I don't know why you philosophize a tweeter (especially one encased for car) radiated backwards or even at 90deg without high attenuation. I also crossed high ~4khz series 1st-order and took advantage of both drivers' natural fall-off past XO, largely breakup/resonance-free being respectively wide-band and ceramic dome (which went fairly low as was typical). And the effective center-to-center distance with bounce was within 1/4 wavelength at XO, rather than 10X that threshold. As I said, frequency response above 12khz exceeded my hearing confidence; up to 11.5khz I heard no significant unevenness for a broad area of head positions and proper angles.
One could imagine all kinds of reasons in order to dismiss an idea that is outside of one's experience. How does the saying go? First the experts all say the idea is wrong. Then they say it is right but not important. Finally they say it is both right and important, but they had known it all along!
p.s. 15" Tannoy Dual-Concentric would be too dangerous for me to try to lift. Would large high-efficiency pro-sound coaxial do better near-field/mid-field? Without DSP/EQ/custom active XO? Maybe someone can make a comparison or recommendation.
Last edited:
Well, don't take concerns as personal attacks wchang: our ears are easily tricked.
Without measurement all we can say is like NC535 said only an opinion, a subjective assessment. Which by itself is fine but without evidence ( measurements) ...
Eg many listeners find perfectly fine and acceptable ( sometime raving about) the sound of their tv... where drivers use wall as reflector and where there is nothing past 8khz...
Without measurement all we can say is like NC535 said only an opinion, a subjective assessment. Which by itself is fine but without evidence ( measurements) ...
Eg many listeners find perfectly fine and acceptable ( sometime raving about) the sound of their tv... where drivers use wall as reflector and where there is nothing past 8khz...
Hi,
also I'm not dismissing your idea wchang, I just do not understand how it would work because sound behaves the same no matter what we do and thats what I'm imagining. The way I understand the concept from your description could be different what you have actually implemented.
also I'm not dismissing your idea wchang, I just do not understand how it would work because sound behaves the same no matter what we do and thats what I'm imagining. The way I understand the concept from your description could be different what you have actually implemented.
Some data, 4kHz xo and 1/4wl bounce means 8.6cm / 4 round trip at max, right, so tweeter dome is within about 1cm from the big woofer dustcap? The tweeter is in front of the woofer cone, coaxial, right? How big is the tweeter structure, 2.5cm?
I want to illustrate what I've been writing, I have not been dismissing your idea but trying to express that there is no way two different physical objects would not affect each other acoustically and didn't understand your setup in regards of the topic.
Earlier I thought your tweeter was off-axis, out of woofers way, which would make the reflection very late and would require the tweeter to beam (have big structure) in order to prevent the direct sound hit ear first. However your last post hints the tweeter is right at the dust cap. Following illustration plays around with the direct sound of tweeter, and a reflection of the woofer cone. The fact that the tweeter is so close means it doesn't have to beam and you'd benefit as small as tweeter as possible.
Since tweeter is so close to the dust cap, the reflection and direct sound diffracting around the tweeter would be within 2cm as per above math. Ideal reflection of tweeter on-axis sound 2cm late with direct sound that diffracts from the tweeter directly to listener (without reflecting from the cone), and ignoring the tweeter being there in the way physically obstructing the reflection, gives quite good response.
Here is ideal 25mm tweeter in housing that is no bigger. It's got some directivity for the bandwidth, around 4kHz its about 6db down to 90deg or 180deg:
If I rotate reference angle to see what's the sound exactly behind:
Here both sound 0deg and 180deg combined, as if the on-axis sound of the tweeter had reflected ideally and has no extra path length to it. On this ideal simulation this results flat response.
If the reflection was from object 10mm away, it makes 20mm round trip extra path length for the reflected sound. Still pretty good since there is constructive interference and above that the 180deg sound fades off so there is no too bad interference ripple in the summed response.
However, if observing from off-axis, say 40deg, and speculating the reflection is not 2cm away but bit further (the tweeter is not at focal point of the reflector), then something like this happens, the top end starts to get ragged. This would not happen if the tweeter was there pointing at you without the reflector.
