Agree your math would sound logic but numbers are bit different
.2nd order LR are Q0,5 same as 2x cascaded 1st order BW (phaseshift 180º), 4th order LR are Q0,7071 same as 2x cascaded 2nd order BW (phaseshift 360º), 1st order All-pass are Q0,5 and phaseshift is surprisingly not 90º but 180º, and 2nd order All-pass are Q0,7071 and phaseshift is surprisingly not 180º but 360º.
All the order of Linkwintz-Riley get a Q=0.5
they are all two cascaded Butterworth (2,4,6,8,10...).
8 order (as an example) has a Q=0.5 (all the Linkwitz-Riley).
or,an another way is to apply directly the right coeff.In compensate mode,"inverse filter" is a mirror of the slope.
Active Filters
they are all two cascaded Butterworth (2,4,6,8,10...).
8 order (as an example) has a Q=0.5 (all the Linkwitz-Riley).
or,an another way is to apply directly the right coeff.In compensate mode,"inverse filter" is a mirror of the slope.
Active Filters
All the order of Linkwintz-Riley get a Q=0.5
they are all two cascaded Butterworth (2,4,6,8,10...).
8 order (as an example) has a Q=0.5 (all the Linkwitz-Riley).
or,an another way is to apply directly the right coeff.In compensate mode,"inverse filter" is a mirror of the slope.
Active Filters
Mr thierry38 seems right !
Sometimes I'm so glad I do not run any crossovers at all
It is too complicated for you
PinocQ<1 damp the signal
Q>1 undamp the signal
A Q of 0.7 damp your signal, if you damp two times your signal, the resulting signal is more damped (you lower the Q)
The LR filter should be damped with a Q of 0.5, so for each filter order you must adjust all the stages Q in order to obtain the desired Q (of 0.5).
That can it be but its a confusing way isn't it
.First pos say:
Second theirry38 say:
So far only understand thierry38's point about the Q's when multiplied sum Q0,5 but is confused why suggest two of them all-pass when 2nd order all-pass have 360º phase shift and pos say one in compensate mode Q0.707 will repair 4th order LR, which below page 6 from Grimm AUDIO LS1 paper also suggest.
Attachments
Thanks Pos, thanks all,
Sorry to ask question and disappear...travel jumped up.
Ok, I now see how abrupt what I was trying to do is, and the need for many more taps.
I started toying with all-pass filters as way of working on phase when I can't use many taps in live sound situations. So I was using rephase to see their behaviour.
I guess the lesson is, there is no free lunch in the world of phase correction either.
Filters linearization it is, and a wee bit of Para EQ
Sorry to ask question and disappear...travel jumped up.
Ok, I now see how abrupt what I was trying to do is, and the need for many more taps.
I started toying with all-pass filters as way of working on phase when I can't use many taps in live sound situations. So I was using rephase to see their behaviour.
I guess the lesson is, there is no free lunch in the world of phase correction either.
Filters linearization it is, and a wee bit of Para EQ
Why not use the allpass filters as IIR in MiniDSP no need to worry about taps then
Yeah, that was my idea, to try to use IIR all-pass to cut down latency. ..
But I wasn't thinking about it with miniDSP IIR, which doesn't have all-pass capabilities as far as I know...unless it can be done with biquads?????
It's no problem with any of the miniDSP units. Just generate the proper coefficients with a utility and paste into the advanced mode of any miniDSP software plugin.
Dave.
Yes it can be done with IIR, use the advanced biquad programming, they have a spreadsheet that can do the calculations
Advanced Biquad Programming
It doesn't make much sense to me to create an allpass in FIR unless you are trying to copy an existing design. They aren't as flexible but could be made to work in correcting phase with no latency.
Edit: Dave beat me to it
Yes it can be done with IIR, use the advanced biquad programming, they have a spreadsheet that can do the calculations
Advanced Biquad Programming
It doesn't make much sense to me to create an allpass in FIR unless you are trying to copy an existing design. They aren't as flexible but could be made to work in correcting phase with no latency.
Edit: Dave beat me to it
Yes, one can create allpass filters with biquads and I tried it but quickly gave up because I didn't know how to compute the biquad coefficients that would result in the phase curve that I needed.
I think passive all pass filters came into use when DSP wasn't commonly available to provide a simple delay, constant over a limited frequency range, for time alignment of drivers in a multi-way speaker,. Today we use DSP delay for this.
Using IIR all pass filters to flatten the phase curve of a speaker after time alignment has been achieved is more difficult compared to just dialing in the correction with RePhase's paragraphic phase eq tab. It seems if you have a smoothly changing phase you might find a reasonable fit with an all pass filter just by trial and error in the spreadsheet. If there is a way other than trial and error I'd certainly like to know about it.
You've got that backward. Normally, you would use an IIR all-pass filter to achieve alignment. However, it doesn't flatten the phase curve.
An IIR/analog all-pass filter only adds further phase distortion to what's already there.
This subject is getting confusing and it seems like there's about three different conversations with three different premises.
Dave.
You've got that backward. Normally, you would use an IIR all-pass filter to achieve alignment. However, it doesn't flatten the phase curve.
An IIR/analog all-pass filter only adds further phase distortion to what's already there.
This subject is getting confusing and it seems like there's about three different conversations with three different premises.
Dave.
read my previous paragraph:
"I think passive all pass filters came into use when DSP wasn't commonly available to provide a simple delay, constant over a limited frequency range, for time alignment of drivers in a multi-way speaker,. Today we use DSP delay for this"
I don't have it backwards and you seem to agree with me at least on that one point.
What needs clarifying/exploration is how to use APF to correct, not an arbitrary phase curve, put perhaps the slowly rising delay curve that we see in our speakers as frequency lowers in the bass and midbass. That is where use of FIR adds the latency we'd like to avoid and even an a partial correction might be preferable to the latency.
Yes, that's exactly what rePhase can achieve. But you can't acheive that with any analog or DSP IIR solution. I'm not sure why adding latency is a problem.....unless this an HT configuration or some such.
Regardless, I assume we're on the same page here, but some of the language in later posts in this thread leaves me puzzled.
A passive all-pass filter is not normally seen in most speaker systems....then or now. (A lattice network would be an example.)
Dave.
Last edited:
Hi Pos, hi all,
Beginning to play with idea of using all pass filters to help flatten phase on a driver-by-driver basis.
I'm particularly interested in 2nd order inverted all-pass, hopefully accomplished by using rePhase's 'compensate'.
One thing I don't understand yet (among many!) is why I get a magnitude notch like the one below.... my meager understanding of all-pass is that it doesn't effect magnitude ????
(Second screen shows parametric eq to overcome the notch.)
Any help understanding this appreciated
Thx, Mark
I think the key thing to notice in the post that created this discussion is the word: inverted.
Last edited:
Maybe of topic: all pass filters can be used with DSP to create time delay less than one sample lenght. So for High frequency it could be useful. It can also be useful for wave shaping using miltiple speaker elements
And 4th order LR is not 2* q=0,701 as the phase responce will be wrong So the cascading q tables must be used(or other tools)
And 4th order LR is not 2* q=0,701 as the phase responce will be wrong
So the cascading q tables must be used(or other tools)
Yes indeed, all the "compensate" filters in the minimum-phase tab are acausal and cannot be replicated with an IIR.
And of course compensate=inverse when talking about all-pass filters.
Last edited: