There is the first order (-6dB / octave) rolloff, that if done correctly has a flat phase shift. Sometimes called a “transient perfect” filter. But this requires a very well behaved flat response drivers without breakups to utilize. This is used in some Dunlavey speakers and I have also used it on my 10F/RS225 FAST/WAW speaker.
https://www.diyaudio.com/community/threads/10f-8424-rs225-8-fast-waw-ref-monitor.273524/
There are also ways to minimize phase shifts such as the Harsch XO that is flat except for a 55deg bump near the XO frequency.
https://www.diyaudio.com/community/threads/s-harsch-xo.277691/
Note the asymmetric slopes (4th low pass and 2nd order high pass) plus there is a time delay needed between the tweeter and woofer that is rather large so usually done with DSP. However, for high XO frequencies, it is possible to do this passively with a waveguide and mechanical offsets to achieve the delay and slopes needed for a passive Harsch XO:
https://www.diyaudio.com/community/...using-ptt6-5-and-rs28f-in-a-waveguide.354778/
A Harsch XO achieves a close to a transient perfect response as can be seen in the measured step response (close to a right triangle shape):
Here is Dunlavey step response as measured by Stereophile for comparison:
https://www.stereophile.com/content/dunlavy-audio-laboratories-sc-iv-loudspeaker-measurements
https://www.diyaudio.com/community/threads/10f-8424-rs225-8-fast-waw-ref-monitor.273524/
There are also ways to minimize phase shifts such as the Harsch XO that is flat except for a 55deg bump near the XO frequency.
https://www.diyaudio.com/community/threads/s-harsch-xo.277691/
Note the asymmetric slopes (4th low pass and 2nd order high pass) plus there is a time delay needed between the tweeter and woofer that is rather large so usually done with DSP. However, for high XO frequencies, it is possible to do this passively with a waveguide and mechanical offsets to achieve the delay and slopes needed for a passive Harsch XO:
https://www.diyaudio.com/community/...using-ptt6-5-and-rs28f-in-a-waveguide.354778/
A Harsch XO achieves a close to a transient perfect response as can be seen in the measured step response (close to a right triangle shape):
Here is Dunlavey step response as measured by Stereophile for comparison:
https://www.stereophile.com/content/dunlavy-audio-laboratories-sc-iv-loudspeaker-measurements

Last edited:
Hi PresscotAs a result, the two active filters won't cause phase shift, or maybe did but cancelled each other, and a remaining passive filter will leave the phase shift of 90 degrees alone. Do I understand correctly?
No, you actually don't. Multiway speakers show almost always (>99% !) allpass behaviour. I.e. they are non minimum-phase. This is independant of the type of filters used (mechanical, passive elecrical or active electrical). You can have a flat amplitude response but not a flat phase-response. You can't correct a phase error on one way with a different phase response in another one.
But the single ways do actually behave as minimum phase devices. If you correct an amplitude response error you will also correct the corresponding phase response of that particular way. But as soon as you combine your ways you will again have the allpass behaviour of the whole system.
There are indeed crossovers that don't behave as allpass filters but these are not recommended for "normal" multiway speakers and implementing them is definitely not for the faint-hearted (and won't work in your case).
Regards
Charles
Edit: XRK was faster and explained more deeply what I mentioned in my last sentence.
Thank you, Charles. You were the one who introduced me to the Harsch XO and I have always liked how it sounds on percussion and strings. It also has great imaging and sound stage.
There's another, much easier "trick" for higher-order...if you can stand a little constant delay.
Simple complementary linear-phase xovers. Choose your order, doesn't matter.
here's a 5-way, all xovers are 16th order (using 85ms delay)
Any order could have been used,
Simple complementary linear-phase xovers. Choose your order, doesn't matter.
here's a 5-way, all xovers are 16th order (using 85ms delay)
Any order could have been used,
Hi Mark100,
This is the ultimate trick. Requires dedicated real time processing engine via DSP/hardware. Usually a PC, sound card, and 3rd party software. Has anyone made their own FPGA to do this?
What hardware and software combo do you use?
This is the ultimate trick. Requires dedicated real time processing engine via DSP/hardware. Usually a PC, sound card, and 3rd party software. Has anyone made their own FPGA to do this?
What hardware and software combo do you use?
Hi xrk971,
Knowing how into things DIYers can get, tt wouldn't surprise me if someone has built the own FPGA implementation. 🙂
I haven't looked around though..
...because over the years in terms of hardware, I just started with a bank of miniDSP OpenDRC, then went to PC for a short while using JRiver convolution, to return to the OpenDRCs.
To finally end up moving to QSC's Q-Sys platform. (Which I absolutely love despite being 48kHz, and having a tap limitation of 16k taps per channel.)
Software has been rePhase, and FirDesigner.
I keep toying with using REW as a FIR generator, but I'm still too slow and error prone with it
In all hardware cases, my technique has been the same. Build a FIR file for each driver that includes both linear-phase xovers and minimum-phase driver EQs.
Then all that's needed to finish up, are levels and delays.
My method is too easy lol.
I admire you guys who really know passive / IIR.....
Knowing how into things DIYers can get, tt wouldn't surprise me if someone has built the own FPGA implementation. 🙂
I haven't looked around though..
...because over the years in terms of hardware, I just started with a bank of miniDSP OpenDRC, then went to PC for a short while using JRiver convolution, to return to the OpenDRCs.
To finally end up moving to QSC's Q-Sys platform. (Which I absolutely love despite being 48kHz, and having a tap limitation of 16k taps per channel.)
Software has been rePhase, and FirDesigner.
I keep toying with using REW as a FIR generator, but I'm still too slow and error prone with it
In all hardware cases, my technique has been the same. Build a FIR file for each driver that includes both linear-phase xovers and minimum-phase driver EQs.
Then all that's needed to finish up, are levels and delays.
My method is too easy lol.
I admire you guys who really know passive / IIR.....
I wonder if Analog Devices ADAU-1452 eval kit can be programmed using Sigma DSP suite to do FIR? It takes data in almost any digital form and also Toslink, and analog.
https://www.analog.com/en/products/adau1452.html
https://www.analog.com/en/products/adau1452.html
I think I can already smell some of the smoke that's (probably) coming out of the poor OP's brains... 😴..here's a 5-way, all xovers are 16th order..
I wonder if Analog Devices ADAU-1452 eval kit can be programmed using Sigma DSP suite to do FIR? It takes data in almost any digital form and also Toslink, and analog.
FIR can be done with Sigma Studio. But I would have to have a closer look to see how it is done. But it is definitley feasible. If you want to do transient perfect crossovers you can still use the classic subtractive-delay approach.
Regards
Charles
- Home
- Loudspeakers
- Multi-Way
- A trick for creating a higher-order slope