coder.excentric not working when compiling with TwinCat Target

Hello,

I am working on a case where I put simulink generated code into an embedded PC. My model works fine in simulink, but when I try to generate the code for my embedded controller, it fails. The fault message is the following.

_Failed to eliminate a call to the MATLAB function 'butter'. For non-simulation builds, calls to unsupported MATLAB functions are eliminated if they do not affect function out puts. Function 'filter1/MATLAB Function' (#34.92.133), line 6, column 1: "[a,b] = butter(w,[x y]/(z/2), 'bandpass')"

I found out that the actual problem is that coder.extrinsic('butter') is not allowed by external compilers. But I really need a way to create a butterworth filter in a matlab function block.Because my system needs a filter with adaptable bandwith. Is there somebody who knows a possible workaround?

     function [a,b] = fcn(w,x,y,z)
      a=zeros(1,(11));
      b=zeros(1,(11));
      coder.extrinsic('butter');

      [a,b] = butter(w,[x y]/(z/2), 'bandpass');

Answer 1

[Tom Bryan]

If you leave out the coder.extrinsic('butter') command, then it will generate code for the embedded processor. But, in that case, the butter() command requires constant inputs.

It is possible to create a table of constant coefficient values that you can index into as a function of discrete w, x, y, z values. Would that be a workable option for you?

[Dries billiet]

Nice idea! I'll try to work this one out!

[Dries billiet]

This does non't work. Because the coeficients are non constants. But it brougt me on the idea of adding a 'database' (read: big array) with the calculated transfert functions already in the array.

Answer 2

[Kaustubha Govind]

You can either implement the functionality yourself in C, or find a third-party library that does so, and call it in your MATLAB code using coder.ceval.

[Ketan]

Here are few other options:

- Implement the desired filter coefficient calculation yourself in MATLAB and call this code in your function block

- Try to leverage some of the more advanced filter functionality in DSP system toolbox:

http://www.mathworks.com/help/coder/ug/functions-supported-for-code-generation--categorical-list.html#bt7ln6w