How to implement an RST controller in Simulink?

Every time I search for help on this I am taken to the following page:

RST Controller

But I don't have access to the Simscape Electrical library!

In any case I would like to implement the equivalent controller using standard Simulink blocks. I'm surprised I haven't found an example or tutorial on this but I did look. If you know of one, please point me to it.

I tried implementing it myself but I'm not sure I did it right since I'm very new to Simulink. Here is my attempt:

Update: I replaced the previous example image with the one above where I am quite confident the polynomials are correct. Since it doesn't produce the desired results (unstable in closed loop even with zero process noise) I suspect there is something wrong with my simulation setup or implementation.

For example, I used the 'Discrete Filter' blocks for T(z), and S(z), and 1/R(z) transfer functions but 'Discrete Transfer Fcn' blocks for the 'B(z)/A(z)' and 'C(z)/D(z)' blocks. Not sure if that was the right choice.

Also I used the 'DiscreteIntegrator' block for the 1/Delta (1/(1-z^-1)) block.

I'm not looking for someone to confirm if my polynomials are correct (but feel free to let me know if you see a problem!). What I'm looking for is an authoritative implementation that I can compare my results to.

[Bill Tubbs]

Okay, well after comparing the output of this model with a hand-coded RST in closed loop with the same process I discovered that the Simulink model above produces identical reslults if I set the discrete delay block in the process to zero timesteps.

I don't understand this as the controller parameters were designed for a process with a delay of 1 timestep (z^-1). Would would it work if I remove this delay? Is there maybe already a delay built in somewhere else?

Any insights appreciated.

Answer 1

[Bill Tubbs]

After some further diagnosis, I think I have solved the problem. The default setting of the Discrete-Time Integrator block is 'Integration: Forward Euler'.

I think the method I need for is 'Backward Euler' and I set it to accumulate, not integrate. From the documentation it seems this is the equivalent of 1/(1-z^-1) or z/(z-1) which is what I want.

Now I get exactly the results expected with the delay set to 1 timestep as it should be.

I guess the learning is don't just copy paste blocks into your model without understanding the defaults!