Sorry spamming, but from wikipedia:
Sound localization - Wikipedia
"
For frequencies below 1000 Hz, mainly ITDs are evaluated (phase delays), for frequencies above 1500 Hz mainly IIDs are evaluated. Between 1000 Hz and 1500 Hz there is a transition zone, where both mechanisms play a role.
Localization accuracy is 1 degree for sources in front of the listener and 15 degrees for sources to the sides. Humans can discern interaural time differences of 10 microseconds or less.[9][10]
"
So my previous writings have a wrong frequencies listed.
Sound localization - Wikipedia
"
For frequencies below 1000 Hz, mainly ITDs are evaluated (phase delays), for frequencies above 1500 Hz mainly IIDs are evaluated. Between 1000 Hz and 1500 Hz there is a transition zone, where both mechanisms play a role.
Localization accuracy is 1 degree for sources in front of the listener and 15 degrees for sources to the sides. Humans can discern interaural time differences of 10 microseconds or less.[9][10]
"
So my previous writings have a wrong frequencies listed.
And Klippel says that we already have a 90 degree phase error at 5kHz at 5m distance, if the temperature changes only 2 degress Celsius. At the 21:00 mark:
Klippel's Near-Field Scanner vs Anechoic Chamber: Discussion with Christian Bellmann - YouTube
In testing loudspeaker part 2 by Joseph D'Appollito - he points out the minor difference that these phaseshifts might make - unless if they are so severe, that we reach 5-15ms in the bass and hereby create a group delay so big that we shift the fundamental bass sound "away" from their upper harmonics - which makes the sound image kinda "torn apart" - making the bass in some instances sound "soft". :
Testing Loudspeakers: Which Measurements Matter, Part 2 | audioXpress
I think I've personally heard this in my own system. If I don't get the transition between my multi-subwoofer-setup and mains correct - then the bass-oomph does not fit with the harmonics from my mid-woofers and then a typical "wack" on a big drum or other deep sound - simply sounds "softer" with less "dynamics" if you will, for a lack of better words.
When listening to different speakers.... this phenomena seems to also fit with poorly integrated midranges and tweets, so that the tweeter seems "left on the top"(read separate) of a midrange, rather than being the harmonic continuum for the midrange.
Klippel's Near-Field Scanner vs Anechoic Chamber: Discussion with Christian Bellmann - YouTube
In testing loudspeaker part 2 by Joseph D'Appollito - he points out the minor difference that these phaseshifts might make - unless if they are so severe, that we reach 5-15ms in the bass and hereby create a group delay so big that we shift the fundamental bass sound "away" from their upper harmonics - which makes the sound image kinda "torn apart" - making the bass in some instances sound "soft". :
Testing Loudspeakers: Which Measurements Matter, Part 2 | audioXpress
I think I've personally heard this in my own system. If I don't get the transition between my multi-subwoofer-setup and mains correct - then the bass-oomph does not fit with the harmonics from my mid-woofers and then a typical "wack" on a big drum or other deep sound - simply sounds "softer" with less "dynamics" if you will, for a lack of better words.
When listening to different speakers.... this phenomena seems to also fit with poorly integrated midranges and tweets, so that the tweeter seems "left on the top"(read separate) of a midrange, rather than being the harmonic continuum for the midrange.
Still nothing wrong with this.
Like I wrote before, strong ‘one note’ diffraction leads to irregular off-axis behavior. Since we not only listen to the speaker but also to the room, such behavior and the resulting uneven first lateral reflections could confuse our sense of direction. Personally I think there is more in there than in the baffle itself.
