# Extracting 60 Hz power line signal

Started by October 28, 2008
I have a wav file (biomedical signal) that is believed to be corrupted by AC line noise; I have been asked to extract the (nominal) 60 Hz component so that it can be further examined (idea is to see if power line frequency variations are partly responsible for causing problems).

The original signal was sampled at 8KHz. I've decimated and bandpass filtered the signal around 59-61 Hz, but get a very ugly looking result (it looks very random...not periodic looking at all).

I also tried using Matlab's specgram function...played around with various combinations of FFT and window size. There is definitely a 60 Hz component, but the resolution is too coarse to be useful for analysis.

Any suggestions on how to achieve a good, clean signal here?

Thanks,

Vinny
Hi
hope this helps

You can increase the resolution of the FTT by increasing the signal length
(append with zeros), however if the signal has important information
between in the 60 Hz area, bandpassfiltering won't give a good result.
Istead you can use "the Goertzel Algorithm" analyse the frequency content
in the area of interrest (59-61Hz).

For instance if you want a resolution of 1mHz you calculate 2Hz/1mHz 00
Goertzels one for each frequency you want to analyse. Then the Goertzel
with the highst result will give you the frequency of the AC disturbance

Afterwords you can remove the AC-line noise with an LMS filter, for which
you will need an reference signal with the exact frequency you want to
remove - to genrate this refrence signal you can also use the Goertzel
algorithm- i.e impulse response of the Goertzel is an sin-wave, so if you
use the same coefficients as in the frequency analysis you will have the
refrence singal you need.

this way you will only remove the specified frequency.

Regards
Ole

On Tue, 28 Oct 2008 02:27:57 +0100, wrote:

> I have a wav file (biomedical signal) that is believed to be corrupted
> by AC line noise; I have been asked to extract the (nominal) 60 Hz
> component so that it can be further examined (idea is to see if power
> line frequency variations are partly responsible for causing problems).
>
> The original signal was sampled at 8KHz. I've decimated and bandpass
> filtered the signal around 59-61 Hz, but get a very ugly looking result
> (it looks very random...not periodic looking at all).
>
> I also tried using Matlab's specgram function...played around with
> various combinations of FFT and window size. There is definitely a 60 Hz
> component, but the resolution is too coarse to be useful for analysis.
>
> Any suggestions on how to achieve a good, clean signal here?
>
> Thanks,
>
> Vinny