Main Content

Correct symbol timing clock skew

The `comm.SymbolSynchronizer`

System object™ corrects symbol timing clock skew between a
single-carrier transmitter and receiver for PAM, PSK, QAM, and OQPSK modulation schemes. For
more information, see Symbol Synchronization Overview.

**Note**

The input signal operates on a sample-rate basis and the output signal operates on a symbol-rate basis.

To correct symbol timing clock skew:

Create the

`comm.SymbolSynchronizer`

object and set its properties.Call the object with arguments, as if it were a function.

To learn more about how System objects work, see What Are System Objects?

creates
a symbol synchronizer System object for correcting the clock skew between a single-carrier transmitter and
receiver.`symbolSync`

= comm.SymbolSynchronizer

sets properties using one or more name-value pairs. For example,
`symbolSync`

= comm.SymbolSynchronizer(`Name`

,`Value`

)`comm.SymbolSynchronizer('Modulation','OQPSK')`

configures the symbol
synchronizer System object for an OQPSK-modulated input signal. Enclose each property name in
quotes.

Tunable `DampingFactor`

,
`NormalizedLoopBandwidth`

, and `DetectorGain`

properties enable you to optimize synchronizer performance in your simulation loop without
releasing the object.

Unless otherwise indicated, properties are *nontunable*, which means you cannot change their
values after calling the object. Objects lock when you call them, and the
`release`

function unlocks them.

If a property is *tunable*, you can change its value at
any time.

For more information on changing property values, see System Design in MATLAB Using System Objects.

`Modulation`

— Modulation type`'PAM/PSK/QAM'`

(default) | `'OQPSK'`

Modulation type, specified as `'PAM/PSK/QAM'`

or
`'OQPSK'`

.

**Tunable: **No

**Data Types: **`char`

| `string`

`TimingErrorDetector`

— Timing error detector method`Zero-Crossing (decision-directed)`

(default) | `Gardner (non-data-aided)`

| `Early-Late (non-data-aided)`

| `Mueller-Muller (decision-directed)`

Timing error detector method, specified as ```
Zero-Crossing
(decision-directed)
```

, `Gardner (non-data-aided)`

,
`Early-Late (non-data-aided)`

, or ```
Mueller-Muller
(decision-directed)
```

. This property assigns the timing error detection scheme
used in the synchronizer. For more information, see Timing Error Detection (TED).

**Tunable: **No

**Data Types: **`char`

| `string`

`SamplesPerSymbol`

— Samples per symbol`2`

(default) | integer greater than 1Samples per symbol, specified as an integer greater than 1.

**Tunable: **No

**Data Types: **`double`

`DampingFactor`

— Damping factor of loop filter`1`

(default) | positive scalarDamping factor of the loop filter, specified as a positive scalar. For more information, see Loop Filter.

**Tunable: **Yes

**Data Types: **`double`

| `single`

`NormalizedLoopBandwidth`

— Normalized bandwidth of loop filter`0.01`

(default) | scalar in the range (0, 1)Normalized bandwidth of the loop filter, specified as a scalar in the range (0, 1). The loop bandwidth is normalized to the sample rate of the input signal. For more information, see Loop Filter.

**Note**

To ensure the symbol synchronizer locks, set the
`NormalizedLoopBandwidth`

property to a value less than
`0.1`

.

**Tunable: **Yes

**Data Types: **`double`

| `single`

`DetectorGain`

— Phase detector gain`2.7`

(default) | positive scalarPhase detector gain, specified as a positive scalar.

**Tunable: **Yes

**Data Types: **`double`

| `single`

**For versions earlier than R2016b, use the step
function to run the System object algorithm. The arguments to step are the
object you created, followed by the arguments shown in this section.**

**For example, y = step(obj,x) and y = obj(x) perform equivalent operations.**

corrects symbol timing clock skew between a single-carrier transmitter and receiver based
on the input samples and outputs synchronized symbols.`symbols`

= symbolSync(`samples`

)

The input operates on a sample-rate basis and the output signal operates on a symbol-rate basis.

You can tune the

`DampingFactor`

,`NormalizedLoopBandwidth`

, and`DetectorGain`

properties to improve the synchronizer performance.

`samples`

— Input samplesscalar (default) | column vector

Input samples, specified as a scalar or column vector of a PAM-, PSK-, QAM-, or OQPSK-modulated single-carrier signal.

