How to select proper values for low or high cut-off freqency for butterworth filter?

[edmonton]

Hi Engineers,

I am going to use Butterworth filter to remove a few high value noises, I played around and could not get ride of the spikes. 

The data was from a acquisition channel, data rate is 5 samples per second, the spikes are approximately 4s wide

Could someone help?

Attached is my Vi of LB2015

Thanks,

 Gu

[joshdoe420]

Try increasing the filter window from 2 to maybe 10 or more.

[Kevin_Price · Accepted solution]

Filtering can be part science, part art (based on domain knowledge).   Here I'd be tempted to use a median filter.

A median filter can be more effective at removing short-duration "spiky" outliers, but you should *also* have a good reason for erasing those outliers.  Same as needing a reason for setting your Butterworth cutoff frequency -- which stuff is ok to attenuate and why?

Using a median filter with both left & right rank set to 20 samples "erased" all the spikes in your provided data set.  Whether or not that's a valid manipulation of the data is for you to decide.

-Kevin P

[edmonton]

Hi Kevin,

An excellent solution. Thanks.

I also tried peak detection vi and successfully removed the spikes, which are abnormal signals and must be removed.  Using a median filter is the best method, simple code and efficient. 

Gu