pick any 3 random files from this database, : https://physionet.org/pn3/ecgiddb/ They are subject to 50 Hz powerline interference.

We wish to convert from time domain to frequency domain and remove the 50 Hz frequency then transfer back to time domain showing clean signal.

I'm looking for some pointers in how I go about designing a Butterworth low pass filter which will eliminate 50 Hz powerline interference from an ECG.

How do I determine which order of filter, and how I determine the poles and zeros in order to calculate magnitude, phase and filter coefficients. Where do I start?

  • 2
    $\begingroup$ Why would you use a low pass filter to remove a single frequency? And then with the Butterworth one that is maximally flat instead of maximally steep? $\endgroup$ – Marcus Müller May 14 at 15:46
  • 1
    $\begingroup$ (i.e.why did you choose the Butterworth? What's your purpose? Why do you go for a continuous-time filter instead of a discrete-time notch filter, as very sensible for this application?) $\endgroup$ – Marcus Müller May 14 at 16:03
  • $\begingroup$ I am looking to compare a couple of filter design techniques, ie butterworth and pole zero placement and also compare different filter orders, just looking for pointers for starting , then maybe incorporate your suggestions for further comparison $\endgroup$ – Gary Hand May 14 at 16:22
  • $\begingroup$ so, my suggestion then is: Write down an overall system specification. What are you measuring, with what sensors and what processing steps, and why is 50 Hz a problem, exactly. Only from that can stem a decision which filter to use, and how to design it! You're starting with the last part of a long decision chain, and I think somewhere upfront, you've missed the important decisions. So, if you could edit your question to be detailed about your measurement setup, and how it suffers from noise, then, and only then, we'll be very much able to help you with your actual problem! $\endgroup$ – Marcus Müller May 14 at 18:39
  • $\begingroup$ Hi! Thanks! So, yet another non-optimal approach: the transform to frequency domain-remove-transform back approach is almost certainly not a good idea. Also, has nothing to do with the Butterworth mentioned in the next paragraph! Furthermore, your signal is already digital, so the continuous-time-typical Butterworth is almost certainly not your prior choice here! So, what will you do with the "clean" signal? $\endgroup$ – Marcus Müller May 14 at 20:10

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.