A while back i was gathering some ECG data for a colleague at a reasonably high sample rate (around 20kHz - 100Hz is enough for ECG). She came back to me and asked me to regather the data at a lower sample rate - claiming that the high sample rate was introducing extra noise.

Is this plausible, realistic, or based in truth? I can't see how a higher sample rate would make a signal worse - especially considering the data was gathered on an expensive hi-spec oscilloscope

  • 6
    \$\begingroup\$ There's nothing you can accomplish by using a lower sample rate that you can't also accomplish by filtering and decimating the original data set. In other words, your colleague is being lazy, and is pushing the extra work onto you. \$\endgroup\$ – Dave Tweed Dec 3 '15 at 17:03
  • 2
    \$\begingroup\$ Ditto. The "extra noise" is just a more precise measurement. \$\endgroup\$ – Eugene Sh. Dec 3 '15 at 17:04
  • \$\begingroup\$ It depends, if you use a lower sample rate, the oscilloscope may filter out higher frequency information to prevent aliasing... However, as mentioned by Dave, you should be able to filter out the high frequency noise from the resultant data as well. \$\endgroup\$ – Jarrod Christman Dec 3 '15 at 17:04
  • \$\begingroup\$ The only downside of high sample rate is... large data... MSO scope can acquire 10million samples and with 8gig RAM (corporate PC...) matlab struggles... \$\endgroup\$ – JonRB Dec 3 '15 at 17:06
  • 1
    \$\begingroup\$ 100 Hz is probably NOT enough for accurate ECG. 200 is more like it. Don't forget Nyquist. \$\endgroup\$ – Scott Seidman Dec 3 '15 at 17:12

Your college does not seem to remember her statistics lessons. The additional higher frequency noise can be trivially filtered with a low pass filter, and the filtered signal might be better than one sampled at low frequency (oversampling).

Note that using a standard oscilloscope to record live ECG violates patient protection, as medical equipment requires better electrical isolation than those devices usually provide.

  • \$\begingroup\$ The filtered and DECIMATED signal might be more useful-- depending of course on what needs to be done with the resulting data. Filter design can get tricky when all your data is compressed in one part of your spectrum, and decimating conveniently spreads it out. \$\endgroup\$ – Scott Seidman Dec 3 '15 at 17:14
  • \$\begingroup\$ This was not in a patient situation, and was not using anything close to standard ECG equipment - don't worry, we aren't trying to hurt anyone! \$\endgroup\$ – Mauvai Dec 3 '15 at 17:26
  • \$\begingroup\$ Also, thanks! I was reasonably sure that was the answer, but I wasn't in a position to argue and i didn't know for sure \$\endgroup\$ – Mauvai Dec 3 '15 at 17:27

Your colleague is correct. Using a high sample rate captures a wider bandwidth, which captures more noise. Downsampling, or capturing with a lower sample rate in the first place is similar to averaging several adjacent samples into one sample, which would remove the noise.

As the commenters point out, your colleague could resample the file herself to remove the extra noise. I would argue that the data should be captured at a sample rate appropriate for the signal in question in the first place. It will produce smaller files and removes one processing step.

  • 2
    \$\begingroup\$ I agree that higher sampling rate increases what is captured. I disagree with compromising via capturing at a lower rate. Capture at as high a rate as possible and post-process \$\endgroup\$ – JonRB Dec 3 '15 at 17:08
  • 1
    \$\begingroup\$ The colleague would be correct if the noise is introduced by the scope. But you are talking about the noise which is already a part of the signal. \$\endgroup\$ – Eugene Sh. Dec 3 '15 at 17:08
  • \$\begingroup\$ That would depend on the bandwidth of the noise in the signal. If the noise extends all the way to the Nyquist limit a lower sample rate will remove it. \$\endgroup\$ – Austin Dec 3 '15 at 17:20
  • \$\begingroup\$ @Austin would that approach not simply alias the higher frequency noise and increase the total low frequency noise? \$\endgroup\$ – Mauvai Dec 3 '15 at 17:28
  • 1
    \$\begingroup\$ What you just said is absolutely true, provided that the sample rate is actually fast enough in the first place. The oscilloscope must first have adequate information on hand. If one samples too slowly, legitimate information is lost to aliasing, and no filter can bring it back because the information is simply not there. This is the argument for oversampling and post-processing. \$\endgroup\$ – Sean Boddy Dec 3 '15 at 18:37

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.