Physical time alignment of drivers

@CharlieLaub:
Could you point to the research supporting your claim about the brain combining reflections >4,5ms together please?
I'm interested.

It's not my experience ( and it can easily be demonstrated too) so would like to know more about it, maybe you made a shortcut describing this research?

Don't take me wrong, what you describe might work but from something different than that particular point imo.
 
In general these claims are following Linkwitz but are supported by well established research. There is a web site by Richard Taylor that has some stuff, like this page:
https://rtaylor.sites.tru.ca/2013/05/08/loudspeaker-placement-in-small-rooms/

To read up on this subject start with the Wikipedia entry for the precedence effect. The authors J. Blauert and R.Y. Litovsky have published on this general topic (see link below). The biggest confounding factor IMO is that pretty much all this research on precedence and so on uses two laterally separated sources (e.g. right and left speakers, 45deg apart), one delayed and possibly attenuated. The tests attempt to identify the threshold at which two distinct sounds can be perceived and so on. The problem is that this is not at all what is happening with a dipole loudspeaker reflecting off of the front wall. With the dipole the first arrival and reflection appear at the exact same angle from the listening position and the reflection will be relatively lower in SPL because it travels a longer distance to the listener. I do not know of any research that has looked at that setup exactly. Typically special test tones such as impulses or clicks are easier to differentiate as distinct sounds compared to music type signals. Trying to triangulate the published research with what is going on with a loudspeaker is a bit of a challenge.


In any case here are a couple of papers you can digest:
"Investigation of the relationship among three common measures of precedence: Fusion, localization dominance, and discrimination suppression." PDF is available online here: https://www.cmu.edu/dietrich/psychology/shinn/publications/pdfs/2001/2001jasa_litovsky.pdf
"The Precedence Effect in Sound Localization." Full article available online here: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4310855/

A very good review of human hearing with some relevance to loudspeakers in room may be found in the attached course lecture notes.
See e.g. the section "The precedence effect" on page 6. Look up the references cited at end.

Also I have attached a Litovsky review on the Precedence effect.

This is just hinting to the many other papers that are out there...

Perhaps others can chime in with relevant papers on the subject.
 

Attachments

Last edited:
From the attached lectures notes of post #43:
Audibility of a single reflection
The audibility of a single reflection depends on the level, delay-time, direction,
and signal type. For speech in a living room (reverberation time 0.8 s), echo
disturbance is observed for a single reflection if the delay time exceed approximately
30 ms, see [3], (reflection 0 dB re. to the direct sound and both arriving
from the frontal direction
). The threshold is about 10 dB lower if the reflection
arrives from about 45 degrees.

The above has one example of a reflection coming from the same angular position as the initial sound. If we can equate speech with music, you can see that these sounds are fused together up to 30msec in this case. But as I highlighted in blue, this is for the case that the reflection has the same SPL as the direct sound. For a loudspeaker's reflection from the front wall, the SPL of the reflection will be less by at least 6dB, which should make this time even longer. So long that it is not happening in any domestic listening space.
 
  • Like
Reactions: GM and krivium
Yeah and its "easy" to hear, you hear only one sound while listening to speakers, the whole thing is perceived as one.

Its easy to hear multiple sounds, go outside, take 10 steps from the front door and clap your hands to hear two separated sounds, direct sound and echo from wall of the house.

Or, if you have dsp with your speaker system, delay tweeter some 10ms or so to hear two distinct sounds from mid and tweet. When delay is only few ms there is only one sound, the sound sources fuse as one. Changing delay of tweeter changes frequency response, which you will hear, perhaps some other qualities like changes in stereo image which you might also hear, or "blurry" sound, less definition to it, but its still just one sound, there is no echo.

Another example, take a portable bluetooth speaker and move it closer and further to a boundary, just holding it in your hand, and you'll hear effect of the nearby boundary. Its still one sound but the tone changes with distance, you hear change in frequency response with small delay strong reflections, the reflection fuses with the direct sound as one.

We gotta trust our ears, try to connect perceived effect to the studies and terminology 🙂
 
Last edited:
  • Like
Reactions: GM and krivium
Hi Charlie,
Thank you for your answer. I wasn't sure you talked about Haas effect.

I agree the dipole case is difficult to 'simulate' with example ( eg: through a delay/reverb plug ins). Been there, tried,... failed.

I agree with Tmuikku, there is some colorations happening and it was my point.

In case of dipole the difference in level as well as the polarity flip and attenuation at 90* must all concur to give the effect?

