Simulate a random walk that starts at a certain time and then stops before the end of the simulation

I want to simulate a white noise that begins at a certain time and then stops at another time before the end of the simulation. The goal is to generate a random-walk disturbance but I want to have a period at the beginning where the disturbance is zero and a final period where the disturbance is static.

Based on this answer I did it using two switches:

Is this the best way or is there something simpler?

Answer 1

[Walter Roberson]

You can use an if action block https://www.mathworks.com/help/simulink/slref/if.html as that permits you to use the & operator. So if time > 300 & time < 1000 ...

Or, use two relational operator blocks https://www.mathworks.com/help/simulink/slref/relationaloperator.html with their output type set as 'double', and multiply the two results, and multiply that by the output of the noise signal: the result will be 0 where either or both of the comparisons is false, and will otherwise be the noise.

[Bill Tubbs]

Thanks very much. It works, although not sure it's any simpler!

[Walter Roberson]

I think using the relation operator blocks and multiplication would probably be easier.

[Bill Tubbs]

For the record, here is the second alternative. Note: I didn't have to set the logical output type to float (it wasn't an option in any case). Simply multiplying the white noise by the boolean worked.

Both the above models produce identicial signals.

[Paul]

I think that this method will produce a new output from the integrator for one time step after t = 1000 (assuming zero crossing detection in use). Everything to the left of the integrator is running for the entire simulation, even when not really needed for t<=300 and t>=1000. If either of these are a concern, consider encapsulating the white noise source and the integrator in an Enabled Subsystem and make the output of the AND block the enabling signal input to the subsystem. Also, use of Compare to Constant can simplify things a bit. The Interval Test block can simplify things even further by combining the upper/lower limit test into a single block. However, it doesn't support zero crossing detection, if that's important.

[Bill Tubbs]

@Paul, I'm okay with the Noise running for the entire simlation but it's good to know that I could put it inside an enabled subsystem to turn it on and off as and when required.

But I'm confused, what's the difference between an Enabled Subsystem and the If Action Subsystem, which Walter suggested?

[Paul]

In this particular case, they are bascially functionally equivalent, except that the band limited white noise streams won't be the same becasue it's running from t = 0 in the If Action case, but only starts running at t > 300 in the Enabled case.

Conceptually, I think of the If Action as a basic if-then-else type of thing in an algorithm, whereas the Enabled Subsystem is used to turn portions of the model on or off, or select between different implementations for the same dynamics, like a high-fidelity or low-fidelity implementation, or model the change in dynamics due to some dynamica condition, like static/kinetic friction. I'm having a difficult time verbalizing my own views and there probably is some (a lot?) overlap between these. Maybe the doc has some discussion on when one should use one type of construct versus another. I'm sure the doc has plenty of examples for all of the types of conditionally executed subsystems.

[Walter Roberson]

I recall that there was a post a few months ago to the effect that the different streams of if/then/else all kept running at the same time, unless you did specific things, and therefore you could end up with an error arising from a grouping that was not selected. I didn't really catch the details, unfortunately.

[Bill Tubbs]

Thanks. That explains it!