Rational Acoustics

June 29th, 2015, 09:56 AM
Im wondering if anyone can shed some light on the topic of using all pass filters instead of delay to achieve driver to driver (like mid to hi freq.) phase coherence thru crossover. Ever since learning the method reading these forums and associated links, I've come to understand and apply the method of phase aligning using delay, where you overlap the phase traces of two adjacent pass-bands. What I would like to know is under what circumstances is that an inferior method to using all pass filters? And even with all pass filters, would you still use delay to account for the distance differences between, let's say, a mains and a sub cabinet.?

June 30th, 2015, 02:58 AM
Hi Niick,

As always, it depends.

Pure (electronic) delay is frequency independent delay for all frequencies.

All-pass filters is frequency dependent delay for some frequencies.

E.g. which system could exhibit more phase shift ergo group delay. A 2-way speaker with one crossover or a 3-way speaker with two crossovers? An all-pass filter would allow you to introduce an amount of phase shift equal (frequency dependable delay) to the missing 2nd crossover in the 2-way speaker making it equally "slow"

Another example would be a situation where the main system is physically so far to the back in relation to the subject(s), that you might not want to apply additional pure delay for the alignment tot the subwoofers because the sound will arrive after the acoustical contribution of stage. Since it's important only those frequencies that are within 10 dB of relative level compared to the subwoofers are aligned, all-pass filter(s) could do the job by introducing frequency dependable delay. It doesn't matter if the MF and HF are misaligned because they're dominant by 10 dB or more over the subwoofers limited LF only band.

10 dB of relative level separation will reduce frequency response ripple (comb filter) from misalignment down to 6 dB or less.

If you wanna experiment and see the results in Smaart consider using the plugin called CrossoverUnit. Follow the link below for more information.


Arthur Skudra
July 1st, 2015, 07:38 PM

Check out this thread where I describe how I aligned two different speakers whose placement were physically compromised, causing a perfect cancellation in the HF due to diverging phase traces despite using the appropriate amount of time delay to get the two boxes to arrive at the same time. I used all pass filters to make the two speakers align better in the overlap region in the HF. A classic case where time alignment only gets you part of the way there.


July 2nd, 2015, 12:42 PM
Arthur, thanks again! This is exactly the type of example I was looking for.

Timo Beckman
May 16th, 2016, 07:35 AM
Since I use Meyer Sound i do not have to worry about the driver to driver stuff but between the UPQ1p and 900LFC's i have you can do some pretty nice things. Getting stuff aligned in a stacked PA for example without using delay but via Allpass filters.....



Timo Beckman
October 11th, 2016, 11:23 AM
I did a new blog post where i use allpass filters to get a high and low driver to align (the ap2 stuff happens around +/- 14min). enjoy



Timo Beckman
October 12th, 2016, 11:03 AM
And the newest post where i use Allpass filters to correct the phase response of the high driver.


Timo Beckman
October 24th, 2016, 12:36 PM
So here's a little explanation on how I look at things. Hope it helps enjoy


January 6th, 2017, 06:05 AM
The CrossoverUnit plug-in allows you to model all single slope filters like Bessel, Butterworth and Linkwitz-Riley up to 8th order (48 dB/oct.). In addition it also offers complex slope filters like Chebyschev and Inverse Chebyschev used in e.g. Meyer Sound's Galileo and Callisto processors.

Unfortunately CrossoverUnit doesn't support elliptical filters a.k.a. Cauer filters, also used in e.g. Meyer Sound's Galileo and Callisto processors. Fortunately there's another plug-in called EngineersFilter.

January 18th, 2017, 10:56 PM
Thank you everyone!! Some excellent examples. I sincerely appreciate all the replies. I work in the aftermarket car audio world and as of yet all pass filters are not widely known in DSP, save one notable exception.

The OEM's use them frequently, and a DSP manufacturer that I work closely with is soon going to be implementing this functionality.

Crossover unit is super cool... Unless, like me, you're running Windows 7. If this is the case, I've found a seriously awesome ASIO software DSP called EKIO. Using this with a completely free 8ch ASIO "virtual interface" called Voicemeeter allows one to model not only LR, BW, and Bessel crossovers but a seemingly unlimited number of parametric peak/sym type filters as well. Or EKIO can be used with an actual hardware interface, so long as it's using an ASIO driver. One yeah, both EKIO and Voicemeeter are free. There is a paid version of EKIO that gets you unlimited ins/outs, but the free version gives you 2 in/4 out.

So to Timo and anyone else who's used to measuring the TF of a processor, I have a question.

What would cause the IR to bounce back and forth, continously, by less than a single sample? Of course this makes the upper end of the phase trace bounce up and down non-stop, as the measure signal is constantly changing its time relation to the reference. I've narrowed down that this is absolutely not my measurement setup. I can switch in a different processor, or multiple other processors and the problem goes away. I've only seen it with one specific manufacturer.

I don't know a whole hell of a lot about this kinda stuff, but it seems to me that possibly what I'm seeing is an extremely severe case of what is referred to as clock Jitter? Maybe? One of the problem processors is a pre-production unit, but the other two I've tested from the same manufacturer also do the same thing.

It's exactly as if I'm sitting there adding (somehow) about a half second of delay to the measured processor channel, then immediately taking it away, then adding it again, taking it away, back and forth, back and forth... This is exactly what it would look like. The live IR bounces back and forth and the high end of the phase trace bounces up and down by about 10-15 degrees or so.

It's strange, I've never seen it before on any other DSP.

Again, thanks to everyone for the replies and happy 2017 to all!! :)

Timo Beckman
January 20th, 2017, 05:40 AM
Is the divide under test you're talking about running at a sample rate of 96K?

January 20th, 2017, 12:19 PM
The DUT runs at a SR of 48k. i tried measuring the "problem" device with Both Smaart and SysTune Pro. Exact same results. I also tried to run my analysis at 96k as to hopefully get a better idea of the amount of IR "movement" back and forth, and I could see it clearly either way really.

I can immediately swap in a processor from a different manufacturer and the "problem" goes away. I think Ill make a video of what Im talking about to clearly show the measurement setup and analysis parameters, and maybe with any luck somewhone will catch something i'm missing. Or at least confirm that the device does seem to be behaving in an unexpected manner.

Oh yeah, one more thing, I can input either digital or analog audio straight out of my interface into the DUT and the problem persists, and the same is true of the output from the DUT to my interface. Analog or digital, doesnt matter, same problem....or, "problem"??

January 20th, 2017, 02:56 PM
Here is a video where you can kinda see it in action


Timo, I'm going to PM you with the link, on second thought Id rather not have the video in the public domain.

Timo Beckman
May 17th, 2017, 07:00 AM
This not really about driver to driver stuff and the implementation of Allpass filters but never the less an example of the use of them.....Enjoy.