In analog filter design, I have a problem using IFFT with cheb2ap function.

I'd like to design chebyshev type2(as known as inverse chebyshev filter) by using 'cheb2ap'.

 t = -100:0.001:100;
 dt = t(2)-t(1);
 f = linspace(-1./(2.*dt),1./(2.*dt),length(t));

 N=input('Enter the value of the N: ');
 [z,p,k] = cheb2ap(N,50);
 [num,den] = zp2tf(z,p,k);
 f_cutoff = 0.06;
 w_c = 2*pi*f_cutoff;
 [num,den] = lp2lp(num,den,w_c);
 w = 2.*pi.*f;
 H = freqs(num,den,w);
 irf_chebyshev2 = abs(ifft(ifftshift(H)));
 irf = irf_chebyshev2./max(irf_chebyshev2);
 figure(1);
 plot(t,irf);
 grid;

I choose odd number N, irf is reasonalbe form. but N is even, irf is unreasonable form.

this is N=5

and this is N=6

same problem happen to 'ellipap' function. how can i do to fix this problem?

Answer 1

[Star Strider]

You may be able to get a better result from the freqs function by increasing the number of frequency points it calculates. (The default is 200.)

From the documentation:

[h,w] = freqs(b,a,n) uses n frequency points to compute the frequency response, h, where n is a real, scalar value. The frequency vector w is auto-generated and has length n.

I would start with:

n = 2^12;