The reason to build a good baffle, is to make the drivers work their best. A lot of problems in speaker drivers can be solved with EQ. Baffle problems are in 3D and cannot be solved with EQ. Any type of issue with edge diffraction on a wrongly designed baffle, result possibly in many smaller sound sources, that are within a short period of time - in relation to the direct sound. This easily "smears" the direct sound, and we have trouble perceptually discerning closely time-related sounds. But if they are delayed more than around 20 ms or so.... we start to perceive them as added spaciousness. This should be the reason why a loudspeaker with a good smooth FR - both on- and off-axis, should sound better in most rooms - when we keep the speaker at a moderate distance to sidewalls or at least dampen or disperse these early reflection in the horizontal plane.
If the speaker in any way has a bad off-axis response. Then you will get all kinds of weird reflections, that in no way sounds like the direct sound. This results in a chaotic FR in the listening position that is absolutely not possible to do anything to, by the power of any DSP or EQ.
Either you dampen your room like crazy to avoid the reflections or you "simply" build/design a better speaker
A good midrange - IMO - Should be very flat and uncolored - and at the same time seamlessly blend with tweeter, so that we get a very clean and smooth on- and off-axis response throughout the most sensitive frequency area of the human hearing.
Mostly I think we run into trouble with building loudspeakers, because of looks, complexity and size... rather than actually solving the sound issue
Please correct me if my conclusions are wrong since I'm literally thinking this through as I write and am eager to learn since it is interesting topic
Before yesterday my concept about speaker diffraction was built on notions like "roundovers reduce diffraction", "roundovers should be big enough to reduce diffraction lower in frequency", "diffraction on low frequencies is not as detrimental than high" and some other similar but all these are were without context. To me it looks logical that hearing system has evolved to take advantage of head geometry. Literally the head provides predictable anomaly for incoming sound and brain has learned how to compute information about it. Head geometry relates to diffraction related delay about at the below 4000Hz so I think this would be the range delay related issues in loudspeakers "distract" hearing system the most.
FLETCHER-MUNSON. The curve has nothing to do with head geometry. It is physiological how we perceive that 4kHz sinewave is much louder than 50Hz sinewave even if the sound pressure is the same.
DUPLEX THEORY. It is how we localize sound based on head geometry (distance between ears, etc.). Frequencies are related to the size of the head.
SPEAKER BAFFLE DIFFRACTION. At LF we have "baffle-step" due to sound propagation to the back of speaker (4-radian) and at HF we have extra ripples due to edge diffraction.
Major effect of this diffraction is to frequency response, so if the frequency response can be made flat (through crossover work) the effect is minimal.
But the effect to on-axis and off-axis is different, especially to LF, that's why I think Linkwitz introduced the dipole woofers (sorry if I'm wrong). Linkwitz also experimented on the effect of baffle to HF and said that the effect is minimal at best.
Like I said before, HF diffraction is audible to me (small ripples in FR due to symmetrical positioning of tweeter on the baffle), but it may only be audible for a few people with sensitive ears.
As the effect of diffraction to off-axis listening of bass, or the accuracy of bass, imho is not really critical to most listeners with far-from-perfect audio system. Try to listen to drums live. It's fine that my speaker doesn't produce drum sound like that.
But I heard/read that Linkwitz dipole woofers (as in the Orion speaker, etc.) can produce drums sound like real.
Last edited:
Linkwitz was skeptical about HF diffraction including the edge rounding. The radius should at least 1/4 of the wavelength to be effective, so the common edge rounding is only cosmetics he said. But i have no experience here.
With music. This is I think just about FR differences, one with smooth FR the other with "rugged" FR. The difference is less than 1dB and this in inaudible for most ears. If the FR is overall not flat of course small difference doesn't matter.
Comparison is the same tweeters, baffle and crossovers (which is developed with diffraction taken into account), only location of the tweeter is different. And tweeter flange was flush-mounted. My conclusion was that for best result it was necessary to vary the distances from tweeter to corners to minimize diffraction effect. I don't like asymmetry aesthetically so varying distance from top and sides (same left/right) is sufficient.
I don't round the corner either as I don't believe it is audible as the effect is minimized by the fact that at higher frequency the tweeter is very directional (and baffle width is usually too wide for the frequency)
But rounded edge and small baffle for tweeter like in the trapezium-style enclosure might have merit I think.