If you did the same in a BEM simulation, the tweeter structure would obscure the reflection and make the response vary to various directions on the whole tweeter pass band, less so on the bottom of it and mostly on the top, sound diffracts and reflects from it and the smaller it is the better in this regard. This is not special to this example setup, but with any objects, it's feature of sound, wavelength.
If the tweeter structure was inside the woofer these reflections and diffractions wouldn't be there, but the woofer acts as waveguide more or less successfully so this typical setup also has issues in the vein of the topic. Hopefully this illustrates what I've been thinking while writing the posts, a tweeter could be further or closer from the woofer, pointing any which way, and the woofer would still affect tweeters response being there at close proximity, and the tweeter would affect woofers response if disabling use of dust cap, hence coaxials have ragged response.
Now, this doesn't mean your setup wouldn't sound good, and your setup might be completely different than what I tried to emulate here, even though it likely has as ragged response as any coaxial implementation. If I may speculate having very small tweeter in front of very big woofer has about no effect on the woofers response, and the woofer might have nice response since the dust cap is implemented. In this sense it's very different implementation that most coaxials. I speculate your 4kHz xo means the woofer makes most of the meat for good sound preserving the phase info as single point source, and the tweeter would be there to distribute more HF energy everywhere to make the beaming of the woofer less obvious and sound more balanced in general. Perhaps you can achieve more even sound this way than by using single fullrange driver. Phase information isn't important anymore at that high frequency so it doesn't matter so much there is interference due to reflections and diffraction. Perhaps shape of the dust cap is such there is not much reflection from the woofer toward off-axis reducing the interference I speculated above. Compare to situation where the tweeter is in place of the dust cap pointing out, like it typically is, you'd likely have the woofer breakup worse so likely lower crossover and so on, so different sound.
If you found good new setup that's cool and I'm really happy for you 🙂
Last edited:
Comparing to tweeter suspended at same location, but pointing towards the listener like in many coaxial speakers would reduce the high frequency interference what I tried to illustrate above, tweeter being in the way of HF reflection when it's pointed to the woofer dust cap. But the trade-off is that one cant get a real world tweeter as close to the cone so more interference would be lower in frequency as trade-off. Unless the tweeter is embeded in the woofer, in place of a dust cap. These are all continuum of same thought process with interference/raggedness visible in measured response, attempt to make less. From technical point of view it makes sense, but perhaps the trick is to use high crossover and not mind about the HF interference too much, perhaps there isn't that much that it would bother.
(They are time-aligned per your assumptions.)Here both sound 0deg and 180deg combined, as if the on-axis sound of the tweeter had reflected ideally and has no extra path length to it. On this ideal simulation this results flat response.
I appreciate your attempt at simulation and the long write-ups. I had a high school teacher, MIT PhD physics, who told us the formulae were but successively finer approximations. Here, I think the first-order effects are time-alignment with the woofer (which you inadvertently simulated, sort of), and phase-alignment around the XO. In a first-order approximation, the dustcap bisects the distance between the drivers' acoustic centers, so the tweeter bounce at the dustcap is equally delayed as the midwoofer, so they are time-aligned as if the tweeter were well-back of the cone. This placement of the tweeter was estimated and empirically nailed down with a test-tone (of any frequency as approximation goes), to maximize the combined SPL. Then the XO was inserted and tweaked if necessary, to ensure phase-alignment around the XO frequency, again by moving the tweeter listening for maxima/minima while playing XO frequency (yes I got a null half-wavelength from max). This is all intuitive common sense.
Then the second-order effects (approximation) of impedances, filters (series 1st-order electrical as well as acoustic), individual and combined phase response etc. -- result in frequency response and transcient response that again can be readily heard using tone-sweep and appropriate music.
Third-order effects (not less important, but harder to quantify and control) off-axis directivity/dispersion, center-to-center-distance/comb-filtering, etc.
Finally (in my opinion) lower-order effects and issues that can be reasonably mitigated: diffraction, reflection, intermoduation, doppler etc.
The tweeter I used is unobtainium as standard-35mm-diameter-casing nominal 4ohm 92dB auto tweeter, but still available with face-plate nominal 6ohm (used in my stacked "lumin77" and "Maevan" TLonken vs WAW comp); I just bought a spare. The car pair can be seen front and back in the photo.
