Should the filter frequency response contain magnitude greater than 3dB in magnitude?

Should the filter frequency response contain magnitude greater than 3dB in magnitude at why axis?

Below is the frequency response of filter. As one can see that the magnitude is in power of 10^-6. IS this design correct?

Is it possible to have low magnitude?

Can one conclude the roll off?

Please clarify my doubts

Answer 1

[Chunru]

if you have 10^-6 dB or lower, then the response is almost flat in amplitude. The design or something else is most likely wrong if you are not design a all-pass filter. To further check the problem, you may want to post the code to produce the plot.

[632541]

Thank you !

In magnitude resonse of filter 3dB has to be there?

Need to do the design for HPF for 5Hz cutoff. Is it possible ?

[Chunru]

Yes. It is definitely possible to design HPF for 5Hz cut-off. But the filter design require more specifications.

[632541]

Sampling frequeny is 50KHz , These all are the parameters i have.

Can you please provide matlab code ..

Still need more specification for design ?

[Chunru]

% Choose filter type: ellip
% Filter order: 12
% Passband Ripple: 1dB
% Stopband Atten: 50dB
fs = 50000;
fc = 5;     % This is very low
[z,p,k] = ellip(12,1,50, fc/(fs/2),'high');
% Convert to 2nd order sections
sos = zp2sos(z,p,k);

freqz(sos, 8192*16, fs)         % Need very large FFT length 8192*16
h = gcf;
set(h.Children, 'XLim', [0 100]);  % zoom in to 0-100Hz

[632541]

Thank you so much..

%Passband Ripple: 1dB

% Stopband Atten: 50dB

You have assumed these two?

Order can be reduced futher to get less number of coefficients ?

Can this be done for butterworth?

[Chunru]

Those parameters can be adjusted. Try it out yourself. Order can be reduced (or estimated). You can try some small values and oberve the transition band around fc. "doc ellip" for more details.

Sure you can use other types of filter. But for a very small value of fc/(fs/2), ellip may result in sharper transition band.