**Data Types: **`double`

| `single`

**Complex Number Support: **Yes

`symbols`

— Synchronized symbolscolumn vector

Synchronized symbols, returned as a variable-sized column vector. The output
symbols inherit the data type from the input samples. For an input with dimensions
*N*_{samp}-by-1, this output has dimensions
*N*_{sym}-by-1.
*N*_{sym} is approximately equal to
*N*_{samp} divided by
*N*_{sps}, where
*N*_{sps} is equal to the
`SamplesPerSymbol`

property value. If the output exceeds the
maximum output size of $$\lceil \frac{{N}_{\text{samp}}}{{N}_{\text{sps}}}\times 1.1\rceil $$, it is truncated.

`timingErr`

— Estimated timing errorscalar in the range [0, 1] | column vector of elements in the range [0, 1]

Estimated timing error for each input sample, returned as a scalar in the range
[0, 1] or column vector of elements in the range [0, 1]. The estimated timing error is
normalized to the input sample rate. `timingErr`

has the same data
type and size as input `samples`

.

To use an object function, specify the
System object as the first input argument. For
example, to release system resources of a System object named `obj`

, use
this syntax:

release(obj)

`comm.SymbolSynchronizer`

Correct a fixed symbol timing error on a noisy QPSK-modulated signal. Check the bit error rate (BER) of the synchronized received signal.

Initialize simulation parameters.

M = 4; % Modulation order for QPSK nSym = 5000; % Number of symbols in a packet sps = 4; % Samples per symbol timingErr = 2; % Samples of timing error snr = 15; % Signal-to-noise ratio (dB)

Create root raised cosine (RRC) transmit and receive filter System objects.

txfilter = comm.RaisedCosineTransmitFilter( ... 'OutputSamplesPerSymbol',sps); rxfilter = comm.RaisedCosineReceiveFilter( ... 'InputSamplesPerSymbol',sps,'DecimationFactor',2);

Create a symbol synchronizer System object to correct the timing error.

symbolSync = comm.SymbolSynchronizer;

Generate random M-ary symbols and apply QPSK modulation.

data = randi([0 M-1],nSym,1); modSig = pskmod(data,M,pi/4);

Create a delay object to introduce a fixed timing error of 2 samples. Because the transmit RRC filter outputs 4 samples per symbol, 1 sample is equivalent to a 1/4 symbol through the fixed delay and channel.

```
fixedDelay = dsp.Delay(timingErr);
fixedDelaySym = ceil(fixedDelay.Length/sps); % Round fixed delay to nearest integer in symbols
```

Filter the modulated signal through a transmit RRC filter by using the `txfilter`

object. Apply a signal timing error by using the `fixedDelay`

object.

txSig = txfilter(modSig); delaySig = fixedDelay(txSig);

Pass the delayed signal through an AWGN channel with a 15 dB signal-to-noise ratio.

`rxSig = awgn(delaySig,snr,'measured');`

Filter the modulated signal through a receive RRC filter by using the `rxfilter`

object. Display the scatter plot. Due to the timing error, the received signal does not align with the expected QPSK reference constellation.

rxSample = rxfilter(rxSig); scatterplot(rxSample(1001:end),2)

Correct the symbol timing error by using the `symbolSync`

object. Display the scatter plot. The synchronized signal now aligns with the expected QPSK constellation.

rxSync = symbolSync(rxSample); scatterplot(rxSync(1001:end),2)

Demodulate the QPSK signal.

recData = pskdemod(rxSync,M,pi/4);

Compute, in symbols, the total system delay due to the fixed delay and the transmit and receive RRC filters.

```
sysDelay = dsp.Delay(fixedDelaySym + txfilter.FilterSpanInSymbols/2 + ...
rxfilter.FilterSpanInSymbols/2);
```

Compute the BER, taking into account the system delay.

[numErr,ber] = biterr(sysDelay(data),recData)

numErr = 12

ber = 0.0012

Correct a fixed symbol timing error on a noisy BPSK transmission signal. Check the bit error rate (BER) of the synchronized received signal.

Initialize simulation parameters.

