure.
Set the TestParameter2
property
to M
. At the MATLAB command line, enter:
grayResults.TestParameter2 = 'M'
Previously, the simulation ran for multiple modulation order
(M) values. The x-axis displays the TestParameter1
property
of grayResults
, which contains EbNo values. Although
the simulation ran for multiple M values, this run contains data for
M=2.
Plot multiple error rate curves by entering the following at the MATLAB command line.
semilogy(grayResults)
This script generates the following figure.
Run Parallel Simulations Using Parallel Computing Toolbox Software. If you have a Parallel Computing Toolbox user license and you create a parpool, the test console runs the simulation in parallel. This approach reduces the processing time.
Note: If you do not have a Parallel Computing Toolbox user license you are unable to perform this section of the tutorial. |
If you have a Parallel Computing Toolbox license, run the following command to start your default parpool:
mypool = parpool()
If you have a multicore computer, then the default parpool uses the cores as workers.
Using the workers, run the simulation. At the MATLAB command line, enter:
tic; run(testConsole); toc
MATLAB returns output similar to the following:
4 workers available for parallel computing. Simulations ..., will be distributed among these workers. Running simulations... Elapsed time is 87.449652 seconds.
Notice that the simulation runs more than three times as fast than in the previous section.
Create a System File and Attach It to the Test Console. In the previous sections, you used an existing Gray coded modulator system file to generate data. In this section, you create a system file and then attach it to the Error Rate Test Console.
This example outlines the tasks necessary for converting legacy code to a system file you can attach to the Error Rate Test Console. Use commdoc_gray as the starting point for your system file. The files you use in this section of the tutorial reside in the following folder:
matlab\help\toolbox\comm\examples
Copy the system basic API template, SystemBasicTemplate.m, as MyGrayCodedModulation.m.
Rename the references to the system name in the file. First, rename the system definition by changing the class name to MyGrayCodedModulation. Replace the following lines, lines 1 and 2, of the file:
classdef SystemBasicTemplate < testconsole.SystemBasicAPI %SystemBasicTemplate Template for creating a system
with these lines:
classdef MyGrayCodedModulation < testconsole.SystemBasicAPI %MyGrayCodedModulation Gray coded modulation system
Rename the constructor by replacing:
function obj = SystemBasicTemplate %SystemBasicTemplate Construct a system
function obj = MyGrayCodedModulation %MyGrayCodedModulation Construct a Gray coded modulation system
Enter a description for your system. Update the obj.Description
parameter
with the following information:
obj.Description = 'Gray coded modulation';
Because you are not using the reset
and setup
methods
for this system, leave these methods empty.
Copy lines 12–44 from commdoc_gray.m to the
body of the run
method.
Copy Lines 54–57 from commdoc_gray.m to the
body of the run
method.
Change EbNo to a test parameter. This change allows the system to obtain EbNo values from the Error Rate Test Console. As a test parameter, EbNo becomes a variable, which allows simulations to run for different values. Locate the following line of syntax in the file:
EbNo = 10; % In dB
EbNo = getTestParameter(obj,'EbNo');
Add modulation order, M, as a new test parameter for the simulation. Locate the following syntax:
M = 16; % Size of signal constellation
M = getTestParameter(obj,'M');
Register the test parameters to the test console.
Declare EbNo as a test parameter by placing the following
line of code in the body of the register
method:
registerTestParameter(obj,'EbNo',0,[-50 50]);
Declare M as a test parameter by placing the following
line of code in the body of the register
method:
registerTestParameter(obj,'M',16,[2 1024]);
The parameter defaults to 16 QAM Modulation and can take values from 2 through 1024.
Add EbNo and M to the test parameters list in the MyGrayCodedModulationFile file.
% Test Parameters properties EbNo = 0; M = 16; end
Define test probe locations in the run
method.
In this example, you are calculating end-to-end error rate. This calculation
requires transmitted bits and received bits. Add one probe for obtaining
transmitted bits and one probe for received bits.
Locate the random binary data stream creation code by searching for the following lines:
% Create a binary data stream as a column vector. x = randi([0 1],n,1); % Random binary data stream
Add a probe, TxBits, after the random binary data stream creation:
% Create a binary data stream as a column vector. x = randi([0 1],n,1); % Random binary data stream setTestProbeData(obj,'TxBits',x);
Locate the demodulation code by searching for the following lines:
% Demodulate signal using 16-QAM. z = demodulate(hDemod,yRx);
Add a probe, RxBits, after the demodulation code.
% Demodulate signal using 16-QAM. z = demodulate(hDemod,yRx); setTestProbeData(obj,'RxBits',z);
This code sends the binary received data stream, z, to the probe RxBits.
Register the test probes to the Error Rate Test Console,
making it possible to obtain data from the system. Add these probes
to the function register(obj)
by adding two lines
to the register
method:
function register(obj) % REGISTER Register the system with a test console % REGISTER(H) registers test parameters and test probes of the % system, H, with a test console. registerTestParameter(obj,'EbNo',0,[-50 50]); registerTestParameter(obj,'M',16,[2 1024]); registerTestProbe(obj,'TxBits') registerTestProbe(obj,'RxBits') end
Save the file. The file is ready for use with the system.
