I have a continuous-time state-space model and I designed in matlab a controller based to this model so yet the whole process is in continuous-time. In Simulink I simulate the model so that there is a sample time Ts (runge kutta). Of course the controller will be digital in the end, so 1. Am I simulating the process in discrete because of sample time? 2. Do zero-hold block enough to simulate desecrating continuous-time state-space model of controller? What I want to do is to convert my controller to desecrate (digital) and simulate it for continues time model…..
Hi Ali,
You may either work in continuous (transfer function/Laplace transform) space or discrete (differential equations in state-space form/Z-transform) space interchangeably. Dedicated transforms allow to switch from one representation model to the other.
In order to address your issue specifically, check for the following links:
Best regards
Change out all the components to digital or install converters digital to analog.
Yes, what I have done is emulation, next step will be to discretize the controller, it is possible in Matlab by using "c2d" function. The other way is to discretize the plant by using "c2d" function and design digital controller. I just found it easy and more practically to do it in emulation way, because of simulating. Ok, lets ask this way, please open the attachment, there should be a basic model. Is it in a blue circle the same thing as c2d(controller state-space model)? if yes then it represents my digital controller which I want to simulate in continuous time process. But there is still what I can not understand, what step-time in Simulink represents? Is it discretizing a whole process and if I set step time as same as zero-order hold block, is the block useless because it represents the same thing? Do I need some D/A after Zero-Order hold block?
Thank you Khan it clarify one step more, how I just forget the step time is related to rungekutta...
Hi Ali,
As people already wrote here, there are two main approaches: 1. Design everything in :the continuous world and then implement the controller in the discrete form, or 2. Discretize the plant and do everything in the discrete world.
Of course, neither one is exact and you can read about hybrid systems, whichtry to take into account the two worlds.
Now we should not forget reality: your plant is continuous and only the controller will be discrete. So, the general practice is 1.: you do the design in the continuous world and then translate the controller only for its discrete implementation. You must be sure that sampling time is sufficiently small for the expected time-constants of your plant. With present day microprocessors this should not be a problem.
To have a realistic simulation, you should use the runge-kuta for the continuous plant part only and use the discrete facilities of SIMULINK for the controller. Still, if you use the discrete simulation for all and it works, I would think that you even have some more safety margin that the rel-world implementation will also work.
Best,
Itzhak
We can sample a continuous plant with a zero-order hold (ZOH). So , we sample a continuous controller. In the book "Computer-controlled systems" by K.J. Astrom and B. Wittenmark , we use an approximation s=(z-1)/(Tz) to convert C(s) to C(z). It is my opinion.
Dear Ali
it is more simple to work in discrete (differential equations in state-space form/Z-transform) space.
you can first discretize your model, then design your desired digital controller.
I suggest Model predictive control.
it is simply implementable and has great performance.
if you are interested I can help you with the coding and consepts.
kind regards
- Badges
- Science topic
- Similar topics
- Engineering
- Control Theory
- Control Systems
- Control Systems Engineering