what does different filter order,signal processing specialist please help

[fenny]

I'm trying to filter out the high frequency noise(>2000hz), so I use a lowpass filter(butterworth filter) and set the frequency to be 2000 hz, but i noticed that when i use different order of filters, i got different results.
i have tried order=20,40,80,200, what on earth does the filter order mean?? how does it inflence my filtered result? any singal processing expert here?
thanks.

[Nanxiong_Zhang]

Hi,

An ideal lowpass filter will let all the frequencies under the cutoff frequency (2KHz in your case) pass without any change and filt out those greater than the cutoff frequency totally. Theoretically, when the order is infinte, you'll get an ideal lowpass filter. However, infinte order is not practical. Firstly, you do not have all the history data of the signal (since minor infinte). Secondly, the computing is impossible for infinte filter length. So when the order is a finite value, the lowpass filter is an approximation to the ideal filter. That means the frequencies under cutoff frequency are not passed exactly and the frequencies greater than the cutoff frequency are not really removed totally. The greater the order, the more the filter close to the
ideal one and the more computations. This is the reason why you got different result when using different order. But it is not necessary to have filter with so high order you tried. Typically an order less than 10 will work well.

[fenny]

i also find that compared to the original data, the filtered data (i mean in frequency domain ) changed the magnitude of the peaks, the position of the peaks are the same(this is as expected), why the magnitude changed? and the higher the order of the filter, the more the magnitude changed.

[fenny]

i also tried very high order (=300)filter, and found that, when the filter order is too high, the filtered signal is disfigured. I mean very different from the original signal. why? is there any guidance as to what order of filter should i use for my signal.

[Nanxiong_Zhang]

The peak in the signal may introduce high frequency components (> 2000Hz in your case). So when you apply a lowpass filter, it will be affected. The disstortion when the order is high is because of your signal is of finite length (You do not have all the history data of the signal). A less than 10 order should work for most of cases.

If you like, you may attach your signal and VI so that I can look at it in more details.