Does anyone have current scientific (not anecdotal) data regarding humans' ability to detect phase coherence with regard to loudspeaker crossover design and associated phase between drivers?
Thank you very much!
Thank you very much!
I know there is a study here: http://lib.tkk.fi/Dipl/2008/urn011933.pdf. I guess its conclusion is that sometimes phase distortion can be heard, but it's quite a subtle effect and not all people hear it. Perhaps people with an AES subscription might reference something better.
I guess it's possible to set up an ABX test for yourself and find out whether you can hear it, but the advent of FIR equalizers makes this problem kinda moot. If you add EQ to the frequency response, might as well equalize the phase flat, right? Note that with loudspeakers one can't really isolate the effects of phase, as different crossovers will have different power response. This is certainly way more audible than phase.
I guess it's possible to set up an ABX test for yourself and find out whether you can hear it, but the advent of FIR equalizers makes this problem kinda moot. If you add EQ to the frequency response, might as well equalize the phase flat, right? Note that with loudspeakers one can't really isolate the effects of phase, as different crossovers will have different power response. This is certainly way more audible than phase.
It's so easy, it's logic and it's scientific.
The 2 or 3 drivers emit in phase, each has the same content SO the final mix GOT to be in phase. It doesn't ? Why? do the adiacent drivers have a phase shift so some frequencies get annihilated by strong interference? Does this happen especially at crossover? What do the theory say? What happens after a 1/2 second after the sound is emitted in a room? What happens to the lower frequencies which result to be 17 m long waves at 20 Hz and 1,7 cm at 20 kHz, can you hear 180° phase shift at 20 kHz? Can you detect it @ 20 Hz? I guess that at 20 Hz you can detect it, specially at high power, where the amplitude shows more the interference produced.
Dave
You could probably specify a little better what you want to know: Are you talking about the transient behaviour of the complete speaker and its audibility (what the Finnish paper refers to). Or do you mean the off-axis peaks and dips caused by inter-diver phase offset ?
To make things more complicated : The constant-voltage crossovers that offer a "perfect" transient response on-axis have the greatest inter-driver phase-shift of them all, resulting in severe off-axis peaks and dips.
Regards
Charles
You could probably specify a little better what you want to know: Are you talking about the transient behaviour of the complete speaker and its audibility (what the Finnish paper refers to). Or do you mean the off-axis peaks and dips caused by inter-diver phase offset ?
To make things more complicated : The constant-voltage crossovers that offer a "perfect" transient response on-axis have the greatest inter-driver phase-shift of them all, resulting in severe off-axis peaks and dips.
Regards
Charles
I think OP is talking about getting the drivers' phase to line up and overlap.. I didn't know it was necessary.
It's necessary if you want them to sum at the crossover frequency. That way, your primary lobe is on the measurement (and hopefully listening) axis.
Once the phase tracks well through the crossover frequency (and half an octave or so either side, depending on the slopes), the phase curve of the resulting speaker will be non-flat.
What I do at the moment is to use a passive crossover to integrate the drivers, and then an FIR filter to linearise the phase response, plus a couple of IIR filters in the LF region.
Chris
Once the phase tracks well through the crossover frequency (and half an octave or so either side, depending on the slopes), the phase curve of the resulting speaker will be non-flat.
What I do at the moment is to use a passive crossover to integrate the drivers, and then an FIR filter to linearise the phase response, plus a couple of IIR filters in the LF region.
Chris
That's if...
And I say this to make a point, considering I doubt OP is going to find any proof on this.
By dropping crossover frequency 😉 When the crossover frequency is below 1/2 wavelength distance of the drivers the lobes start to disappear.
You can simulate lobing with Sound Wave Interference Applet for example.
You can simulate lobing with Sound Wave Interference Applet for example.
Last edited:
I don't have my copy of Toole's book with me, but I am pretty certain that he presented evidence that humans can only detect phase differences when listening to test tones with headphones. With program material coming from real speakers in real rooms, there was no statistical evidence that trained listeners could hear the difference between phase coherent time aligned speakers and those with high order crossovers.
I truly hope someone links us to such research, as i'm always on the watch for this.
So far, all i've seen is very dated studies when today's comparative tools and methods and weren't available.
(If Toole did any real phase-audibility research in valid isolating testing conditions, I've yet to see it..)
Or i've seen studies rooted in near ridiculous comparisons, like the paper linked in second post for example.
That paper rests on audibility of a >500th order xover at a single 3kHz frequency, with maybe even with some delay thrown in. Just some kid needing a master's thesis project. Lol
Anyway, I'd love to see some real science.
Because relative phase audibility truly needs light of day, whatever falls out to be true .. imho.
Because if coherent phase is not audible, why even look at an impulse response? That's all it really depicts..... plain ole phase response.