M = 2; % Modulation order for BPSK nSym = 20000; % Number of symbols in a packet sps = 4; % Samples per symbol timingErr = 2; % Samples of timing error snr = 15; % Signal-to-noise ratio (dB)

Create root raised cosine (RRC) transmit and receive filter System objects.

txfilter = comm.RaisedCosineTransmitFilter(... 'OutputSamplesPerSymbol',sps); rxfilter = comm.RaisedCosineReceiveFilter(... 'InputSamplesPerSymbol',sps,'DecimationFactor',1);

Create a symbol synchronizer System object™ to correct the timing error.

symbolSync = comm.SymbolSynchronizer(... 'SamplesPerSymbol',sps, ... 'NormalizedLoopBandwidth',0.01, ... 'DampingFactor',1.0, ... 'TimingErrorDetector','Early-Late (non-data-aided)');

Generate random data symbols and apply BPSK modulation.

data = randi([0 M-1],nSym,1); modSig = pskmod(data,M);

Create a delay object to introduce a fixed timing error of 2 samples. Because the transmit RRC filter outputs 4 samples per symbol, 1 sample is equivalent to a 1/4 symbol through the fixed delay and channel.

```
fixedDelay = dsp.Delay(timingErr);
fixedDelaySym = ceil(fixedDelay.Length/sps); % Round fixed delay to nearest integer in symbols
```

Filter the modulated signal through a transmit RRC filter by using the `txfilter`

object. Apply a signal timing error by using the `fixedDelay`

object.

txSig = txfilter(modSig); delayedSig = fixedDelay(txSig);

Pass the delayed signal through an AWGN channel.

`rxSig = awgn(delayedSig,snr,'measured');`

Filter the modulated signal through a receive RRC filter by using the `rxfilter`

object. Display the scatter plot. Due to the timing error, the received signal does not align with the expected BPSK reference constellation.

rxSample = rxfilter(rxSig); scatterplot(rxSample(10000:end),2)

Correct the symbol timing error by using the `symbolSync`

object. Display the scatter plot. The synchronized signal now aligns with the expected BPSK constellation.

rxSync = symbolSync(rxSample); scatterplot(rxSync(10000:end),2)

Demodulate the BPSK signal.

recData = pskdemod(rxSync,M);

Compute, in symbols, the total system delay due to the fixed delay and the transmit and receive RRC filters.

```
sysDelay = dsp.Delay(fixedDelaySym + txfilter.FilterSpanInSymbols/2 + ...
rxfilter.FilterSpanInSymbols/2);
```

Compute the BER, taking into account the system delay.

[numErr1,ber1] = biterr(sysDelay(data),recData)

numErr1 = 8

ber1 = 4.0000e-04

Correct symbol timing and frequency offset errors by using the `comm.SymbolSynchronizer`

and `comm.CarrierSynchronizer`

System objects.

**Configuration**

Initialize simulation parameters.

M = 16; % Modulation order nSym = 2000; % Number of symbols in a packet sps = 2; % Samples per symbol spsFilt = 8; % Samples per symbol for filters and channel spsSync = 2; % Samples per symbol for synchronizers lenFilt = 10; % RRC filter length

Create a matched pair of root raised cosine (RRC) filter System objects for transmitter and receiver.

txfilter = comm.RaisedCosineTransmitFilter('FilterSpanInSymbols',lenFilt, ... 'OutputSamplesPerSymbol',spsFilt,'Gain',sqrt(spsFilt)); rxfilter = comm.RaisedCosineReceiveFilter('FilterSpanInSymbols',lenFilt, ... 'InputSamplesPerSymbol',spsFilt,'DecimationFactor',spsFilt/2,'Gain',sqrt(1/spsFilt));

Create a phase-frequency offset System object to introduce a 100 Hz Doppler shift.

doppler = comm.PhaseFrequencyOffset('FrequencyOffset',100, ... 'PhaseOffset',45,'SampleRate',1e6);

Create a variable delay System object to introduce timing offsets.

varDelay = dsp.VariableFractionalDelay;

Create carrier and symbol synchronizer System objects to correct for Doppler shift and timing offset, respectively.

carrierSync = comm.CarrierSynchronizer('SamplesPerSymbol',spsSync); symbolSync = comm.SymbolSynchronizer( ... 'TimingErrorDetector','Early-Late (non-data-aided)', ... 'SamplesPerSymbol',spsSync);

Create constellation diagram System objects to view the results.

refConst = qammod(0:M-1,M,'UnitAveragePower',true); cdReceive = comm.ConstellationDiagram('ReferenceConstellation',refConst, ... 'SamplesPerSymbol',spsFilt,'Title','Received Signal'); cdDoppler = comm.ConstellationDiagram('ReferenceConstellation',refConst, ... 'SamplesPerSymbol',spsSync,'Title','Frequency Corrected Signal'); cdTiming = comm.ConstellationDiagram('ReferenceConstellation',refConst, ... 'SamplesPerSymbol',spsSync,'Title','Frequency and Timing Synchronized Signal');

**Main Processing Loop**

The main processing loop:

Generates random symbols and apply QAM modulation.

