# Question on Chebyshev filter

What does the output signal look like when you insert $2000 Hz$ square pulses to chebyshev filter with parameters: (Matlab syntax). I tried to do it like this but I don't understand where it goes wrong.

%Making the square pulses

Fs = 1000000;
t = [0:1/Fs:50e-4];
f = 2000;

%square pulse with 50% duty cycle

xt = 1*square(2*pi*f*t, 50);

%Making the filter

Wp = [1000 3000];
Ws = [100 3900];
[n,Wn] = cheb1ord(Wp,Ws,3,20,'s');
[b,a] = cheby1(n,3,Wn,'s');

%Filtering with the coefficient

xt_filtered = filter(b,a,xt);

%Plotting the filtered signal

plot(t,xt_filtered);


But I get something like this which doesn't make sense..? • I think this is a duplicate of dsp.stackexchange.com/a/8420/29 Basically you want to use oversampling to make the digital filter approximate the analog filter more closely, and you want to generate a bandlimited square wave for your input signal (with highest harmonic significantly below the sampling frequency) – endolith Nov 20 '14 at 14:39
• Your plot is wrong because you're generating analog filter coefficients and then using them in a digital filter. you need to remove 's'. You can generate the bandlimited square using additive synthesis: gist.github.com/endolith/407991 – endolith Nov 20 '14 at 21:21
• Thank's for the link! I'll try it when I have more time to get to know it for sure. My matlab script gets error if I lose the 's' from cheb1ord and cheby1 functions. Wp = [1000 3000]; Ws = [100 3900]; [n,Wn] = cheb1ord(Wp,Ws,3,20,'s'); [b,a] = cheby1(n,3,Wn);  What would be the correct syntax then? – ELEC Nov 20 '14 at 22:19
• @ELEC - What error do you get if you remove 's' ? – JRE Nov 21 '14 at 10:58
• Your square wave generator doesn't seem to generate a square wave. When I do "Plot(t,xt)" I get a straight line at level 1.0 (DC signal that stays put at 1.0) This may be a difference between Gnu Octave and Matlab - I will see what Octave needs to generate a square wave. – JRE Nov 21 '14 at 11:01

Allowing for differences between Matlab and Octave ( I only have Gnu Octave to play with,) and correcting the errors noted in the comments, I get the following: Needed fixes:

1. Values for Wp and Ws divided by half of sampling rate - the cheb1ord function requires frequencies be given relative to the Nyquist frequency.
2. Removed 's' (not allowed for Octave anyway.)

To make the code work for Octave, I had to make the following changes:

1. Remove duty cycle value from square
2. Remove 's' from cheb1ord and cheby1

Code for Octave (should run under Matlab as well:)

%Making the square pulses
Fs = 1000000;
nyq=Fs/2;
t = 0:1/Fs:50e-4;
f = 2000;

%square pulse with 50% duty cycle
xt = square(2*pi*f*t);

%Making the filter
Wp = [1000/nyq 3000/nyq];
Ws = [100/nyq 3900/nyq];
[n,Wn] = cheb1ord(Wp,Ws,3,20);
[b,a] = cheby1(n,3,Wn,'s');

%Filtering with the coefficients
xt_filtered = filter(b,a,xt);

%Plotting the filtered signal
subplot (211);
plot(t,xt);
subplot(212);
plot(t,xt_filtered);