ProSoundWeb Community

Please login or register.

Login with username, password and session length
Advanced search  

Pages: 1 [2] 3  All   Go Down

Author Topic: Combining drivers  (Read 7201 times)

Mark Wilkinson

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1104
Re: Combining drivers
« Reply #10 on: March 03, 2021, 08:31:20 PM »

So, I've just been having another think about this, and I feel like there must be more to this alignment than just the drivers' natural roll-off. If this speaker uses any kind of an HF waveguide (horn, etc.) there's almost certainly a physical offset between the HF and LF, and compensating for this requires some kind of time-domain solution, wouldn't it? Without knowing the speaker's design it's difficult to say for sure, but I still feel like the only way for this speaker to have a "pc75" phase response is going to involve delaying the LF and then dealing with the resulting wrap one way or another.

-Russ

For sure Russ, (and thanks for the earlier vote of confidence, but i'd trade what i know for what you know in a heartbeat !)
with both drivers having flat phase on their own, it still takes some time delay to tie them together so that the entire trace is mag and phase flat.

After flattening each driver's out-of-band response, the difference in Smaart's Delay Finder between the two, will exactly equal the processor delay needed to tie them together.
And that processor delay also equals  physical distance between acoustic centers of the two drivers.

That's one of the beauties of flat phase....makes timing easy peasy....and also allows common sense distance observation/verification like you were mentioning.

If Helge gets the out of bound flattening in place on the upper end of the lower driver, and the lower end of the upper driver, he will find he can move the FIR brickwall xover to any freq  within the jointly flattened range, and not have to change processor delay at all (as found via the above).
That's another of the beauties.



Logged

Russell Ault

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 2484
  • Edmonton, AB
Re: Combining drivers
« Reply #11 on: March 03, 2021, 10:57:54 PM »

{...} with both drivers having flat phase on their own, it still takes some time delay to tie them together so that the entire trace is mag and phase flat. {...}

Sorry, something still isn't working for me, but my experience with zero-phase filters is basically, er, zero, so let me see if I have this correct:

  • Setup a speaker and a measurement system somewhere where no one will want to punch you if you run pink noise for an hour straight
  • Put pink noise through HF driver
  • Use Delay Finder, note the delay compensation time
  • Adjust EQ/phase response of the driver until everything is as flat as you can get
  • Put pink noise through LF driver
  • Use Delay Finder, note the delay compensation time
  • Adjust EQ/phase response of the driver until everything is as flat as you can get
  • Insert zero-phase crossovers to taste

Right so far? If so, here's the problem I'm seeing:

We have our beautifully flat HF trace and our beautifully flat LF trace, but the LF is arriving early. As you say, no problem, we just delay the LF channel by the Delay Finder difference. Now everything lines up, but the LF trace is no longer flat, because adding delay to the LF causes its phase trace to start sloping.

The only way to avoid this entirely that I can think of would be to do all the component-level flattening at the single, fixed delay compensation setting found with the latest-arriving component. Done this way, any needed alignment delay would be already baked into any phase-altering processing.

In the case of Helge's traces, flattening the phase trace of the HF is going to take some significant processing muscle since it's lagging by ~180 degrees at the crossover point. An FIR maximum-phase all-pass filter will do the trick, but will effectively result in the highest HF frequencies arriving later than they originally were, which will necessitate re-setting the delay compensation. (Doing this IIR with a whole mess of minimum-phase all-pass filters would have the same effect since, of course, you can't actually make the sound at 1.2 kHz arrive earlier, you can only make the sound at higher frequencies arrive late enough to match.)

Now that the HF is flat it's time to face the bad news waiting for us in the LF. Since the delay compensation has increased from 20.90 ms to ~21.31 ms (or more), the LF trace that was once within 15 degrees of in time through the crossover is now significantly leading, so we'll have to add one (or more) minimum-phase all-pass filters to re-flatten it, effectively using all-pass filters instead of channel delay.