Last edited:
Maybe i don't get what you stated ( might be lost in translation) but i see issue with what being put in bold.
What works for frequency doesn't garantee you it's true for transients ( time domain) : if you have 360* ( or multiples of it) delta you'll have a maxima/minima in frequency but transients can be delayed by one ( or many more) cycles.
Time and frequency domain can't really be analyzed using same technique. At least not in the way i understand what you did ( and despite it seems common sense). And again our ears are hopeless to define an issue with time domain issue ( for low values, of course past a treshold we start to hear an issue but ime, it can be very long until we find it objectionable with 'real music' and it is freq dependant).
A step response could tell but you'll need full 2 chanel measurement set up with time reference ( to nullifying time of fly and other aberations included in a one chanel measurement).
Another way to approach this would be through wavelet analysis at xover freq ( it works surprisingly well ime).
https://www.diyaudio.com/community/...ind-delay-phase-and-polarity-at-xover.370287/
Thank you very much for the suggestions. I had thought about the issues you raised, if I understand correctly --
Having first done time-alignment of acoustic centers in the ~4khz frequency range (~8cm wavelength) was prerequisite to attempting phase-alignment; impossible to be off by a full cycle. An assumption is that in the critical XO frequency range, but without XO filters involved, the drivers are close enough to linear that acoustic centers are well-defined and can be made approximately coincident. One question is, what if they have very different phase in a nonlinear way. Then maximizing sum shouldn't work. In practice though I haven't encountered it (and it would make perfect transcient theoretically impossible I think). Then, keeping track of the time-aligned driver offset, XO phase could be aligned (by tweaking XO if necessary). I try to do 1st-order which helps with linearity. The result I get from all this is much clearer "depth" than just phase-alignment without time-alignment. As for "perfect transcients", frequency response will have to be nearly flawless too.
Having first done time-alignment of acoustic centers in the ~4khz frequency range (~8cm wavelength) was prerequisite to attempting phase-alignment; impossible to be off by a full cycle. An assumption is that in the critical XO frequency range, but without XO filters involved, the drivers are close enough to linear that acoustic centers are well-defined and can be made approximately coincident. One question is, what if they have very different phase in a nonlinear way. Then maximizing sum shouldn't work. In practice though I haven't encountered it (and it would make perfect transcient theoretically impossible I think). Then, keeping track of the time-aligned driver offset, XO phase could be aligned (by tweaking XO if necessary). I try to do 1st-order which helps with linearity. The result I get from all this is much clearer "depth" than just phase-alignment without time-alignment. As for "perfect transcients", frequency response will have to be nearly flawless too.
Last edited:
Are you sure where the emmissive center of your woofer is at 4khz?
I mean i already tried to define it on a number of driver and it was'nt easy.
That said Dunlavy's work prooved it can be done ( even on vertically stacked pairs of drivers).
To be blunt if you want to go this way (time alignement) i would forget passive and go dsp from start. Don't take me wrong it can be done but it's so much hassle even Dunlavy followed this way at the end of his life ( despite the 'Magnus' never seen completion as he went too sick to finish them... but dsp/FIR based they were).
Wchang, don't you have measurement setup? Imho you should invest in it given your experiments ( not something fancy, a 2 chanels soundcards and a measurement mic).
You won't have buggers like us teasing you! 😉
I mean i already tried to define it on a number of driver and it was'nt easy.
That said Dunlavy's work prooved it can be done ( even on vertically stacked pairs of drivers).
To be blunt if you want to go this way (time alignement) i would forget passive and go dsp from start. Don't take me wrong it can be done but it's so much hassle even Dunlavy followed this way at the end of his life ( despite the 'Magnus' never seen completion as he went too sick to finish them... but dsp/FIR based they were).
Wchang, don't you have measurement setup? Imho you should invest in it given your experiments ( not something fancy, a 2 chanels soundcards and a measurement mic).
You won't have buggers like us teasing you! 😉
Let me first report that bounce-aligned was clearly more "holographic" than LX; whereas unaligned was "flat". What I mean by these terms is that, playing monophonically, a clear sense of spatial projection of everything present, a somewhat palpable, dimensional, floating "soundstage" ("acoustic space") out beyond the speaker, depth and height well-defined but not horizontal localization. (Playing stereophonically adds horizontal.) This "image" is hard to depict in words but easy enough to reproduce using a single coherent speaker, such as a fullrange driver, especially if firing upward. A less-well aligned 2-way speaker will not have this effect. Like stereo imaging with depth, it's there or it's not (please try it).