Some says they hear 0,5 db variations. Some others claim, beyond the power response, front-bafles give simply different trade offs. Big à la Sonus Faber Stradivari is not bad for soundstage, more rounded or large flanges between 20° to 45° with sharp edges between the front and its sides, are good enough (Kef 104/2) for the speaker to disappear and good soundstage. Some others claim than a monkey coffins if less good with the imaging instead gives a feeling of more life and dynamic in the frequency range involved, because ears are also sensible to what makes the music : variations.
The Hansen Prince is a good compromise to my ears but I of course draw no conclusion about that, there is also an happy mariage or not between a driver and its enclosure. If you compare Vivid rounded speaker and the ugly What Pappy speakers that are enhanced monkey coffins, both are said to sound good and in between some YG Acoustics and Magicos are said to be as good.
I took the most expensive loudspeakers for a reason, the enclosures are a strong marketing factor for those brands.
I imagine maybe the Cabasse Sphere beat them all for the best said shape, but I'm even not sure it's the best at the end from a blind test.
I have the idea power response still the main factor with filter topology, look at the "petite" D&D... a baby monkey coffin. DSP/active is the soundstaging friend... Can a better shape enhance that ????
The Hansen Prince is a good compromise to my ears but I of course draw no conclusion about that, there is also an happy mariage or not between a driver and its enclosure. If you compare Vivid rounded speaker and the ugly What Pappy speakers that are enhanced monkey coffins, both are said to sound good and in between some YG Acoustics and Magicos are said to be as good.
I took the most expensive loudspeakers for a reason, the enclosures are a strong marketing factor for those brands.
I imagine maybe the Cabasse Sphere beat them all for the best said shape, but I'm even not sure it's the best at the end from a blind test.
I have the idea power response still the main factor with filter topology, look at the "petite" D&D... a baby monkey coffin. DSP/active is the soundstaging friend... Can a better shape enhance that ????
Asymmetric designs can be quite convenient, indeed, but come along with the hefty penalty of an asymmetric radiation behavior into the room, also. I therefore abandonned asymmetric designs altogether in my later designs for non-nearfield systems.
You can very well archieve a decent diffraction behavior along even with a crosswise mirrored symmetrical design/positioning of any driver on a baffle instead. Symmetry is not primarly correlated as such with a problematic diffraction behavior.
You may try to shape your baffle edges not linearly in order to steadily increase the time delay till diffraction along the rotating angle (from center). So you may end up with symmetrical, curved baffle shapes instead of asymmetric straight ones.
Attachments
https://www.audiovintage.fr/leforum/viewtopic.php?t=1379
bizarre ! they at least succeeded to center the woofer !
bizarre ! they at least succeeded to center the woofer !
Last edited:
Some modelling / simulation?
In the example below, a circular baffle with a radius of 11.32cm was modeled on an experimental basis, which would be equipped with a smaller midrange driver.
A radius of 11.32 cm corresponds to a time delay of 0.33 ms and thus half of the maximum possible physiological ITD of 0.66 ms at an azimuth of 90 ° for a natural sound event. An ITD of 0.33ms thus roughly corresponds to a sound event from a horizontal angle of 30° (sin 30° = 0.5 and 0.5 * 0.66ms = 0.33ms). Which corresponds to a lateral installation of a loudspeaker in an equilateral 60 ° stereo triangle.
Graphs:
The red curve models a 4th order LR bandpass with crossover frequencies at 200Hz and 2kHz, and corresponds to the target-on-axis frequency response of the midrange driver (and also of the whole midrange of the whole speaker).
The gray curve roughly models the off-axis drop in the sound pressure of the midrange driver towards high frequencies.