With all that done, we will have two drivers that are in in time with each other throughout both pass-bands, but it'll take a while.

Personally, I think I'd rather do what Helge's done and then phase-tweak the combined response. It just seems easier to me! :)

-Russ
Logged

Mark Wilkinson

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1104
Re: Combining drivers
« Reply #12 on: March 04, 2021, 01:21:38 AM »

Sorry, something still isn't working for me, but my experience with zero-phase filters is basically, er, zero, so let me see if I have this correct:

  • Setup a speaker and a measurement system somewhere where no one will want to punch you if you run pink noise for an hour straight
  • Put pink noise through HF driver
  • Use Delay Finder, note the delay compensation time
  • Adjust EQ/phase response of the driver until everything is as flat as you can get
  • Put pink noise through LF driver
  • Use Delay Finder, note the delay compensation time
  • Adjust EQ/phase response of the driver until everything is as flat as you can get
  • Insert zero-phase crossovers to taste

Right so far? If so, here's the problem I'm seeing:




Personally, I think I'd rather do what Helge's done and then phase-tweak the combined response. It just seems easier to me! :)

-Russ

Aaah, my apologies Russ, i wasn't clear in last post when to use Delay Finder's readings..
And just to be clearer backing up further to first post, I was trying to answer Helge's question about how to remove the final remaining phase wrap, not suggest an overall different methodology to what you two were discussing.)

May I rearrange your sequence:
  • Put pink noise through HF driver
  • Adjust EQ/phase response of the driver until everything is as flat as you can get
  • Put pink noise through LF driver
  • Adjust EQ/phase response of the driver until everything is as flat as you can get
  • Insert zero-phase crossovers to taste
  • Use Delay Finder, note the HF delay compensation time
  • Use Delay Finder, note the LF delay compensation time

   *Use the Delay Finder difference as the delay to be put into processor for delaying whichever driver had the lower Delay Finder reading.
   *Done, in hopefully a lot less than an hour  :)

I guess we each have learned our methods, and they become the easiest for us. I use phase trace overlay a lot too, because dang, we most often have to.
Honestly, I consider guys like you and all the others that have mastered aligning sloping phase traces to have learned the universal tool, the more difficult tool to learn.
Because the additional flattening steps i proposed only work when both flattening and linear phase xovers are practical.
But when they are practical, alignment gets easier than phase overlay ime/imo.



Logged

Mark Wilkinson

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1104
Re: Combining drivers
« Reply #13 on: March 04, 2021, 01:27:05 AM »

Ok, I'll try to get outside next week and refine my pre-alignemt EQ and see what happens.
IIRC it's a linear phase brick wall filter, 1.2Khz/79db/oct.


Thanks :)

You bet !
Yep, looked like a brick wall.
Hope flattening works, please see some further comments about timing made in posts with Russ.
Also, must say I agree with Russ you're unlikely to hear a change at that freq, but it does make for nice looking traces haha. 
Logged

Russell Ault

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 2484
  • Edmonton, AB
Re: Combining drivers
« Reply #14 on: March 04, 2021, 04:43:09 PM »

Aaah, my apologies Russ, i wasn't clear in last post when to use Delay Finder's readings.. {...}

Sorry Mark, looking back I think you were clear. I was pretty tired yesterday and I think my imagination was leaving out a delay compensation adjustment. Now that I've thought it through again I totally get what you're describing.

That said, I will stand by a little portion of my statement: flattening the phase response of Helge's HF driver is going to take muscle. :)

{...} Also, must say I agree with Russ you're unlikely to hear a change at that freq, but it does make for nice looking traces haha. 

This is a fascinating can of worms. Here's my take:

Humans' ability to perceive time in the audio domain isn't particularly precise. For audio technicians this is very useful, since it gives us wiggle room to do things that would otherwise be impossible (the best example I can think of is the "latency budget" for IEMs).