Last night I specifically compared tweeter/listening angles: bounce-aligned (tweeter toward dustcap, listening on-axis or on the unobstructed side) the best holographic projection I had ever heard; LX time-aligned (tweeter toward listener near 90deg off-axis) good but not as good; unaligned (tweeter forward toward on-axis listener) was just flat, no sense of space.
Last night I specifically compared tweeter/listening angles: bounce-aligned (tweeter toward dustcap, listening on-axis or on the unobstructed side) the best holographic projection I had ever heard; LX time-aligned (tweeter toward listener near 90deg off-axis) good but not as good; unaligned (tweeter forward toward on-axis listener) was just flat, no sense of space.
The necessary winning condition, I think, is aligning the drivers' acoustic centers as opposed to measuring precisely where they are (if well-defined in the first place). They may depend on frequency, but within the target XO region they best not vary too much.
@AllenB please help! How is acoustic center simualated @tmuikku?
I don't have the proper consistent physical environment (I live at work) but this is a problem I really must solve cost-no-object.
If you want to read a story someone has written about time alignment sounding different, then you must check they did it right. You shouldn't be able to hear differences in typical small time alignment changes by itself.
1. The phase must be correct in both cases. It must compensate the delay electrically (yes, this is possible). Not every case confirms this before declaring that it sounds different.
2. The total response needs to be the same. When delay is compensated for electrically, you need to ensure the response doesn't also change.
3. The baffle needs to be the same. This is unlikely in practice and will result in power variations so you should try to avoid sharp edges at least.
When compensating electrically, you can manage both normal delay and group delay simply by looking at phase. Vituixcad normally shows you both anyway so you can see what you have.
1. The phase must be correct in both cases. It must compensate the delay electrically (yes, this is possible). Not every case confirms this before declaring that it sounds different.
2. The total response needs to be the same. When delay is compensated for electrically, you need to ensure the response doesn't also change.
3. The baffle needs to be the same. This is unlikely in practice and will result in power variations so you should try to avoid sharp edges at least.
You will see the variations in a plot of group delay.
When compensating electrically, you can manage both normal delay and group delay simply by looking at phase. Vituixcad normally shows you both anyway so you can see what you have.
Quick reply: conditions 1-3 imply the time-aligned and unaligned speakers must use the same baffle but different XO filters so that both have the same frequency response; phase response would necessarily be different but both should be phase-aligned at their XO frequency (is this right?). To isolate time-alignment as responsible for any differences, the experiment needed to control for all other factors.
Last night's listening comp used only one XO and clearly frequency response differed (especially suboptimal for LX; even at >45deg the bounce was preferred over LX). Back when I had the same tweeter forward with small whizzer (maybe a couple years ago), phase was aligned at XO and frequency response was flat enough within my hearing range -- believe me it was not holographic like the bounce -- but the XO was different, 15" straight-through not 1st-order. Yes this is important so I need to re-do it with 1st-order.
There is no question in my mind the requirements given in msg#246 are necessary conditions to achieve fidelity. I'm dubious though, that time-unaligned could be made as flat and transcient and phase-coherent and musically nuanced using my current toolkit, as time-aligned.
Last night's listening comp used only one XO and clearly frequency response differed (especially suboptimal for LX; even at >45deg the bounce was preferred over LX). Back when I had the same tweeter forward with small whizzer (maybe a couple years ago), phase was aligned at XO and frequency response was flat enough within my hearing range -- believe me it was not holographic like the bounce -- but the XO was different, 15" straight-through not 1st-order. Yes this is important so I need to re-do it with 1st-order.
There is no question in my mind the requirements given in msg#246 are necessary conditions to achieve fidelity. I'm dubious though, that time-unaligned could be made as flat and transcient and phase-coherent and musically nuanced using my current toolkit, as time-aligned.
