From PrattWiki
Jump to: navigation, search

The creators of MATLAB understand that block diagrams are used to visually depict complicated systems, and also that the mathematical characteristics of those systems, once defined, can be used to determine the values for the various signals throughout the system. For that reason, they developed the Simulink package as an add-on to MATLAB. Simulink allows you to draw your system as if you were drawing a block diagram, and then to simulate the system using a variety of computational methods, usually with some component of integration involved.

This introduction will cover starting Simulink, understanding the basic blocks and connections, and running simulations. It will then move on to creating arbitrary functions, transferring data between the model and the workspace, and printing models. By the end of this introduction, you should be able to generate a wide variety of signals, perform basic and intermediate mathematical operations, visualize and save data, control models with a script (not completed yet), and print your model.

Note - figures are numbered by section. I haven't figured out a better way to do this...

Starting Simulink

Simulink is an add-on package to MATLAB, and OIT has purchased several licenses for use on the Linux workstations. To start Simulink, you must first start MATLAB (by typing matlab &), then type simulink at the command prompt within MATLAB. At that point, the Simulink library window will come up. While you will not be creating your model in this window, you will be using it to obtain access to the many different built-in blocks that Simulink provides.

To start your own model, go to File->New->Model. Now you have a blank canvas to work with. To fill it in, you need to know what the basic blocks are, how to bring them into your model, and how to connect them.

Basic Blocks and Connections

Figure 1

The Simulink library is broken up into several categories, and each category contains several building blocks for modeling signals and systems. To look at the contents of a particular category, just double-click its name in the library window. For example, if you double-click on the Sources icon, you will open a new library window containing several different ways that MATLAB can generate a signal. If you want to add one of these to your model, all you have to do is click and hold on it with the left mouse button, then drag it onto your canvas. Try this with the Signal Generator block. Your model should now have a single Signal Generator (fig. 1).

To examine and change the possible parameters of a block, simply double-click on it in your model window. The Signal Generator, for example, allows you to choose what wave form you would like, the time base to use, the amplitude of the signal, the frequency, and what units of frequency to use. If you were to want to use an input of x(t)=3\sin(8t), for example, you would choose a sine wave, using the simulation time, with an amplitude of 3 and a frequency of 8 radians per second. Go ahead and make those changes, then click the OK button to accept them.

Figure 2

Once you have some kind of input signal, you may want to process it. There are many different ways to process a signal in Simulink, ranging from basic mathematics to advanced image processing and frequency analysis. For now, open up the Math Operations category by double-clicking it in the original library window. Among the different operations you can perform is the Gain block, which acts like a multiplier. Drag a copy of this into your model, so that now you have a Signal Generator and a Gain block. Move the Gain block so it is about an inch to the right of the Signal Generator (fig. 2).

Figure 3

If you double-click this block, you will see that it has more options than you may have thought for a simple gain. Generally, however, you will only need to change the Gain itself. Go ahead and set the gain to 2 and click the Apply button. Notice that the number inside the block itself changes as a result (fig. 3).

Figure 4
Figure 5

The number in the gain block will reflect the actual gain if there is room for it - otherwise the gain block will have the letter K in it. For example, if you put in 0.0001 for the gain and hit Apply again, the block will just have a K in it (fig. 4). If you really want to see the gain, you can change the size of the block to make it large enough that the number will fit inside (fig. 5) though generally you will not want to resize blocks just to see the contents. The main blocks you will be resizing are those with multiple inputs or outputs, for example, summation blocks there the larger size will make it easier to track and attach the different inputs. In this case, change the gain back to 2 and hit OK, then make the gain block a more normal size.

Figure 6

Finally, you will want to have some way of viewing or saving the data. If you go back to the original Library window and open up the Sinks category, you will see the various ways that MATLAB can take a signal and either send the information to the screen, to the workspace, or to a file. In this case, you are going to want to both view and manipulate the data, so drag both a Scope and a To Workspace to your model. Put both about an inch to the right of the Gain block, with the Scope about an inch above and the To Workspace about an inch below (fig. 6).

If you double-click the scope, an oscilloscope window will appear. The scope is useful for viewing the signals throughout a simulation in real time. Generally, you will add these during the construction of a model, but once you have completed the model, you will want a more permanent record of the data obtained. This is where the To Workspace block comes in.

Figure 7

The To Workspace block will record the signal values sent to it as a variable in the MATLAB workspace. If you double-click on the block, you will note that you can change the variable name - go ahead and make this y. The Save format that is most readily used is Array, so go ahead and make that change as well. The default cases for the other parameters will generally work fine. Once done, click OK and notice that the name of the variable is now showing inside of the block (fig. 7).

Figure 8

Now that the blocks are in place, you need to connect them. You will notice if you run the cursor over the small port symbol (the >) of a block that the cursor changes from an arrow to crosshairs. The way you connect blocks is to pick the port of one block, hold down the left mouse button, and drag until you hit another port (or another wire). Go ahead and connect the Signal Generator to the Gain block by putting the cursor on top of the output port of the Signal Generator, holding down the left mouse button, and dragging until the cursor is over the input port of the Gain block. You will notice that the cursor changes to double crosshairs when you have found a terminus for your wire. Now connect the output of the Gain block to the Scope (fig. 8).

If you also want to connect the Gain block to the To Workspace block, you will need to first click on the input of the To Workspace then drag until you hit the wire connecting the scope to the gain block. You cannot go back to the Gain block's output port, because every input or output port can only be used once.

Since every wire should have at least one new port connection, you must begin from that port. Again, you will get the double crosshairs when you are over a wire such that a connection will be made and you will notice a node symbol (small black dot) at the connection. If you make a mistake and drop the wire before it is connected, you will get a dashed red wire with a port symbol at the end of it. You can remove this wire by left clicking on it (thereby selecting it) and hitting delete.