The experiments into human hearing that ultimately gave us the precedence effect suggest that humans will typically "fuse" two identical clicks that arrive within 1 to 5 ms of each other, perceiving them to be just a single click. For more complex sounds that fusion can happen with delays of up to 40 ms depending on the nature of the sound.

In my mind, since phase is ultimately just time, these findings are directly applicable to the question of "can humans hear phase?". Given that phase offset results in time smear (as opposed to separate, distinct arrivals), it's safe to say that the fusion window is going to be bigger than the 1-5 ms range, although how much bigger will likely come down to the individual as much as the material. Conservatively, though, a fairly safe assumption is that any time smearing of less than 5 ms will be entirely imperceptible to practically all Western ears.

In phase terms, then, a second-order all-pass filter inserted at 200 Hz would produce ~5 ms of time offset and might just be perceptible in a double-blind study. At 1.2 kHz, producing that same time offset would require 1800 degrees of phase shift. Conversely, at 50 Hz even a 90 degree phase offset might be perceptible. Anecdotally, I once had the opportunity to hear a sound system that had been processed to be phase-flat from 30 Hz to 20 kHz, and I can tell you that the A/B difference was not only perceptible, it was visceral.

Of course all filters, by their nature, produce delay in the form of phase offset. While zero-phase filters (and FIR processing in general) appear to be magic, the truth is that under the hood they are still bound by the same laws of physics as any other filters, and the only way to "undo" do the phase offset caused by the filtering is to delay the rest of the signal to match the delay inherent to the filters being used (moar taps!). While this is a fairly transparent process with FIR filtering (and a real pain with IIR), either way it produces one very obvious side-effect: latency.

It's this latency, ultimately, that is the real pisser, at least in live sound. The only way to "undo" some entirely-perceivable phase shift at 50 Hz is to, in effect, delay the rest of the frequencies to match it, which can mean introducing 10s of ms of additional latency. As the frequencies go lower (i.e. the periods get longer) phase shift becomes more and more perceptible, but fixing that phase shift becomes more and more expensive (in both latency and dollars).

Now, that all being said, in my mind the real advantage of speakers that are largely flat-phase is that it makes array-building much easier.

-Russ
Logged

Mark Wilkinson

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1104
Re: Combining drivers
« Reply #15 on: March 05, 2021, 12:17:33 PM »

Sorry Mark, looking back I think you were clear. I was pretty tired yesterday and I think my imagination was leaving out a delay compensation adjustment. Now that I've thought it through again I totally get what you're describing.

That said, I will stand by a little portion of my statement: flattening the phase response of Helge's HF driver is going to take muscle. :)

This is a fascinating can of worms. Here's my take:

Humans' ability to perceive time in the audio domain isn't particularly precise. For audio technicians this is very useful, since it gives us wiggle room to do things that would otherwise be impossible (the best example I can think of is the "latency budget" for IEMs).

The experiments into human hearing that ultimately gave us the precedence effect suggest that humans will typically "fuse" two identical clicks that arrive within 1 to 5 ms of each other, perceiving them to be just a single click. For more complex sounds that fusion can happen with delays of up to 40 ms depending on the nature of the sound.

In my mind, since phase is ultimately just time, these findings are directly applicable to the question of "can humans hear phase?". Given that phase offset results in time smear (as opposed to separate, distinct arrivals), it's safe to say that the fusion window is going to be bigger than the 1-5 ms range, although how much bigger will likely come down to the individual as much as the material. Conservatively, though, a fairly safe assumption is that any time smearing of less than 5 ms will be entirely imperceptible to practically all Western ears.

In phase terms, then, a second-order all-pass filter inserted at 200 Hz would produce ~5 ms of time offset and might just be perceptible in a double-blind study. At 1.2 kHz, producing that same time offset would require 1800 degrees of phase shift. Conversely, at 50 Hz even a 90 degree phase offset might be perceptible. Anecdotally, I once had the opportunity to hear a sound system that had been processed to be phase-flat from 30 Hz to 20 kHz, and I can tell you that the A/B difference was not only perceptible, it was visceral.