Last edited:
@tmuikku Please consider simulation 15" wide-band 94dB smooth roll-off above 4.5khz; coaxial ceramic tweeter 92dB response curve like Eton (published); tweeter at various offsets from behind the cone, coincident acoustic centers (i.e. assume ideal reflection), to in front of dustcap (i.e. bridged coaxial). Plausible rising impedance curves and phase response, 1st-order XO (series if possible) phase-aligned ~4khz. Maybe this would illustrate time-alignment vs not? Later, model dustcap-deflector as a widening of tweeter directivity. Still later model tweeter off-axis by 1/4 XO wavelength vs normal CtoC. Finally diffraction etc.
Makes sense or not?
Makes sense or not?
Hi, ideal drivers could be simulated in ABEC but haven't had time to do it yet. I've got already projects for 15" so it is matter of adding a tweeter there so it's not too much time to test this stuff. I'm quite sure effect of varying acoustic center is not included in BEM simulation, because it is at least partly due to cone breakup, so which parts of cone are at which phase at which frequency, and this information is outside BEM simulation. It would be possible to see effects of the "SBIR" and diffraction though.
VituixCAD could show you phase data and allow to play with the distance and see how the phase aligns, if you measure your real drivers with proper procedure. This would not show what happens with the reflection though, how the two objects affect each other sound acoustically when the distance is changed. You could try it quite fast though, make just axial measurements, measuring both the woofer and tweeter individually, then move the tweeter another location and measure again. This would not show off-axis behaviour but perhaps would give some information about the phase, which you are mostly interested in. In VituixCAD, you could load the axial measurements with various off-axis designations to exploit the reference angle functionality: look at SPL and Phase graphs, and use the reference angle adjustment box "to move the tweeter" and see what changes in real time 😉
To my knowledge reflections and diffraction also affect perception similarly as acoustic offset you've been adjusting. Ear detects sounds based on their harmonics, and when phase information comes through the system intact enough brain pays attention to the sound (and not consider it noise, doesn't pay attention). Basically, when phase information is preserved all the harmonics align every fundamental frequency cycle and make huge amplitude peak, which makes signal to noise ratio high and brain considers it important. The sound pokes through above all the sounds surrounding us (Stuff Griesinger writes about). At least in simplified terms, I'm not scientist on the field.
Thinking a bit: group delay, so crossover frequency and slope affects this stuff as well. All of these effects including early diffraction, reflections, distortion, that either rotate phase so that harmonics don't align, or add to the noise, reduce signal to noise ratio and likely affects the holographic effect in perception. Although, Griesinger hints brain either pays attentio to the sound or not, so perhaps there is no need to get much further than just make sure the effect happens, and everything is good enough now, brain pays attention and provides you the holographic effect. This happens with stacked multiway speakers as well.
With stereo system both speakers make noise to the other as they do not colocate affecting phantom images, so early reflections included, off-axis sound, so listening triangle positioning in room and directivity and how well the L and R speakers response match each other all affect some, toe-in. Perhaps this stuff is important to very high frequency as long as harmonics have meaningful amplitude to add to the peak. Even though ear didn't detect phase at very high frequencies, the amplitude peaks the phase would still affect.
VituixCAD could show you phase data and allow to play with the distance and see how the phase aligns, if you measure your real drivers with proper procedure. This would not show what happens with the reflection though, how the two objects affect each other sound acoustically when the distance is changed. You could try it quite fast though, make just axial measurements, measuring both the woofer and tweeter individually, then move the tweeter another location and measure again. This would not show off-axis behaviour but perhaps would give some information about the phase, which you are mostly interested in. In VituixCAD, you could load the axial measurements with various off-axis designations to exploit the reference angle functionality: look at SPL and Phase graphs, and use the reference angle adjustment box "to move the tweeter" and see what changes in real time 😉
To my knowledge reflections and diffraction also affect perception similarly as acoustic offset you've been adjusting. Ear detects sounds based on their harmonics, and when phase information comes through the system intact enough brain pays attention to the sound (and not consider it noise, doesn't pay attention). Basically, when phase information is preserved all the harmonics align every fundamental frequency cycle and make huge amplitude peak, which makes signal to noise ratio high and brain considers it important. The sound pokes through above all the sounds surrounding us (Stuff Griesinger writes about). At least in simplified terms, I'm not scientist on the field.
