Error: Index exceeds matrix dimensions. Error in Edesign8_2 (line 21) plot(real(​z(i)),imag​(z(i)),'o'​)

I cannot for the life of me figure out why my Matlab code keeps breaking. For some reason when the stopband frequency Ws exceeds 1.2, the code breaks due to the extra pole.

%Using Matlab for lowpass Elliptic filter design

close all; clear all;

Wp=1; %passband frequency

Ws=1.2; %stopband frequency

Rp=1; %maximum attenuation in passband (in dB), alpha_max

Rs=30; %minimum attenuation in stopband (in dB), alpha_min

[n,Wn]=ellipord(Wp,Ws,Rp,Rs,'s') %n is filter order, Wn is cutoff

% frequency, 's' denotes an analog filter

% denormalize to Ws for InvCheby

[b,a]=ellip(n,Rp,Rs,1,'s') %b is coefficients of numerator, a is coefficients of

% denominator where 1st entry is highorder

% coefficient. This is the transfer function of a

% normalized inverse Cheby filter.

% Pass the stopband attenuation Rs for InvCheby

sys=tf(b,a) %calculates the transfer function of the normalized butterworth filter

bode(sys) %plots the bode plot of the transfer function

[z,p,k]=ellip(n,Rp,Rs,Wp,'s') %returns zeros (z) and poles (p) and gain (k) of filter

figure; hold on;

for i=1:n

plot(real(p(i)),imag(p(i)),'x')

plot(real(z(i)),imag(z(i)),'o')

end

%axis([-1 1 -1 1])

%calculate stages from poles

poles=p(1);

zeros=z(1);

for i=3:2:n%n-mod(n,2)

poles=[poles,p(i)];

zeros=[zeros,z(i)];

end

%poles/zeros matching

d11=abs(zeros(1)-poles(1))+abs(zeros(2)-poles(2))

d12=abs(zeros(1)-poles(2))+abs(zeros(2)-poles(1))

% for j=1:n/2

% d(i+j-1)=abs(poles(i)-zeros(i))+abs(imag(poles(i)))

w0=sqrt(real(poles).^2+imag(poles).^2)

Q=-w0./(2*real(poles))

for i=1:(n-mod(n,2))/2

den(i,:)=[1 w0(i)/Q(i) w0(i)^2];

num(i,:)=[1 0 abs(zeros(i))^2];

end

if mod(n,2) %odd order polynomial

den(i+1,:)=[0 1 w0(i+1)];

num(i+1,:)=[0 0 1];

end

%order the stages

[Qs,ind]=sort(Q);

w0s=w0(ind);

nums=num(ind,:);

dens=den(ind,:);

for i=1:(n+mod(n,2))/2

t(i)=tf(nums(i,:),dens(i,:));

end

%K=(dens(1,3)*dens(2,3))/(nums(1,3)*nums(2,3))

K=k

M2=1/K

[m,w]=bode(t(1));

M1=max(m)

k1=K*M2/M1

k2=M1/M2

Answer 1

[Ishu]

Hi Jerrod,

I understand that the issue you're encountering with the code breaking when the stopband frequency "Ws" exceeds 1.2 is due to the calculation and handling of poles and zeros.

Like for "ellipord" function where the filter order and the distribution of poles and zeros can significantly change with the specifications. Elliptic filters can have complex pole-zero placements and can generate different number of poles and zeros. As your "n" and "p" are 5 but the value of "z" is 4, so when you plot the real and imaginary part of poles and zeros, the loop breaks and you get the error.

In order to plot the poles and zeros when the number of poles and zeros might not match exactly, you can refer to the below approach:

figure; hold on; grid on;
plot(real(p), imag(p), 'x', 'MarkerSize', 10);  % Poles
plot(real(z), imag(z), 'o', 'MarkerSize', 10);  % Zeros

For more information on "ellipord" you can refer the below MathWorks documentation;

https://www.mathworks.com/help/signal/ref/ellipord.html

[Jerrod]

Thanks for your help. The code now breaks on line 27, 31, 39, and 43 (and 44) with the errors:

27:

Index exceeds matrix dimensions.

Error in Edesign8_2 (line 27)

zeros=[zeros,z(i)];

31:

Index exceeds matrix dimensions.

Error in Edesign8_2 (line 31)

d11=abs(zeros(1)-poles(1))+abs(zeros(2)-poles(2))

39:

Index exceeds matrix dimensions.

Error in Edesign8_2 (line 39)

den(i,:)=[1 w0(i)/Q(i) w0(i)^2];

43:

Subscript indices must either be real positive integers or

logicals.

Error in Edesign8_2 (line 43)

den(i+1,:)=[0 1 w0(i+1)];

num(i+1,:)=[0 0 1];(line 44)

Any help you could provide with these as well would be much appreciated. I'm at my wits end.