Simulink^{®} models can process both discrete-time and continuous-time signals. Models built with DSP System Toolbox™ software are often intended to process discrete-time signals only. A discrete-time signal is a sequence of values that correspond to particular instants in time. The time instants at which the signal is defined are the signal's sample times, and the associated signal values are the signal's samples. Traditionally, a discrete-time signal is considered to be undefined at points in time between the sample times. For a periodically sampled signal, the equal interval between any pair of consecutive sample times is the signal's sample period, T_{s}. The sample rate, F_{s}, is the reciprocal of the sample period, or 1/T_{s}. The sample rate is the number of samples in the signal per second.
The 7.5-second triangle wave segment below has a sample period of 0.5 second, and sample times of 0.0, 0.5, 1.0, 1.5, ...,7.5. The sample rate of the sequence is therefore 1/0.5, or 2 Hz.
A number of different terms are used to describe the characteristics of discrete-time signals found in Simulink models. These terms, which are listed in the following table, are frequently used to describe the way that various blocks operate on sample-based and frame-based signals.
Term | Symbol | Units | Notes |
---|---|---|---|
Sample period | T_{s }T_{si }T_{so} | Seconds | The time interval between consecutive samples in a sequence, as the input to a block (T_{si}) or the output from a block (T_{so}). |
Frame period | T_{f }T_{fi }T_{fo} | Seconds | The time interval between consecutive frames in a sequence, as the input to a block (T_{fi}) or the output from a block (T_{fo}). |
Signal period | T | Seconds | The time elapsed during a single repetition of a periodic signal. |
Sample frequency | F_{s} | Hz (samples per second) | The number of samples per unit time, F_{s = }1/T_{s}. |
Frequency | f | Hz (cycles per second) | The number of repetitions per unit time of a periodic signal or signal component, f = 1/T. |
Nyquist rate | Hz (cycles per second) | The minimum sample rate that avoids aliasing, usually twice the highest frequency in the signal being sampled. | |
Nyquist frequency | f_{nyq} | Hz (cycles per second) | Half the Nyquist rate. |
Normalized frequency | f_{n} | Two cycles per sample | Frequency (linear) of a periodic signal normalized to half the sample rate, f_{n} = ω/π = 2f/F_{s}. |
Angular frequency | Ω | Radians per second | Frequency of a periodic signal in angular units, Ω = 2πf. |
Digital (normalized angular) frequency | ω | Radians per sample | Frequency (angular) of a periodic signal normalized to the sample rate, ω = Ω/F_{s} = πf_{n}. |
In the Block Parameters dialog boxes, the term sample time is used to refer to the sample period, T_{s}. For example, the Sample time parameter in the Signal From Workspace block specifies the imported signal's sample period.
Simulink allows you to select from several different simulation solver algorithms. You can access these solver algorithms from a Simulink model:
In the Modeling tab, click Model Settings. The Configuration Parameters dialog box opens.
The selections you make in the Solver pane determine how discrete-time signals are processed in Simulink. The recommended Solver settings for signal processing simulations are
Type: Fixed-step
Solver: Discrete (no
continuous states)
Fixed step size (fundamental sample
time):
auto
Treat each discrete rate as a separate
task: Off
You can automatically set the above solver options for all new models by using DSP Simulink model templates. For more information, see Configure the Simulink Environment for Signal Processing Models.
In the fixed-step, single-tasking mode, discrete-time signals differ from the prototype described in Time and Frequency Terminology by remaining defined between sample times. For example, the representation of the discrete-time triangle wave looks like this.
The above signal's value at t=3.112 seconds is the same as the signal's value at t=3 seconds. In the fixed-step, single-tasking mode, a signal's sample times are the instants where the signal is allowed to change values, rather than where the signal is defined. Between the sample times, the signal takes on the value at the previous sample time.
As a result, in the fixed-step, single-tasking mode, Simulink permits cross-rate operations such as the addition of two signals of different rates. This is explained further in Cross-Rate Operations.
It is useful to know how the other solver options available in Simulink affect discrete-time signals. In particular, you should be aware of the properties of discrete-time signals under the following settings:
When the fixed-step, multitasking solver is selected, discrete signals in Simulink are undefined between sample times. Simulink generates an error when operations attempt to reference the undefined region of a signal, as, for example, when signals with different sample rates are added.
When the Variable-step
solver is selected, discrete
time signals remain defined between sample times, just as in the fixed-step,
single-tasking case described in Recommended Settings for Discrete-Time Simulations. When the
Variable-step
solver is selected, cross-rate
operations are allowed by Simulink.
SeeSimulink Tasking Mode for a description of the criteria that Simulink uses to set the tasking mode. For the typical model containing multiple rates, Simulink selects the multitasking mode.
When the fixed-step, multitasking solver is selected, discrete signals in Simulink are undefined between sample times. Therefore, to perform cross-rate operations like the addition of two signals with different sample rates, you must convert the two signals to a common sample rate. Several blocks in the Signal Operations and Multirate Filters libraries can accomplish this task. See Convert Sample and Frame Rates in Simulink for more information. Rate change can happen implicitly, depending on diagnostic settings. See Multitask rate transition (Simulink), Single task rate transition (Simulink). However, this is not recommended. By requiring explicit rate conversions for cross-rate operations in discrete mode, Simulink helps you to identify sample rate conversion issues early in the design process.
When the Variable-step
solver or fixed-step,
single-tasking solver is selected, discrete time signals remain defined
between sample times. Therefore, if you sample the signal with a rate or
phase that is different from the signal's own rate and phase, you will
still measure meaningful values:
At the MATLAB^{®} command line, type ex_sum_tut1
.
The Cross-Rate Sum Example model opens. This model sums two signals with different sample periods.
Double-click the upper Signal From Workspace block. The Block Parameters: Signal From Workspace dialog box opens.
Set the Sample time
parameter to 1
.
This creates a fast signal, (T_{s}=1), with sample times 1, 2, 3, ...
Double-click the lower Signal From Workspace block
Set the Sample time
parameter to 2
.
This creates a slow signal, (T_{s}=2), with sample times 1, 3, 5, ...
In the Debug tab, select Information Overlays > Colors.
Selecting Colors allows you to see the different sampling rates in action. For more information about the color coding of the sample times see View Sample Time Information (Simulink).
Run the model.
Using the DSP Simulink model templates with cross-rate operations
generates errors even though a fixed-step, single-tasking
solver is selected. This is due to the fact that
Single task rate transition is set
to error
in the
Sample Time pane of the
Diagnostics section of the
Configuration Parameters
dialog box.
At the MATLAB command line, type
dsp_examples_yout
.
The following output is displayed:
dsp_examples_yout = 1 1 2 2 1 3 3 2 5 4 2 6 5 3 8 6 3 9 7 4 11 8 4 12 9 5 14 10 5 15 0 6 6
The first column of the matrix is the fast signal, (T_{s}=1). The second column of the matrix is the slow signal (T_{s}=2). The third column is the sum of the two signals. As expected, the slow signal changes once every 2 seconds, half as often as the fast signal. Nevertheless, the slow signal is defined at every moment because Simulink holds the previous value of the slower signal during time instances that the block doesn't run.
In general, for Variable-step
and the
fixed-step, single-tasking modes, when you measure the value of a discrete
signal between sample times, you are observing the value of the signal at
the previous sample time.