Why work at getting something right that doesn't matter?
(By 'phase response', I mean both relative phase between drivers and geometric time alignment between drivers, such that 'phase response' represents a complete coherent alignment in the time domain.)
Same thing for step response, why bother unless coherent phase is audible ?
As step is just frequency response and phase together in one plot...
Same thing with wavelets, etc, etc...
We (including myself) are often either paying way too much attention to measurements that don't matter,
or way too little attention to an all important fundamental property of reproduction...
Floyd Toole : pg419
"Many investigators over many years have attempted to determine whether phase shift mattered to sound quality (e.g., Greenfi eld and Hawksford, 1990; Hansen and Madsen, 1974a, 1974b; Lipshitz et al., 1982; Van Keulen, 1991). In every case, it has been shown that if it is audible, it is a subtle effect, most easily heard through headphones or in an anechoic chamber, using carefully chosen or contrived signals. There is quite general agreement that with music reproduced through loudspeakers in normally reflective rooms, phase shift is substantially or completely inaudible. When it has been audible as a difference, when it is switched in and out, it is not clear that listeners had a preference."
-------
I have tried this myself. I EQ'd the speaker FR flattish (smooth, non-min phase), then used a seperate phase correction convolution (RePhase) that that I can turn on/off without effecting the FR and I listen to any differences. They are subtle
"Many investigators over many years have attempted to determine whether phase shift mattered to sound quality (e.g., Greenfi eld and Hawksford, 1990; Hansen and Madsen, 1974a, 1974b; Lipshitz et al., 1982; Van Keulen, 1991). In every case, it has been shown that if it is audible, it is a subtle effect, most easily heard through headphones or in an anechoic chamber, using carefully chosen or contrived signals. There is quite general agreement that with music reproduced through loudspeakers in normally reflective rooms, phase shift is substantially or completely inaudible. When it has been audible as a difference, when it is switched in and out, it is not clear that listeners had a preference."
-------
I have tried this myself. I EQ'd the speaker FR flattish (smooth, non-min phase), then used a seperate phase correction convolution (RePhase) that that I can turn on/off without effecting the FR and I listen to any differences. They are subtle
Last edited:
The question seems quite specific. There's hopefully enough here to keep you amused Phase, Time and Distortion in Loudspeakers
What quality? We are talking about sound, its origin. The 'subtle' may refer to 'quality', but at this stage the only quality that matters is filter Q
Sound quality as defined by listening impression, not the "Q" quality factor used filters. If you have good phase alignment (no excess phase) your speaker can produce a reasonably clean impulse (transient) response. As most multiway speakers cannot do this, the question is fair "is it important"? Can you hear the sound difference and is one preferred over the other.
The "subtle" (IMO) refers to minor sound quality differences. Like does the voice sound more emotive? or does a reed instrument sound more accurate. My trial was not blind, but I found the differences to be small (ie. subtle). I should add that these tests are usually on a mono signal (one speaker).
The "subtle" (IMO) refers to minor sound quality differences. Like does the voice sound more emotive? or does a reed instrument sound more accurate. My trial was not blind, but I found the differences to be small (ie. subtle). I should add that these tests are usually on a mono signal (one speaker).
Last edited:
It's a vector sum. It's complex (pun intended) addition, but it's not that complicated.
This is the question. Group delay at low frequencies is probably more audible than system transient response when listening to music.
Thx DonVK,
Yep, i've seen that quote many times, many places. It's even in the excellent Elliot Sound link subsequently posted.
For me however, that's more anecdotal evidence than not,
that has unfortunately turned into parrot-repeat.
I think the investigations quoted are dated to when it was simply too difficult, too impractical, to achieve good science re phase audibility.
2000 is about the earliest I'd trust any relative phase audibility research;
and really, pragmatic dsp comparisons probably started becoming feasible in the last decade i think.
And i don't believe Toole did any definitive research of his own here.
As well as Harman documented their listening preference research, if they had done any real phase research it stands to reason they would have touted the actual tests and the their methodology.
Because it's not easy to make fair and valid comparisons.
It's cool you've worked on such for yourself. I can't count the number of times I've compared minimum phase tunings with linear phase tunings.
Like you, in mono, seamless switching, and outdoors whenever possible.
I don't think the headphone tests where folks apply all-pass phase warp to compare to flat phase work.
Because ime a major benefit of flat phase (or group delay elimination) is the timing relationship of strong bass bursts with their accompanying higher-in-frequency transients or harmonics.
Headphones simply can't convey the impact of bass transients like speakers can, imo.
But i digress....let's hope for some recent research to hit this thread...
Whether the research is presented as audibility of relative phase, or audibility of group delay, of course makes no difference as they are the same thing....
Let's see it! 😀