is phase shifting a side effect of digital filtering? the higher the order of filter, the more phase shifting? how to select the right order?

[fenny]

Seems that higher order filter has better filter effect if cut-off frequnecy is set to be constant.
So filter order the higher the better?
is there any side effects for higher order? such as phase delay?

[johnsold]

All filters have delay or phase shifting effects. Steep roll-off (high order) filters tend to have bigger phase effects than lower order filters of the same type and nominal frequency. An all pass filter exists which can have constant amplitude but varying phase versus frequency. These can be used to compensate or equalize the delay.

To specify a filter you need to specify the attenuation as a function of frequency for both the frequencies you want to pass through the filter and those which the filter must reject. If phase or delay is important to your application, it should also be specified. It is easy to specify a filter which is quite difficult to implement, so do not overspecify.

Tell us more about your application and someone on the forum may be able to help.

Lynn

[fenny]

how does phase shifting come, theoretically?
So is there a upper limit for filter order to minimize the phase shifting effect?

[fenny]

is the all-pass filter available in LABVIEW? I just did not see it on the pallette.

[johnsold]

Theoretically, hmmm... I haven't thought about that for quite a few years. Think of a wide bandwidth filter like a smooth four-lane highway - vehicles can travel at high speed. Compare that with a goat path over the mountain. Only specialized vehicles can take that route and then only slowly. A wide bandwidth filter has less delay than a narrow one. You can look at the transfer functions of simple filters and compare the time domain responses with the frequency domain (Fourier or Laplace transformations) if you want to get into the mathematics.

Higher order filters will tend to have faster phase changes than lower order filters for the same cut-off frequency. Butterworth and elliptical filters of the same order and same cut-off frequency have quite different phase responses. So there are not simple rules. There are lots of books and papers of filter design.

As I said earlier, try to specify your filter or describe your application so someone can help.

Lynn

[rpursley8]

A quick way to compensate for most of the phase shift (most of it is time delay from passing through the filter) is to select your filter order (even number) and then feed your signal through a filter with half that order then reverse your 1D array signal and run it through the same filter again. Reverse it again to get it the way it was before and the resulting signal should be OK. I would use Butterworth filters, I think this procedure works best with them.

[fenny]

sounds interesting! I will try it!
thanks!

[TylerS]

fenny,

An all-pass filter would be no filter at all.

Tyler S

[fenny]

thank you all!

[johnsold]

Tyler, Fenny,

"Up until now, the networks we discussed were used to obtain a desired amplitude versus frequency characteristic. No less important is the all-pass family of filters. This class of networks exhibits a flat frequency response but introduces a prescribed phase shift versus frequency. All-pass filters are frequently called "delay equalizers.""
Arthur B. Williams, Electronic Filter Design Handbook, McGraw-Hill, 1981, page 7-1.

The poles and zeros of an all-pass filter are mirror images of each other. I have built hardware equalizers but never tried it in LV.

Lynn