Thinking a bit: group delay, so crossover frequency and slope affects this stuff as well. All of these effects including early diffraction, reflections, distortion, that either rotate phase so that harmonics don't align, or add to the noise, reduce signal to noise ratio and likely affects the holographic effect in perception. Although, Griesinger hints brain either pays attentio to the sound or not, so perhaps there is no need to get much further than just make sure the effect happens, and everything is good enough now, brain pays attention and provides you the holographic effect. This happens with stacked multiway speakers as well.
With stereo system both speakers make noise to the other as they do not colocate affecting phantom images, so early reflections included, off-axis sound, so listening triangle positioning in room and directivity and how well the L and R speakers response match each other all affect some, toe-in. Perhaps this stuff is important to very high frequency as long as harmonics have meaningful amplitude to add to the peak. Even though ear didn't detect phase at very high frequencies, the amplitude peaks the phase would still affect.
Ok. Let me first say this 'holographic' thingy, i don't get. I heard even collegue sound engineers talk about it for microphones but... it doesn't mean anything to me.
Because even stereo can't be as for height informations to be presents you need at least a 'triangulated' recording device (minimum 3 capsules impliyed in 3d) and the same for reproducing system ( an Atmos system).
In each case you will hear height or elevation in mono or stereo then it's either your room playing a trick on you or a phase variation within the source signal ( some of my synth have filters which can give this feeling of elevation when filters open up) or the reproducing system ( interactions between drivers and or room giving aberation in rendering).
See my previous sentence about room/loudspeaker interaction. The effect you talk about is an illusion created by your room mainly.
Don't believe me? Bring your loudspeaker outside and test them to see/hear if it is still there when performing in hemi anechoic situation.
Stereo with depth is unrelated to the kind of loudspeaker used, it's related to the kind of microphone couple used and if it use delta phase ( A/B) or delta level ( X/Y) or a mix or both ( Ortf,Nos,...). Difference of phase technique will bring 'ambiance' but an undefined soundstage ( A/B or grand A/B, often used in western 'classical music' recording) and they are not monocompatible ( you'll loose information by phase cancelation once monoed), coincident technique ( delta level, X/Y, M/S, Blumlein) will bring a very defined soundstage ( except blumlein as it record an inverted image of what happen back to it's main axys) but with little ambiance. Those one are monocompatible.
Technique using both principle at once brings good from both techniques but are not monocompatible either.
Your loudspeakers can 'blur' the image created by this recording technique ( add distortions, and not only harmonic distortions, but things like compression or extension of initial position within stereo field, things like that) and this can be detrimental or euphonic ( bring enjoyment to listener) depending from listener.
When mixing/mastering we have tools at our disposal to manipulate things to give listener the impression there is more width, depth, things like that. So yes i'm aware of the 'effects' you can bring by modifying initial signal. As a result i can evaluate if loudspeaker ( or their interaction with room) bring (or take) something that is present in the initial signal recorded as i can compare real time to what microphone captures... and so i'm cautious about the effect you depict.
If there is no depth in a stereo recording then it's either a flaw with the mic set up, either there was no depth in the initial signal ( some recording room are so acousticaly 'dead' that there is no clues for depth of image to be created), intentionaly or unintentionaly.
Or it's a 'fake' stereo created from mono signals + effects ( reverberations). We can give the 'sense' of depth by manipulation of level+eq too. Even faking it by using Haas effect... we can be liars with a bag of tricks... 🙂
Your loudspeaker configuration, how and where they are located as well as your listening spot can bring or kill those effects to. As having your eyes open when listening ( i'm really serious about this*).
What were you listening? A record you made or a track you wasn't present when it was recorded? 😉
That's another reason i urge you to invest into a 2 chanel soundcard and a couple of mics: you can do your own records and can compare from known source.
You talked about cost is no object for measurement set up. It's not needed imho.
A measurement mic like B&K ( now DPA), Earthworks,... make sense if you have space which offer blameless condition to use them ( anechoic room, or very big sound isolated ones) with the whole recording chain on par.
For the cost of one of this mic you can have soundcard+computer+one 'standard' measurement mic ( Dayton or Behringer or if you want to spend a bit more Isemi) + a pair of small capsule with multiple directivity capsule + etc,etc,etc,....
