Re: Problems with Arx-algorithm

[Lukas Reichlin]

On 14.05.2014, at 10:58, Flo89 <address@hidden> wrote:

> Hello,
> 
> I did some measurements and data analysis on a system. Now I want to let
> octave calculate a transfer function with arx algorithm but I'm running into
> problems.
> Here's the code:
> 
> InputData =
> dlmread('/home/flo/Dokumente/Octaveprojects/c_output_frequencies_high_speed.csv','\t');
> input = InputData(:,1);
> input_short = input(10 : 10 : end);
> output = InputData(:,2);
> output_short = output(10 : 10 : end);
> 
> sampletime = 1/(200000)*10;
> t = 0 : sampletime :sampletime*(rows(input_short)-1);
> dat = iddata(output_short, input_short, sampletime);
> 
> [sys, x0] = arx(dat,  'na',i, 'nb', l);
> [y, t, x] = lsim (sys,input_short,t,x0);
> figure
> plot(t,y,'r');
> hold on;
> plot(t,output_short,'g');
> 
> I'm doing this for example in two for loops to find good values for na and
> nb. But if I'm doing na = 6 and nb = 1 to nb = 6 theres always the same
> simulated system. I doubt it will ever chance when changing the parameters. 
> So my question is: What am I doing wrong in general?
> And the second: I knot that the delay between input and output is around 100
> sampletimes. Is nk then 100?
> 
> Here are images of the real system:
> <http://octave.1599824.n4.nabble.com/file/n4664059/real_system.png> 
> And here of the real output and the simulated one:
> <http://octave.1599824.n4.nabble.com/file/n4664059/simulated.png> 
> 
> Thank you!

Hi Florian

Could you please send me the file "c_output_frequencies_high_speed.csv"? I 
would like to run some tests. Note that the "iddata" class has a "resample" 
method (help @iddata/resample). Regarding the second question: Yes, "nk" should 
be 100 in your case.

Lukas