Filters the modulated signal.

Applies frequency and timing offsets.

Passes the transmitted signal through an AWGN channel.

Filters the received signal.

Corrects the Doppler shift.

Corrects the timing offset.

for k = 1:15 data = randi([0 M-1],nSym,1); modSig = qammod(data,M,'UnitAveragePower',true); txSig = txfilter(modSig); txDoppler = doppler(txSig); txDelay = varDelay(txDoppler,k/15); rxSig = awgn(txDelay,25); rxFiltSig = rxfilter(rxSig); rxCorr = carrierSync(rxFiltSig); rxData = symbolSync(rxCorr); end

**Visualization**

Plot the constellation diagrams of the received signal, frequency corrected signal, and frequency and timing synchronized signal. Specific constellation points cannot be identified in the received signal and can be only partially identified in the frequency corrected signal. However, the timing and frequency synchronized signal aligns with the expected QAM constellation points.

cdReceive(rxSig)

cdDoppler(rxCorr)

cdTiming(rxData)

Correct a monotonically increasing symbol timing error on a noisy 8-PSK signal. Display the normalized timing error.

Initialize simulation parameters.

M = 8; % Modulation order nSym = 5000; % Number of symbol in a packet sps = 2; % Samples per symbol nSamp = sps*nSym; % Number of samples in a packet

Create root raised cosine (RRC) transmit and receive filter System objects.

txfilter = comm.RaisedCosineTransmitFilter( ... 'OutputSamplesPerSymbol',sps); rxfilter = comm.RaisedCosineReceiveFilter( ... 'InputSamplesPerSymbol',sps, ... 'DecimationFactor',1);

Create a variable fractional delay System object™ to introduce a monotonically increasing timing error.

varDelay = dsp.VariableFractionalDelay;

Create a symbol synchronizer System object to correct the timing error.

symbolSync = comm.SymbolSynchronizer(... 'TimingErrorDetector','Mueller-Muller (decision-directed)', ... 'SamplesPerSymbol',sps);

Generate random 8-ary symbols and apply 8-PSK modulation.

data = randi([0 M-1],nSym,1); modSig = pskmod(data,M,pi/8);

Filter the modulated signal through a raised cosine transmit filter and apply a monotonically increasing timing delay.

vdelay = (0:1/nSamp:1-1/nSamp)'; txSig = txfilter(modSig); delaySig = varDelay(txSig,vdelay);

Pass the delayed signal through an AWGN channel with a 15 dB signal-to-noise ratio.

`rxSig = awgn(delaySig,15,'measured');`

Filter the modulated signal through a receive RRC filter. Display the scatter plot. Due to the timing error, the received signal does not align with the expected 8-PSK reference constellation.

rxSample = rxfilter(rxSig); scatterplot(rxSample,sps)

Correct the symbol timing error by using the `symbolSync`

object. Display the scatter plot. The synchronized signal now aligns with the expected 8-PSK constellation.

[rxSym,tError] = symbolSync(rxSample); scatterplot(rxSym(1001:end))

Plot the timing error estimate. Over time, the normalized timing error increases to 1 sample.

figure plot(vdelay,tError) xlabel('Time (s)') ylabel('Timing Error (samples)')

The symbol timing synchronizer algorithm is based on a phased lock loop (PLL) algorithm that consists of four components:

A timing error detector (TED)

An interpolator

An interpolation controller

A loop filter

For OQPSK modulation, the in-phase and quadrature signal components are first aligned (as in QPSK modulation) using a state buffer to cache the last half symbol of the previous input. After initial alignment, the remaining synchronization process is the same as for QPSK modulation.

This block diagram shows an example of a timing synchronizer. In the figure, the symbol timing
PLL operates on *x*(*t*), the received sample signal after
matched filtering. The symbol timing PLL outputs the symbol signal, $$x(k{T}_{\text{s}}+\widehat{\tau})$$, after correcting for the clock skew between the transmitter and
receiver.