Figure 9

At this point, your model should have all four blocks properly wired, with a gain of 2 and a To Workspace variable called y (fig. 9). There should also be an open Scope window that came about when your double-clicked the Scope block.

Running Simulations

Fig 1: Plot with no limit on maximum step size
Fig 2: Plot with limit of 0.01 on maximum step size

Now that you have a basic block diagram with an input signal going through a system that multiplies that signal by 2 as well as two different ways of examining the data, you are ready to simulate the system. Within the Simulation menu of your model, go to Configuration Parameters and you will discover that Simulink has an extensive array of options from which to choose. For now, the critical part is telling MATLAB to store all the data and telling MATLAB the simulation time.

Go to the Data Import/Export selection and un-check the Limit data points... option in the Save options. For whatever reason, MATLAB defaults on only storing the last thousand data points unless you specify otherwise. Generally, you will want Simulink to report all the data.

With respect to timing, go to the Solver selection. The default case - which goes from 0 to 10 seconds - might work, so go ahead and hit OK. Then, to run the simulation, go to Simulation and Start. The scope will show the values of the output, and when the simulation is finished, MATLAB will beep.

If you type whos in MATLAB, you will discover that not only did you generate an array called y but that Simulink also created an array called tout for you. Now you can run the simulation and, when finished, issue the command

plot(tout, y)

in MATLAB's command window to plot the signal - you should get fig. 1.

When you do this, you will notice something very disturbing - the output signal does not look like a sine wave a 8 radians per second. This is because Simulink, while processing your model, tries to solve as quickly as possible by taking as large a time step as possible. For this particular model, allowing it to do that has produces a somewhat disastrous result - the locations at which data points were taken, if connected using lines, do not accurately represent the signal. To forestall this, you can go back into the Simulation menu's Configuration Parameters and make some changes. Specifically, within the Solver selection change the Max step size to a reasonable value for the speed of the signal - perhaps 0.01. Re-run the simulation, and re-plot your results to get fig. 2.

Creating Arbitrary Inputs

Fig. 1: After deleting blocks
Fig. 2: After adding new blocks
Fig. 3: Plot of new output

There are several ways to create different inputs signals. The Signal Generator is certainly the easiest way to create a basic sinusoid, square wave, or sawtooth. For other signals, you can use a combination of sources and basic math to get what you want. There is even a Signal Builder block that can be used to draw signals. Generally, however, the easiest way to create an input is to use a Clock block and feed it into a Fcn block. Go ahead and save your current model as SDDemo1.mdl, then do a save as and call it SDDemo2.mdl.

Now select and delete both the Signal Generator and the Gain block. Among other changes, you will now be left with unconnected lines, denoted by red dotted lines (fig 1). These can be deleted by selecting them and hitting delete, or they can be reused by putting a block's input or output ports close enough that Simulink thinks they should be connected. For example, drag a Clock block from the Sources category such that the output of the Clock is near the left-most part of the left line. It will attach itself to that wire, though the wire will remain red since its output end is still unconnected.

Now open the User-Defined Functions category and drag a Fcn block into your model, placing it in the space vacated by the gain block. The Fcn block is larger that the Gain block so you can only snag one of the wires. To get the other one, you can drag the Fcn block again or use the arrow keys to move it until it connects and the wire turns solid black. You should now have the model in fig 2.

To enter a function, double-click the Fcn block. Note that the expression is currently a function of the vector u. This vector represents the different inputs to the block. Right now, we are only using a single input - though later you can use multiplexers to send multiple streams along a single line. Your function, therefore, should be written in terms of u[1]. For example, if you want the output of the function to be e^{-t/3}\cos(4t), then the input to this block should be exp(-u[1]/3)*cos(4*u[1]). Note - Simulink does not like using the .* version of multiplication.

Once you have that function entered, re-run the simulation. If you make a plot of y as a function of tout using the command plot(tout, y), you should get the plot in fig. 3.

Transferring Data

As mentioned earlier, you will often want to transfer data from the Simulink simulation to the workspace. The To Workspace block is the easiest way to get a signal's information to the workspace. Just make sure that the variable name is unique for each of the blocks and that you have used the Array method for storing the data. There are also From Workspace blocks if you want to pass in value and time data from the workspace.

Naming and Arranging Blocks

You can change the name of blocks in Simulink by double-clicking the name you want to change. This can be very useful if your blocks have real-world analogs. You can also probably come up with much more interesting names than To Workspace for your outputs. Avoid using dots or slashes in the block names, however. Stick with letters and numbers if at all possible. This will become more important in the next section.

You may also occasionally want to aim a block in a different direction - to do this, simply pick the block, then go to the Format menu and select either Flip Block or Rotate Block. You have already seen that you can resize blocks. Take note the other options in the Format menu - there may be times with particularly sophisticated models that you will want to use various colors and other options to help clarify your system.

Running Models With Scripts

This section is under construction and not required for EGR 224 Spring 2010

Saving and Printing Models

To save a Simulink model, just go to the File menu and select either Save or Save As. Simulink models should end with .mdl. To print your model, you can either use the Print function in the File menu or type

print -deps -sMODELNAME FILENAME.eps

The following code will not only print the model based on the filename, it will also save and close the model window when finished:

Filename = 'StringGoesHere'
eval(sprintf('print -s%s -deps %smodel', Filename, Filename))

The name of the saved picture of the model will be Filenamemodel.eps. If you are going to do several things with the same model, you may want to add some kind of flag in the sprintf command to differentiate among plots. For example, if you are running code in a loop based on some integer counter k, you could use the following code:

eval(sprintf('print -s%s -deps %smodel_%0.0f', Filename, Filename, k))

Sample Codes and Models

Class Document Protection