Of course all filters, by their nature, produce delay in the form of phase offset. While zero-phase filters (and FIR processing in general) appear to be magic, the truth is that under the hood they are still bound by the same laws of physics as any other filters, and the only way to "undo" do the phase offset caused by the filtering is to delay the rest of the signal to match the delay inherent to the filters being used (moar taps!). While this is a fairly transparent process with FIR filtering (and a real pain with IIR), either way it produces one very obvious side-effect: latency.

It's this latency, ultimately, that is the real pisser, at least in live sound. The only way to "undo" some entirely-perceivable phase shift at 50 Hz is to, in effect, delay the rest of the frequencies to match it, which can mean introducing 10s of ms of additional latency. As the frequencies go lower (i.e. the periods get longer) phase shift becomes more and more perceptible, but fixing that phase shift becomes more and more expensive (in both latency and dollars).

Now, that all being said, in my mind the real advantage of speakers that are largely flat-phase is that it makes array-building much easier.

-Russ

Good stuff Russ ! I found myself nodding my head yes to your comments.

Re:  flattening phase needing muscle.
Yep, i can see if someone is not in the habit of already doing in-band flattening, adding out-of-band flattening would be a task.
If they are accustomed to doing in-band EQs, I think they will be surprised by how easy out-of band flattening is, when using steep xovers.
I find, it doesn't take much work when the out-of-band summation range is narrow, and often helps make the in-band fall into place easier. 
But it takes lots of work with traditional shallow xovers like 24dB/oct or less, in which cases i usually i say screw it.

As an aside, another advantage I've found with steep linear phase xovers, is since the range of xover summation is so narrow, off-axis combing issues between the two drivers is minimized to a much narrower freq range than historically typical.
But again, since steep equals excessive phase wrap/group delay, it only works with linear phase xovers. Which of course means there is a low freq limit where steep works, when latency is an issue.

Re:  audibility of phase

This truly is a fascinating can of worms!  A can I've been digging in for about 5 years now, since i first started playing with FIR.

Here's my 2c take.
Imo, the jury is out.  I'm not convinced the historical research has adequately addressed phase audibility in music reproduction. 
I can find many studies that usually used test tones and occasionally music, where they ask listeners to judge between whether degrees of applied all-pass filters were audible etc.
Or some tests used simpler time delays at given frequencies.
 
Typically, the tests were done with headphones or smaller two-ways, since until recently it wasn't readily possible to get larger speakers with flat phase..
At any rate, phase audibility is continually debated huh?, and my take is generally deemed inaudible.

Which reminds of the beginning of the digital photography era, when DSLR's were in the 2-6 megapixel camp.  Debates raged on professional photography forums as to whether more megapixels were needed for normal sized prints.
The vast consensus said no, that more MP were only needed for enlargements, because study after study had proven the human eye can only see a little better than 200 lines-per-inch.
Well, twas true...the eye  could only see 200 Lpi at the times all the studies were made...because that's all the dang printers could print clearly !!!! Lol

So circling back to 'typically, the tests were done with headphones or smaller two-ways'.
My experience is that headphones and smaller speakers cannot impart the bottom octaves with sufficient impact, to gauge the effect of phase audibility down low.
I think our full-range larger speakers, with realistic bass and dynamics, have been like the old printers, historically incapable of playing music with flat phase for meaningful audible comparisons across the full spectrum.

And to my ears, that is where the advantage of flat phase is....down low. (Like in your 1.2kHz vs 50 Hz example.)

And I completely agree how audible and visceral flat-phase is 30Hz to 20kHz.
It's freaking awesome really...and was why i was belaboring about timing a bit on a recent Transient Response thread in the Sub Forum....saying it's a move beyond punch into the world of slam  ;D

But since the latency required to do such is so great, i've become hesitant to talk about it on the Live forums.
Just doesn't seem appropriate.

Anyway, for live, I've given myself a limit of 15ms total latency. Which sets the transition freq for moving from from FIR to IIR.
For playback, i use up to 170ms.  (16K taps in an Ebay QSys Core 500...not that $$ at all.)