I should try to see if i can dig up the files of the 'simulation' i tried to achieve ( shared them with member BButerfield at the time) it made some interesting example of what ER does ( headphones mandatory).

It made an interesting case on how our brain react to louder signal too as i had to match all files to the softest one generated as i kept coming to the 'worst' one in first rank without it...

Apologize for the OT.
 
Just a quick comment as I don't follow these threads much anymore. Form my point of view, time alignment is the physical alignment of the driver, without crossover, such that the impulse response of each driver arrives at the design point at the same time (or with the same delay). Phase alignment is alignment of the drivers with crossover in place such that the phase difference between drivers, matches the matches the text book phase difference between them for the crossover chosen.
I share that point of view.

Although maybe more honestly said, i don't think it's even a 'point of view'....
I think the distinction between time alignment and phase alignment is such a fundamental concept ....., that it is a 'do-not-pass-go concept' in xover and speakers design, until it's clear.
 
the distinction between time alignment and phase alignment

1234w.png
 
I don't know what you are trying to depict...? Plz explain...🙂

What I think I see is an electrical simulation. One where frequencies are properly in phase with each other.
What I don't think i see is any evidence of time alignment. Like any delays needed for physical offsets, for acoustic centers distances.
 
LC cells as allpass to 'compensate' ( induce delay) to the three drivers ( r1/2/3), so it take care of physical offset.
The physical offset could be estimated through the graph but resolution is a bit coarse though.
 
Last edited:
  • Like
Reactions: AllenB
LC cells as allpass to 'compensate' ( induce delay) to the three drivers ( r1/2/3), so it take care of physical offset.
The physical offset could be estimated through the graph but resolution is a bit coarse though.
Thx krivium,
If that is just a series of all-pass (i really am a LCR ignoramus...and intent to stay that way haha 😀),
it is effecting phase rotations.

I wonder how much delay, in terms of milli or microseconds, do those all pass achieve?
And at what particular frequencies would the times come from?

They're just not a good form of constant delay, imso.
 
  • Like
Reactions: krivium
LC cells as allpass to 'compensate' ( induce delay) to the three drivers ( r1/2/3), so it take care of physical offset.
The physical offset could be estimated through the graph but resolution is a bit coarse though.
Thanks krivium, that is what I was trying to show.. The frequency is 1kHz and the low pass filters are -6dB at that frequency. Each order adds 45 degrees, so 0.5ms for the fourth order filter.
 
  • Like
Reactions: krivium
I share that point of view.

Although maybe more honestly said, i don't think it's even a 'point of view'....
I think the distinction between time alignment and phase alignment is such a fundamental concept ....., that it is a 'do-not-pass-go concept' in xover and speakers design, until it's clear.

The reality is that these are all old school definitions. The bottom line is that when designing a speaker a crossover type is chosen, be it 1st order Butterworth or 8th order LR. Filters are then designed such that the acoustic response of each source matches the theoretical response of the chosen crossover filter as closely as possible. Then the drivers are positioned or delay added such that the inter driver phase difference in the crossover region matches that of the chosen crossover. Call it time alignment, phase alignment, alignment of acoustic centers,.....

Old school wise, systems were designed with arbitrary HP and LP filters that didn't necessarily sum flat on axis. The designers would dick around with filter types and look at the summed response, on axis, and play with driver off sets to try and get the best looking impulse response while achieving reasonable flat response. Spica was one manufacture that took this approach. Since the speaker produce a reasonable impulse they were referred to as being time aligned. These were designs that were based only on amplitude response without consideration of phase.


spica-tc-50_550.jpg
 
The reality is that these are all old school definitions. The bottom line is that when designing a speaker a crossover type is chosen, be it 1st order Butterworth or 8th order LR. Filters are then designed such that the acoustic response of each source matches the theoretical response of the chosen crossover filter as closely as possible. Then the drivers are positioned or delay added such that the inter driver phase difference in the crossover region matches that of the chosen crossover. Call it time alignment, phase alignment, alignment of acoustic centers,.....

The bottom line you describe is how I see it too.
Once the acoustic response target is chosen for a source, the target matching for the source is without regard to constant delay.
Once each source has independently met its acoustic target, it's simply a matter of either time delays or physical offsets to complete the total alignment.

I should probably add, I view acoustic target responses to be necessarily fully complementary, unless some particular beam/lobing steering is purposely being used.

Can't say i see the definitions as old school, though.
The idea that time alignment means constant time, that is not frequency dependent;
and that phase alignment means relative phase rotation between frequencies, and is totally frequency dependent...
are simple fundamental definitions that I think fat too many people get confused over.

