how to differentiation code,?

i want to wirte x*(1+s) into matlab function block in c code

wher x is input signal and s - is s laplace i.e. use for differentation in matlab

please help

Answer 1

[Benjamin Thompson]

If you have x as sampled signal in time, differentiation is easy: (x(t2) - x(t1)) / (t2 - t1)

[RJS]

Thanks @Benjamin Thompson for your answer but how to perform this into matlab funtion block,

I tried but it's not working

[Benjamin Thompson]

Simulink has blocks for differentiating signals. If you are using a continuous time signal, use the Derivative block in the Continuous library. If you are using discrete time signals, use the Discrete Derivative block. You could probably use a MATLAB Function Block, but you need to store the previous value of the input signal which takes some work, and you need access to the value of the time step--see the help article "Define MATLAB Function Block Data" for more information on providing data to this type of block in Simulink.

[Walter Roberson]

https://www.mathworks.com/help/simulink/slref/transferfcn.html

Simulink has transfer function blocks

[RJS]

@Walter Roberson @Benjamin Thompson thanks i know how to use this Transfer function block and derivative blocks , but want to write (k*s+1) in c code and then i want to test this code with my simulink plant model using matlab function block

[Walter Roberson]

Simulink Coder can generate code for a Transfer Function block.

Have you considered building the C code as a Level 2 S-Function instead of calling it from a MATLAB Function Block ?

[Benjamin Thompson]

LaPlace notation is used for convenience to work with differential equations algebraically, for analysis or simplification, solving for parameters, etc. If you want a time based numerical solution to a differential or difference equation, you must convert to the time representation. In the case of differentiation that is fairly trivial.

[Walter Roberson]

For any given numeric k, you can construct the numeric tf(), and use c2d() to find the equivalent discrete transfer function, the coefficients of which can be used to filter()

I just spent a while tracing through c2d, as I expected that there might be a relatively simple conversion for this case of 1/(k*s+1) .

Unfortunately, the internal construction of ss form involves a 2^round(log2(sqrt(1/k))) to construct a scaling factor, which makes the conversion a bit more complicated.

I kept looking through anyhow, thinking that at least in theory that might not be necessary, that you might be able to proceed with state space a=-1/k, b = 1, c = 1/k, d = 0, e = [] .

Sadly, inside the routine to discretize from zero order hold, there is a call to the undocumented mex routine matscale

I see hints that maybe you could use

Phi = exp(-Ts/k)
Gamma = k - k*exp(-Ts/k)

and
x[n+1] = Phi * x[n] + Gamma * u[n]

but I am not sure how that would translate into filter() or conv()