Logged

Mike Caldwell

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 3074
  • Covington, Ohio
    • Mike Caldwell Audio Productions
Re: Combining drivers
« Reply #16 on: March 05, 2021, 05:42:09 PM »

As for finding driver offset timing have you done a comparison between using the delay finder
and response null test?

What I'm calling response null amounts to flipping the polarity between the woofer and high frequency driver and if front mounted delaying the woofer back and watching the transfer function until you get the deepest null at the crossover frequency.

Helge A Bentsen

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1753
  • Oslo, Norway.
Re: Combining drivers
« Reply #17 on: March 05, 2021, 06:39:16 PM »

As for finding driver offset timing have you done a comparison between using the delay finder
and response null test?

What I'm calling response null amounts to flipping the polarity between the woofer and high frequency driver and if front mounted delaying the woofer back and watching the transfer function until you get the deepest null at the crossover frequency.
Yes, did that to fine-tune my delay time using a mic on-axis and one off-axis.
« Last Edit: March 05, 2021, 06:42:13 PM by Helge A Bentsen »
Logged

Russell Ault

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 2484
  • Edmonton, AB
Re: Combining drivers
« Reply #18 on: March 05, 2021, 07:43:00 PM »

Re:  flattening phase needing muscle.
Yep, i can see if someone is not in the habit of already doing in-band flattening, adding out-of-band flattening would be a task.
If they are accustomed to doing in-band EQs, I think they will be surprised by how easy out-of band flattening is, when using steep xovers.
I find, it doesn't take much work when the out-of-band summation range is narrow, and often helps make the in-band fall into place easier. 
But it takes lots of work with traditional shallow xovers like 24dB/oct or less, in which cases i usually i say screw it.
{...}

I think it depends on how you define "muscle". In the world of FIR, flattening out Helge's HF response would be relatively easy. For me, though, basically everything I've interacted with in my career has been IIR only (the first time I actually got my hands on a DSP that was even capable of doing FIR filtering was after the pandemic started, and even then I haven't actually used its FIR functionality yet), and in IIR flattening a phase trace that's lagging by 180 degrees takes a lot of filters (sometimes more filters than an IIR-oriented system processors really wants to deal with, hence "muscle"). :)

-Russ
Logged

Chris Grimshaw

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1822
  • Sheffield, UK
    • Grimshaw Audio
Re: Combining drivers
« Reply #19 on: March 07, 2021, 02:50:24 AM »


The experiments into human hearing that ultimately gave us the precedence effect suggest that humans will typically "fuse" two identical clicks that arrive within 1 to 5 ms of each other, perceiving them to be just a single click. For more complex sounds that fusion can happen with delays of up to 40 ms depending on the nature of the sound.

To throw a slight curve-ball into the mix...

To me, an acoustically reflective room sounds fine at lower levels, but can turn to an awful mess if the levels are brought up a bit. To take an example I'm very familiar with, my living room has the speakers against one wall, and the sofa against the opposite wall. At sensible volumes, all is well. Turn it up, and the slap echo between the two walls becomes very obviously audible on snare drums, claps, etc. Anything transient.

I designed the speakers to be tolerant of relatively high power levels, so they're still acting linearly when the problems become audible.

However, if I go around with the measurement mic, the reflections are always there. It's only at higher levels that I notice them, which makes me suspect there's some filtering going on between my ears, which stops working above a particular SPL.


Following that, I'd suggest that any study into the precedence effect etc would be incomplete without introducing SPL as a variable.

Chris
Logged
Sheffield-based sound engineering.
www.grimshawaudio.com

ProSoundWeb Community

Re: Combining drivers
« Reply #19 on: March 07, 2021, 02:50:24 AM »


Pages: 1 [2] 3  All   Go Up
 



Site Hosted By Ashdown Technologies, Inc.

Page created in 0.044 seconds with 22 queries.