I mean look at how many posts there are about using all-pass or xover induced group delay as a substitute for constant time...
That's confused in my book....



Old school wise, systems were designed with arbitrary HP and LP filters that didn't necessarily sum flat on axis. The designers would dick around with filter types and look at the summed response, on axis, and play with driver off sets to try and get the best looking impulse response while achieving reasonable flat response.
Haha...exactly !!
It's the kind of juggling around, that keeps designs from reaching greater potential, imo.

I often think half the debate over whether or not can hear phase / transient perfect response / group delay / yada yada....
and half the differences in studies conclusions...
is more about the vagaries in comparative phase listening tests, when trying to use all the juggled / compromised designs out there.
It's like trying to measure straight using a crooked ruler !
 
Thx krivium,
If that is just a series of all-pass (i really am a LCR ignoramus...and intent to stay that way haha 😀),
it is effecting phase rotations.

I wonder how much delay, in terms of milli or microseconds, do those all pass achieve?
And at what particular frequencies would the times come from?

They're just not a good form of constant delay, imso.

Yes they are (series of all pass in Allen's example and effecting phase rotation).


They are kind of constant delay below a (more or less) defined threshold which will depend mainly on order of the implemented allpass and it's associated group delay ( which will define the max delay availlable).

See first graph linked

In fact it'll depend of implementation and needs and the associated compromise you are willing to accept:
In the second graph you can expect a relatively constant behavior below ~200hz and above ~5khz. If you plan to use this filter in between those two value then yes, it's not very constant delay generator i agree ( it'll 'work' at a narrow range of frequency).

The main issue i see in a passive implementation are: the relatively low delay value availlable (if low order is implemented which is very likely the case) and the increase in cost ( depending on self and capacitors value - we could add another weak point about capacitors about the technology used for anything bigger than 100uf which could bother some).

Could be implemented active analog though, which lower costs using the kind of circuit in third image linked both type presented 'leading' and 'lagging'.

This is this kind of cells which Tannoy used in the SystemX00A(ctive) circuitry to compensate for Z offset between both drivers of concentric units used in this serie ( 600 and 800 see same issues but at different freq).

They choose to implement them using the 'constant' group delay region too by doubling the freq implemented ( wrt to the target frequency) and using 2 in series ( one of both kind linked) to have more delay availlable and keeping resulting phase relationship in the xover region.

You can see the schematic here and the allpass cells are ic3b and ic2d:
https://www.diyaudio.com/community/threads/tannoy-system-800a-crossover.336733/#post-5762360
 

Attachments

  • 3-s2.0-B978185617505000020X-gr42.gif
    3-s2.0-B978185617505000020X-gr42.gif
    20 KB · Views: 76
  • Fig02Summation.jpg
    Fig02Summation.jpg
    53.2 KB · Views: 68
  • images.jpeg
    images.jpeg
    7.1 KB · Views: 60
Last edited:
They are kind of constant delay below a (more or less) defined threshold which will depend mainly on order of the implemented allpass and it's associated group delay ( which will define the max delay availlable).
Thx again krivium,
Yes I'm familiar with all-pass characteristics.
(Just couldn't recognize a LCR all-pass implementation)


Could be implemented active analog though, which lower costs using the kind of circuit in third image linked both type presented 'leading' and 'lagging'.
Meyer uses all-pass in their active analog. I watch a long vid of theirs on how the UPA-1p was set up, which used a couple or so of them.

Also had some prosound buds show me how to string a boatload of all-pass together in dsp, to effectively flatten phase across the spectrum.
And I do mean a boatload Lol. Tricky it was.
There was never any idea to also use them for delays between driver sections though, because with dsp delays why on earth would you?
 
Mark, yes i know you are familiar with them, but you know: i've been told i like to please my ego so every occasion i can take to do it.... i happily do. Lol.

More seriously it is in case someone follow the subject without as much understanding of this as members in here.

Yes active analog (as well as passive) have their use. Even after dsp.

I agree delay brought by LC cells are limited in range availlable so of almost zero interest as you go lower in frequency imo. But theorically it could be used that way too.

The interest of doing it other than digital for delay is...? It is true about any attempt to delay anything in analog anyway ( except if you want to add 'character' or really modify a signal like with BBD or Tape Delay Line).
 
Oh, I'm very glad you gave such full information....yes, it helps others follow along, and heck it often helps us simply communicate more clearly.
Besides, how many zillion times have i repeated my basic tuning mantras....

Pls don't think for a second i was suggesting less info. 🙂
 
  • Like
Reactions: krivium