The symbol timing synchronizer supports non-data-aided TED and decision-directed TED methods. This table shows the timing estimate expressions for the TED method options.

TED Method | Expression |
---|---|

Zero-crossing (decision-directed) | $$e(k)=x\left((k-1/2){T}_{s}+\widehat{\tau}\right)\left[{\widehat{a}}_{0}(k-1)-{\widehat{a}}_{0}(k)\right]+y\left((k-1/2){T}_{s}+\widehat{\tau}\right)\left[{\widehat{a}}_{1}(k-1)-{\widehat{a}}_{1}(k)\right]$$ |

Gardner (non-data-aided) | $$e(k)=x\left((k-1/2){T}_{s}+\widehat{\tau}\right)\left[x\left((k-1){T}_{s}+\widehat{\tau}\right)-x(k{T}_{s}+\widehat{\tau})\right]+y\left((k-1/2){T}_{s}+\widehat{\tau}\right)\left[y\left((k-1){T}_{s}+\widehat{\tau}\right)-y(k{T}_{s}+\widehat{\tau})\right]$$ |

Early-late (non-data-aided) | $$e(k)=x(k{T}_{s}+\widehat{\tau})\left[x\left((k+1/2){T}_{s}+\widehat{\tau}\right)-x\left((k-1/2){T}_{s}+\widehat{\tau}\right)\right]+y(k{T}_{s}+\widehat{\tau})\left[y\left((k+1/2){T}_{s}+\widehat{\tau}\right)-y\left((k-1/2){T}_{s}+\widehat{\tau}\right)\right]$$ |

Mueller-Muller (decision-directed) | $$e(k)={\widehat{a}}_{0}(k-1)x(k{T}_{s}+\widehat{\tau})-{\widehat{a}}_{0}(k)x\left((k-1){T}_{s}+\widehat{\tau}\right)+{\widehat{a}}_{1}(k-1)y(k{T}_{s}+\widehat{\tau})-{\widehat{a}}_{1}(k)y\left((k-1){T}_{s}+\widehat{\tau}\right)$$ |

Non-data-aided TED uses received samples without any knowledge of the transmitted
signal or the results of the channel estimation. Non-data-aided TED is used to
estimate the timing error for signals with modulation schemes that have
constellation points aligned with the in-phase or quadrature axis. Examples of
signals suitable for the Gardner or early-late methods include QPSK-modulated
signals with a zero phase offset that has points at {1+0*i*,
0+1*i*, -1+0*i*, 0−1*i*}
and BPSK-modulated signals with a zero phase offset.

**Gardner method**— The Gardner method is a non-data-aided feedback method that is independent of carrier phase recovery. It is used for baseband systems and modulated carrier systems. More specifically, this method is used for systems that use a linear modulation type with Nyquist pulses that have an excess bandwidth between approximately 40% and 100%. Examples include systems that use PAM, PSK, QAM, or OQPSK modulation and that shape the signal using raised cosine filters whose rolloff factor is between 0.4 and 1. In the presence of noise, the performance of this timing recovery method improves as the excess bandwidth increases (or rolloff factor increases in the case of a raised cosine filter). The Gardner method is similar to the early-late gate method.**Early-late method**— The early-late method is a non-data-aided feedback method. It is used for systems that use a linear modulation type such as PAM, PSK, QAM, or OQPSK modulation. For example, systems using a raised cosine filter with Nyquist pulses. In the presence of noise, the performance of this timing recovery method improves as the excess bandwidth of the pulse increases (or rolloff factor increases in the case of a raised cosine filter).

The early-late method is similar to the Gardner method. The Gardner method performs better in systems with high SNR values because it has lower self noise than the early-late method.

Decision-directed TED uses the `sign`

function to estimate the
in-phase and quadrature components of received samples, which result in lower
computational complexity than non-data-aided TED.

**Zero-crossing method**— The zero-crossing method is a decision-directed technique that requires 2 samples per symbol at the input to the synchronizer. It is used in low-SNR conditions for all values of excess bandwidth and in moderate-SNR conditions for moderate excess bandwidth factors in the approximate range [0.4, 0.6].**Mueller-Muller method**— The Mueller-Muller method is a decision-directed feedback method that requires prior recovery of the carrier phase. When the input signal has Nyquist pulses (for example, when using a raised cosine filter), the Mueller-Muller method has no self noise. For narrowband signaling in the presence of noise, the performance of the Mueller-Muller method improves as the excess bandwidth factor of the pulse decreases.