This last pair of mic will let you try the different type of couple used for recording, get used to their rendering of stereo effect and from there you can record ambiance of place you know and where you can seat while recording.
Once done you can then compare results you experienced real life to the one you experience trough your loudspeakers in your room. It's usually enlightning.
I'm lucky to have a backyard in which i do this and i find it helped me to define flaws into my realisations. I'm lucky i can bring my loudspeakers too in there to know what the room does too.
So i'm not saying all this just to challenge you: it comes from my own observation too ( and theory as i used to teach this kind of things..., but nothing beats real life experience to validate theory).
* we are mostly sight based animals ( because of evolution and our world we evolved in), so 90% of our brain is dedicated to this sense in the treatment of information. If there is clue of depth in fromt of you ( window, large room with wall facing you away from you) subconciously you'll attibute a sense of depth to the reconstructed image of soundstage in your brain... it can come from physical clues from the room ( when you walk or moove in it) or by 'shortcut' in the interpretation of sight. We are easily tricked... believe me, i have been caught many times by this illusions...
Last edited:
Yes.
However as soon as you say 'time alignment', you kind of disqualify off-axis considerations. This is because time alignment is normally only done on one axis. Even if we check, it may not be time aligned on another axis.
Then if you change the position of the tweeter, you change the geometry. You might be altering the off-axis sound and that is just one more thing that you might hear.
In particular "... everything is good enough now, brain pays attention and provides you the holographic effect. This happens with stacked multiway speakers as well."
I think this is right, the ears-brain (and eyes-brain, see below) has evolved to snap-to-attention and localize predator/prey even amidst a background-noisy environment. The coherence of a sound not its loudness is probably what toggles the brain. Since measurement tools and methods haven't fought for survival over millions of generations (in fact only four human generations), stereo sound perception by machine analysis is primitive at best, compared to animals. When a stereo sound is both coherent and to a high-enough frequency (hence more directional and attenuated by distance), soundstage gains holographic, focused depth; otherwise it sounds flat like how a 2D picture is seen through both eyes (brain detects no parallax). This psycho-acoustic effect is maximally facilitated by so-called stereo-triangle placement with speakers toed-in to aim axially at the ears from front-L/R directions where hearing is most acute (more than straight front or 90deg side). When everything clicks the effect can be incredible. We all know this. Yet the sum-quality of stereo imaging has never been machine-measured -- only some putative component parts like frequency response and phase.
What has surprised me, is that monophonic sound can have depth well-beyond the speaker, in the sense of a psycho-acoustic effect (when things click, coherent phase and high frequencies both), a palpable spatial sense where every sound has its natural place in relation to another. I can detect this phenomenon if, and only if, the speaker is coherent (such as a decent fullrange driver, especially pointing up). Then when two such speakers play stereo the soundstage is spread out horizontally as well. The easiest (only?) way I know of to achieve this level of imaging is by the sequence of steps given earlier: first, time-align acoustic centers; second, tweak XO to align phase around XO frequency (1st-order being best for linear phase and amplitude) and to ensure flat high frequency response. Is it possible to time-align acoustic centers without correcting/normalizing both drivers' phase first? Well I checked the KEF LS50 Meta 1793 coaxial for acoustic-center alignment, by using two speakers one midbass the other tweeter, without XO, and confirmed test-tone max sum at zero offset (it's possible I got lucky by playing intended XO frequency). So I think time-alignment this way works.
Now stereo vision. Twenty-five years ago I serendipitously discovered a surprising psycho-visual effect, that monocular viewing of a picture with just one eye had a strong sense of depth. Back then I was involved in 3D stereo-photography (won an internet 3Dphoto contest for macro) so I had true mixed feelings -- the cheapest 3D viewer just shut one eye! With both eyes open and no parallax the brain says "flat" no second thought. With just one eye open the brain amplifies the information and reconstructs the scene in real-time, to a perceived depth of ~65-70% that of a true stereo-pair of images seen through a 3D-viewer.
With sound, maybe it's similar -- sound-field depth can be perceived from a monophonic source, under the right conditions of information content and internal consistency, or "coherence".