Passband ripple and stopband attenuation

Hello ,

I want to know how to select the passband ripple for my signal , does it depend on specific standered to be selected ? does it hava anything to do with the fft ?

[Mathieu NOE]

hello

the ripple is a certain amount of amplification or attenuation tolerated in the pass band of the filter. So it depends if those effects are critical for your application or not.

if you are doing a FFT and the amplitude of the spectrum is key, then a low or no ripple low pass filter is required.

If the filter is used to remove some high frequency noise in an industrial monitoring or closed loop system, that may not have any critical effect

[Ali]

How does changing the passband and stopband attenuation affect the filter's performance?

[Star Strider]

@Ali — If the passband and stopband attenuation values differ significantly, this can result in a very steep transition region, and the filter can exhibit undesirable characteristics. Reducing the difference in the passband and stopband attenuation values can improve filter performance in this instance. Other than that, the only effects are to change the amplitudes of the signal components in the passband and stopband.

Answer 1

[Star Strider]

Passband ripple (derived from elliptic and Chebychev filters and not strictily applicable to all filter designs, such as Butterworth) is generally set at 1 dB. Stopband ripple or attenuation can be whatever you want it to be (within limits), with usual values being between 50 dB and 100 dB. Note that getting the [z,p,k] outputs of the filter design functions and then using the zp2sos function generally results in a stable filter.

[Omailnoth]

Hello @Star Strider I still have a question, If my signal has avery high amplitued in the very beginning so at the very end , is it ok to remove it directly ? I'm trying to use pan-tompkin's algorethim to detect the qrs , but when I try to square and diffrentiate the signal this high amplitued affects the amplitued of the other signals..

[Star Strider]

I have no idea what is producing those spikes.

It is necessary to use the filtfilt function to do the actual filtering. If the spikes are still present after using filtfilt, it would likely be best to remove them by thresholding or using findpeaks or islocalmax to detect them so they can be removed.

[Omailnoth]

Sorry I atached the wrong figure,,,

if I just removed the data which has more than (30) amplitued , would that effect negativilly on the signal ? I already did that , because these spikes much larger than the desiered signal,(around 200) . I just set them to zero, and here is the resulted figure ...THANK YOU@Star Strider

[Star Strider]

Since the ‘spikes’ are by definition high-frequency noise, use an elliptic bandpass filter with a 1 Hz to 45 Hz passband to filter the EKG signal. That may eliminate the spikes.

[Omailnoth]

Hello @Star Strider could you give me your opinion about this ,, I have extracted the Q,R,S,P,T but still not sure whether I got them correctly or not , this is the figure ,and here is the code I used after I got the R-peaks

for i=1:length(locs)
   [Q_peak(i), Q_locs(i)] = min(ecg_res_fil(locs(i)-numOfsam_tolook:locs(i)));%Getting the min peak before the R-peak
   [S_peak(i), S_locs(i)] = min(ecg_res_fil(locs(i):locs(i)+numOfsam_tolook));%Getting the min peak after the R-peak
    Q_loc(i) = locs(i)-numOfsam_tolook+Q_locs(i)-1;
    S_loc(i) = locs(i)+S_locs(i)-1;  
   [P(i),P_locs(i)] = max(ecg_res_fil(locs(i)-samples_tolook));
   [T_peaks(i),T_locs(i)] = max(ecg_res_fil(S_loc(i):(S_loc(i)+Tsamples_tolook)));
    P_loc(i) = locs(i) - P_locs(i)-samples_tolook-1;
    T_loc(i) =  S_loc(i)+T_locs(i)-1;

end

after that I wanted to get the PR ,QT ,ST intervals , depending on there definition I used this code

RR_R = diff(locs)/Fs;  % RR interval (seconds)
QRS_R = (S_loc - Q_loc)/Fs; % QRS width (seconds)
PR_R = (Q_loc - P_loc)/Fs;   % PR interval 

Is this right ?? THANK YOU

[Star Strider]

That appears to be correct to me. (For the record, I am a Board-Certified Internal Medicine physician, now retired.)