Because the decision-directed methods (zero-crossing and Mueller-Muller) estimate
timing error based on the sign of the in-phase and quadrature components of signals
passed to the synchronizer, they are not recommended for constellations that have
points with either a zero in-phase or a quadrature component. $$x(k{T}_{\text{s}}+\widehat{\tau})$$ and $$y(k{T}_{\text{s}}+\widehat{\tau})$$ are the in-phase and quadrature components of the input signals to
the timing error detector, where $$\widehat{\tau}$$ is the estimated timing error. The Mueller-Muller method
coefficients $${\widehat{a}}_{0}(k)$$ and $${\widehat{a}}_{1}(k)$$ are the estimates of $$x(k{T}_{\text{s}}+\widehat{\tau})$$ and $$y(k{T}_{\text{s}}+\widehat{\tau})$$. The timing estimates are made by applying the `sign`

function to the in-phase
and quadrature components and are used for only the decision-directed TED
methods.

The time delay is estimated from the fixed-rate samples of the matched filter, which are asynchronous with the symbol rate. Because the resulting samples are not aligned with the symbol boundaries, an interpolator is used to "move" the samples. Because the time delay is unknown, the interpolator must be adaptive. Moreover, because the interpolant is a linear combination of the available samples, it can be thought of as the output of a filter.

The interpolator uses a piecewise parabolic interpolator with a Farrow structure and
coefficient *α* set to 1/2 (see Rice,
Michael, *Digital Communications: A Discrete-Time
Approach*).

Interpolation control provides the interpolator with the basepoint index and fractional interval. The basepoint index is the sample index nearest to the interpolant. The fractional interval is the ratio of the time between the interpolant and its basepoint index and the interpolation interval.

Interpolation is performed for every sample, and a strobe signal is used to determine if the interpolant is output. The synchronizer uses a modulo-1 counter interpolation control to provide the strobe and the fractional interval for use with the interpolator.

The synchronizer uses a proportional-plus integrator (PI) loop filter. The proportional gain,
*K*_{1}, and the integrator gain,
*K*_{2}, are calculated by

$${K}_{1}=\frac{-4\zeta \theta}{\left(1+2\zeta \theta +{\theta}^{2}\right){K}_{p}}$$

and

$${K}_{2}=\frac{-4{\theta}^{2}}{\left(1+2\zeta \theta +{\theta}^{2}\right){K}_{p}}\text{\hspace{0.17em}}.$$

The interim term, *θ*, is given by

$$\theta =\frac{{\scriptscriptstyle \frac{{B}_{n}{T}_{\text{s}}}{N}}}{\zeta +{\scriptscriptstyle \frac{1}{4\zeta}}}\text{\hspace{0.17em}},$$

where:

*N*is the number of samples per symbol.*ζ*is the damping factor.*B*_{n}*T*_{s}is the normalized loop bandwidth.*K*_{p}is the detector gain.

[1] Rice, Michael. *Digital Communications: A Discrete-Time Approach*. Upper Saddle River, NJ: Prentice Hall, 2008.

[2] Mengali, Umberto and Aldo N. D’Andrea. *Synchronization Techniques for Digital Receivers.* New York: Plenum Press, 1997.

Generate C and C++ code using MATLAB® Coder™.

Usage notes and limitations:

See System Objects in MATLAB Code Generation (MATLAB Coder).

Tiene una versión modificada de este ejemplo. ¿Desea abrir este ejemplo con sus modificaciones?

Ha hecho clic en un enlace que corresponde a este comando de MATLAB:

Ejecute el comando introduciéndolo en la ventana de comandos de MATLAB. Los navegadores web no admiten comandos de MATLAB.

Choose a web site to get translated content where available and see local events and offers. Based on your location, we recommend that you select: .

Select web siteYou can also select a web site from the following list:

Select the China site (in Chinese or English) for best site performance. Other MathWorks country sites are not optimized for visits from your location.

- América Latina (Español)
- Canada (English)
- United States (English)

- Belgium (English)
- Denmark (English)
- Deutschland (Deutsch)
- España (Español)
- Finland (English)
- France (Français)
- Ireland (English)
- Italia (Italiano)
- Luxembourg (English)

- Netherlands (English)
- Norway (English)
- Österreich (Deutsch)
- Portugal (English)
- Sweden (English)
- Switzerland
- United Kingdom (English)