frep2tf

Arguments

frq

vector of frequencies in Hz.

repf

vector of frequency response

dg

degree of linear system

dom

time domain ('c' or 'd' or dt)

tols

a vector of size 3 giving the relative and absolute tolerance and the maximum number of iterations (default values are rtol=1.e-2; atol=1.e-4, N=10).

weight

vector of weights on frequencies

h

SISO transfer function

err

error (for example if dom='c' sum(abs(h(2i*pi*frq) - rep)^2)/size(frq,*))

Description

Frequency response to transfer function conversion. The order of h is a priori given in dg which must be provided. The following linear system is solved in the least square sense.

weight(k)*(n( phi_k) - d(phi_k)*rep_k)=0, k=1,..,n

where phi_k= 2*%i*%pi*frq when dom='c' and phi_k=exp(2*%i*%pi*dom*frq if not. If the weight vector is not given a default penalization is used (when dom='c').

A stable and minimum phase system can be obtained by using function factors.

Examples

s=poly(0,'s');
h=syslin('c',(s-1)/(s^3+5*s+20))
frq=0:0.05:3;
repf=repfreq(h,frq);
clean(frep2tf(frq,repf,3))

Sys=ssrand(1,1,10);
frq=logspace(-3,2,200);
[frq,rep]=repfreq(Sys,frq);  //Frequency response of Sys
[Sys2,err]=frep2tf(frq,rep,10);
Sys2=clean(Sys2) //Sys2 obtained from freq. resp of Sys

[frq,rep2]=repfreq(Sys2,frq); //Frequency response of Sys2
clf();
bode(frq,[rep;rep2])   //Responses of Sys and Sys2
[gsort(spec(Sys('A'))), gsort(roots(Sys2('den')))] //poles

dom = 1/1000; // Sampling time
z = poly(0,'z');
h = syslin(dom,(z^2+0.5)/(z^3+0.1*z^2-0.5*z+0.08))
frq = (0:0.01:0.5)/dom;
repf = repfreq(h,frq);
[Sys2,err] = frep2tf(frq,repf,3,dom);
[frq,rep2] = repfreq(Sys2,frq); //Frequency response of Sys2
clf();
plot2d(frq, abs([repf;rep2])')

See also

• imrep2ss — state-space realization of an impulse response
• arl2 — SISO model realization by L2 transfer approximation
• time_id — SISO least square identification
• armax — armax identification
• frfit — frequency response fit