i discovered today that the -6db filter type can be used for perfect band splitting; if i'm correct they're linkwitz-riley filters
of course, setting up a band split and increasing the order (roll-off) screws up the phase, as does putting the filters in series
idk whether this is a bug or a feature of the filters, but i'll file something here just in case
-6db filters: higher order screws up phase
- AutumnCheney
- Posts: 507
- Joined: Sun Dec 29, 2019 8:16 am
- Location: tahlequah, ok, usa
- Contact:
Re: -6db filters: higher order screws up phase
Its a feature of the filter, in fact you can alse see filtering in terms of phase shifts, canceling the signal.
First order lp iir: the signal is delayed one sample, which is a 180 degrees phase shift for Nyquist frequency, but hardly any shift for very low frequencies.
This is folded back on the signal - convolution , - canceling or damping the higher frequencies.
At the same time you can see it as an integration with an exponentially decaying window. Both views are interchangeable, but not seperable.
First order lp iir: the signal is delayed one sample, which is a 180 degrees phase shift for Nyquist frequency, but hardly any shift for very low frequencies.
This is folded back on the signal - convolution , - canceling or damping the higher frequencies.
At the same time you can see it as an integration with an exponentially decaying window. Both views are interchangeable, but not seperable.
- AutumnCheney
- Posts: 507
- Joined: Sun Dec 29, 2019 8:16 am
- Location: tahlequah, ok, usa
- Contact:
Re: -6db filters: higher order screws up phase
i am aware of how filters work, but my expectation is that if i'm correct about the -6db filters, increasing the order should change the phase shift in such a way that the sum of the lowpass and highpass are equal to the input, which it isn't
Re: -6db filters: higher order screws up phase
sorry I didnt get you were talking about Sunvox specifically.
You can sum 1 1 pole lowpass highpass, but not more of them in series, you need a Linkwitz Rikey response, two times Butterworth.
In other words, the higher the filter order the steeper the rolloff you need.
The rolloff of a 1 pole covers the whole frequency range, so its not steep enough when you add more of them in series cause it still covers the whole range, the slope gets 12 db but the curve does not change to more brick wall but stays the same.
EDIT: SORRY you can have 2 in series, but not more.
EDIT 2 and you need to invert the highpass route, maybe thats the issue?
You can sum 1 1 pole lowpass highpass, but not more of them in series, you need a Linkwitz Rikey response, two times Butterworth.
In other words, the higher the filter order the steeper the rolloff you need.
The rolloff of a 1 pole covers the whole frequency range, so its not steep enough when you add more of them in series cause it still covers the whole range, the slope gets 12 db but the curve does not change to more brick wall but stays the same.
EDIT: SORRY you can have 2 in series, but not more.
EDIT 2 and you need to invert the highpass route, maybe thats the issue?
Re: -6db filters: higher order screws up phase
sorry I a m bit confused today, but this is never the case I think.AutumnCheney wrote: ↑Sun Oct 16, 2022 9:42 pmincreasing the order should change the phase shift in such a way that the sum of the lowpass and highpass are equal to the input, which it isn't
The signal is delayed around the transition band.
I dont see a way this could be ever avoided.