Create a Gray coded modulation system. At the MATLAB command line, enter:
mySystem = MyGrayCodedModulation
MATLAB returns the following output:
mySystem = Description: 'Gray coded modulation' EbNo: 0 M: 16
Create an Error Rate Test Console by entering the following at the MATLAB command line:
testConsole = commtest.ErrorRate
The MATLAB software returns the following output:
testConsole = Description: 'Error Rate Test Console' SystemUnderTestName: 'commtest.MPSKSystem' FrameLength: 500 IterationMode: 'Combinatorial' SystemResetMode: 'Reset at new simulation point' SimulationLimitOption: 'Number of transmissions' TransmissionCountTestPoint: 'Not set' MaxNumTransmissions: 1000
Attach the system file MyGrayCodedModulation to the error rate test console by entering the following at the MATLAB command line:
attachSystem(testConsole, mySystem)
Configure the Error Rate Test Console and Run a Simulation. Configure the Error Rate Test Console to obtain error rate metrics from the attached system. The Error Rate Test Console defines metrics as number of errors, number of transmissions, and error rate.
At the MATLAB command line, enter:
registerTestPoint(testConsole, 'DemodBitErrors', 'TxBits', 'RxBits');
This line defines the test point, DemodBitErrors, and compares bits from the TxBits probe to the bits from the RxBits probe. The Error Rate Test Console calculated metrics for this test point.
Configure the Error Rate Test Console to run simulations for EbNo values. Start at 2 dB and end at 10 dB, with a step size of 2 dB and M values of 2, 4, 8, and 16. At the MATLAB command line, enter:
setTestParameterSweepValues(testConsole, 'EbNo', 2:2:10) setTestParameterSweepValues(testConsole, 'M', [2 4 8 16])
Set the simulation limit to the number of transmissions.
testConsole.SimulationLimitOption = 'Number of transmissions'
Set the maximum number of transmissions to 1000.
testConsole.MaxNumTransmissions = 1000
Configure the Error Rate Test Console so it uses the demodulator bit error test point for determining the number of transmitted bits.
testConsole.TransmissionCountTestPoint = 'DemodBitErrors'
Run the simulation. At the MATLAB command line, enter:
run(testConsole)
Obtain the results of the simulation. At the MATLAB command line, enter:
grayResults = getResults(testConsole)
To obtain more accurate results, run the simulations for a given minimum number of errors. In this example, you also limit the number of simulation bits so that the simulations do not run indefinitely. At the MATLAB command line, enter:
testConsole.SimulationLimitOption = 'Number of errors or transmissions';
testConsole.MinNumErrors = 100;
testConsole.ErrorCountTestPoint = 'DemodBitErrors';
testConsole.MaxNumTransmissions = 1e8;
testConsole
Run the simulation by entering the following at the MATLAB command line.
run(testConsole);
Generate the new results in a Figure window by entering the following at the MATLAB command line.
grayResults = getResults(testConsole);
grayResults.TestParameter2 = 'M'
semilogy(grayResults)
This script generates the following figure.
Optimize System Performance Using Parameterized Simulations. In the previous example, the system only utilizes the run
method.
Every time the object calls the run
method, which
is every 3e4 bits for this simulation, the object sets the M and SNR
values. This time interval includes: obtaining numbers from the test
console, calculating intermediate values, and setting other variables.
In contrast, the system basic API provides a setup
method
where the Error Rate Test Console configures the system once for each
simulation point. This change relieves the run
method
from getting and setting simulation parameters, thus reducing simulation
time.
The run
method of a system also creates a
new modulator (hMod) and a new demodulator (hDemod). Creating a modulator
or a demodulator is much more time consuming than just modifying a
property of these objects. Create a modulator and a demodulator object
once when the system is constructed. Then, modify its properties in
the setup method of the system to speed up the simulations.
Save the file MyGrayCodedModulation as MyGrayCodedModulationOptimized.
In the MyGrayCodedModulationOptimized file, replace the constructor name and the class definition name.
Locate the following lines of code:
classdef MyGrayCodedModulation < testconsole.SystemBasicAPI %MyGrayCodedModulation Gray coded modulation system
Replace them with:
classdef MyGrayCodedModulationOptimized < testconsole.SystemBasicAPI %MyGrayCodedModulationOptimized Gray coded modulation system
In the MyGrayCodedModulationOptimized ﬁle, replace the constructor name.
Locate the following lines of code:
function obj = MyGrayCodedModulation %MyGrayCodedModulation Construct a Gray coded modulation system
Replace them with:
function obj = MyGrayCodedModulationOptimized %MyGrayCodedModulationOptimized Construct a Gray %coded modulation system
Move the oversampling rate definition from the run
method
to the setup
method.
nSamp = 1; % Oversampling rate
Move code related to setting M to the setup
method.
Cut the following lines from the run
method and
paste to the setup
method.
