It is best not to use the transfer function implementation of discrete (digital) filters, because those are frequently unstable. Use zero-pole-gain and second-order-section implementation instead.
Example —
fc = 100;
fs = 36000;
[z,p,k] = butter(6,fc/(fs/2));
[sos1,g1] = zp2sos(z,p,k);
figure()
freqz(sos1,2^16,fs)
set(subplot(2,1,1),'XLim',[0 200])
set(subplot(2,1,2),'XLim',[0 200])
%%
fc = 10;
fs = 36000;
[z,p,k] = butter(6,fc/(fs/2));
[sos2,g2] = zp2sos(z,p,k);
figure()
freqz(sos2,2^16,fs)
set(subplot(2,1,1),'XLim',[0 200])
set(subplot(2,1,2),'XLim',[0 200])
I also made a few small changes to improve the plots.
signal_filtered1 = filtfilt(sos1,g1,signal);
signal_filtered2 = filtfilt(sos2,g2,signal);
That should produce the correct result.
.