M = getTestParameter(obj,'M'); k = log2(M); % Number of bits per symbol
In the setup
method, replace M
with the object property M.
obj.M = getTestParameter(obj,'M'); k = log2(obj.M); % Number of bits per symbol
run
method.Move code related to setting EbNo to the setup
method.
Cut the following lines from the run
method and
paste to the setup
method.
EbNo = getTestParameter(obj,'EbNo'); SNR = EbNo + 10*log10(k) - 10*log10(nSamp);
In the setup
method, replace EbNo
with the object property EbNo. This change provides access to the
EbNo value from the run method.
obj.EbNo = getTestParameter(obj,'EbNo'); SNR = obj.EbNo + 10*log10(k) - 10*log10(nSamp);
Create a new internal variable called SNR to store
the calculated SNR value. Define the SNR property as a private property;
it is not a test parameter. With this change, the system calculates
SNR in the setup
method and accesses it from the run
method.
Add the following lines of code the system file, after the Test Parameters
block.
%================================================================= % Internal variables properties (Access = private) SNR end
In the setup
method, replace SNR
with object property SNR.
obj.SNR = obj.EbNo + 10*log10(k) - 10*log10(nSamp);
In the run
method, replace M with
obj.M and SNR with obj.SNR.
hMod = comm.RectangularQAMModulator(obj.M); % Create a 16-QAM modulator yNoisy = awgn(yTx,obj.SNR,'measured');
run
method creates the QAM modulator and demodulator.Move the QAM modulator and demodulator creation out
of the run
method. Move following lines from the run
method
to the constructor (i.e the method named MyGrayCodedModulationOptimized)
%% Create Modulator and Demodulator hMod = comm.RectangularQAMModulator(obj.M); % Create a 16-QAM modulator hMod.BitInput = true; % Accept bits as inputs hMod.SymbolMapping = 'Gray'; % Gray coded symbol mapping hDemod = comm.RectangularQAMDemodulator(obj.M); % Create a 16-QAM demodulator hDemod.BitOutput = true; % Output bits hDemod.SymbolMapping = 'Gray'; % Gray coded symbol mapping
Create private properties called Modulator and Demodulator to store the modulator and demodulator objects.
% Internal variables properties (Access = private) SNR Modulator Demodulator end
In the constructor method, replace hMod
and hDemod
with
the object property obj.Modulator
and obj.Demodulator
respectively.
% Create a 16-QAM modulator obj.Modulator = comm.RectangularQAMModulator(obj.M, ... 'BitInput',true,'SymbolMapping','Gray'); % Create a 16-QAM demodulator obj.Demodulator = comm.RectangularQAMDemodulator(obj.M, ... 'BitOutput',true,'SymbolMapping','Gray');
In the run
method, replace hMod
and hDemod
with
object properties obj.Modulator
and obj.Demodulator
.
y = modulate(obj.Modulator,x); z = demodulate(obj.Demodulator,yRx);
Locate the setup region of the ﬁle.
function setup(obj) % SETUP Initialize the system % SETUP(H) gets current test parameter value(s) from the test % console and initializes system, H, accordingly.
Set the M value of the modulator and demodulator by adding the following lines of code to the setup.
obj.Modulator.M = obj.M; obj.Demodulator.M = obj.M;
Save the ﬁle.
Create an optimized system. At the MATLAB command line, enter:
myOptimSystem = MyGrayCodedModulationOptimized
Create an Error Rate Test Console and attach the system to the test console. At the MATLAB command line, type:
testConsole = commtest.ErrorRate(myOptimSystem)
At the MATLAB command line, type:
registerTestPoint(testConsole, 'DemodBitErrors', 'TxBits', 'RxBits');
This line defines the test point, DemodBitErrors, and compares bits from the TxBits probe to the bits from the RxBits probe. The Error Rate Test Console calculated metrics for this test point.
Configure the Error Rate Test Console to run simulations for EbNo values. Start at 2 dB and end at 10 dB, with a step size of 2 dB and M values of 2, 4, 8, and 16. At the MATLAB command line, type:
setTestParameterSweepValues(testConsole, 'EbNo', 2:2:10) setTestParameterSweepValues(testConsole, 'M', [2 4 8 16])
Configure the Error Rate Test Console so it uses the demodulator bit error test point for determining the number of transmitted bits.
testConsole.TransmissionCountTestPoint = 'DemodBitErrors'
To obtain more accurate results, run the simulations for a given minimum number of errors. In this example, you also limit the number of simulation bits so that the simulations do not run indefinitely. At the MATLAB command line, type:
testConsole.SimulationLimitOption = 'Number of errors or transmissions';
testConsole.MinNumErrors = 100;
testConsole.ErrorCountTestPoint = 'DemodBitErrors';
testConsole.MaxNumTransmissions = 1e8;
testConsole
Run the simulation. At the MATLAB command line, type:
tic; run(testConsole); toc
MATLAB returns the following information:
Running simulations... Elapsed time is 191.748359 seconds.
Notice that these optimization changes reduce the simulation run time about 10%.
Generate the new results in a Figure window. At the MATLAB command line, type:
grayResults = getResults(testConsole); grayResults.TestParameter2 = 'M' semilogy(grayResults)