The green curve selectively models the baffle diffraction. The frequency curve of the baffle diffraction corresponds to that of the midrange driver, which was convolved with its off-axis characteristics: Since high frequencies are bundled, they have less of an effect on the edge of the baffle. Since the room geometry changes from half-space to full-space at the edge of the baffle, the diffraction level is reduced by -6dB compared to the useful signal. Furthermore, in the time domain window, the impulse appears inverted, as it is a pressure loss when the wave drops into full space at the baffle edge.
The blue curve is the result of adding the two impulse responses, the ideal midrange (red) and the diffraction (green), into a single impulse.
The black curve corresponds to the difference between the ideal midrange and the diffraction.
Comments:
There is agreement that a well-controlled frequency response is of the utmost importance. This is mainly and traditionally based on tonal considerations. Frequency response and impulse response, however, have a clear, 1: 1 relationship and can be converted into one another. Thus, a pulse curve in the time domain corresponds exactly to a sound pressure curve across the frequency scale and vice versa.
The present simulation, based on a synthesized pulse course and frequency response specifications, assumes a worst-case, circular baffle with a correspondingly problematic, worst-case baffle diffraction. The black curve shows their frequency response anomaly compared to a perfect, diffraction-free system. This lies in the pass range of the bandpass between 200Hz and 2kHz within delta 7dB. That means in the graph in the range of -4dB… + 3dB.
Since this is a worst-case scenario, it can be assumed that even only moderately better baffles are likely to have significantly smaller on-axis level irregularities. This means that even moderate, tonally well-tolerated irregularities in the frequency response also have the potential to impair spatial locatability. For this reason it seems right to aim for “ruler flat” frequency responses in relation to the listening position.
Corresponding meticulous equalization by means of DSP appears to be quite sensible. This ensures that at least one point in the room, the sweet spot for the best possible solo listening pleasure, has optimal conditions for both tonal balance and spatial location. This correction by means of DSP refers exclusively to the direct sound, which in the time domain only applies until the first room interference occurs.
In this very example and in a digital multiway system, you simply could apply perfect DSP by convolving the midrange's driver input signal with an impulse corresponding to the inversed amplitude of the black curve. This would eliminate all (or at least nearly all) of the effects induced by the baffle diffraction. You would end up with a near-perfect impulse corresponding to the initial one from the red data points = the theoretically perfect LR 4 Bandpass from 200Hz to 2kHz.
For non-DSP corrected and therefore less well filtered systems it might prove useful to optimize the baffle diffraction not only within an ITD of 0.66ms (which is a must). But also to take special care of tit’s behavior within 0.33ms. Since this value corresponds to the azimuthal acoustic events within the conventional 60 ° stereo setup.
Whether with or without DSP, the aim for a psychoacoustic diffraction behavior that appears as minimal as possible by means of constructive methods is part of the functional aesthetics of loudspeaker construction. DSP can do practically everything, even if it can be used as a successful but exclusive “brute force” measure to correct problematic diffraction behavior. Enzo Ferrari once remarked the following about the then young aerodynamics: "L’aerodinamica si addice a gente che non sa costruire motori". He was wrong and right in his vanity and his pride. In this sense, the baffle diffraction should be optimized as a primary designer goal. With and in spite of the possibilities of the DSP.
Nota Bene: These statements are based on assumptions. Well-founded contradictions, better values, theories and models are therefore always welcome.
In the example below, a circular baffle with a radius of 11.32cm was modeled on an experimental basis, which would be equipped with a smaller midrange driver.
A radius of 11.32 cm corresponds to a time delay of 0.33 ms and thus half of the maximum possible physiological ITD of 0.66 ms at an azimuth of 90 ° for a natural sound event. An ITD of 0.33ms thus roughly corresponds to a sound event from a horizontal angle of 30° (sin 30° = 0.5 and 0.5 * 0.66ms = 0.33ms). Which corresponds to a lateral installation of a loudspeaker in an equilateral 60 ° stereo triangle.
Graphs:
The red curve models a 4th order LR bandpass with crossover frequencies at 200Hz and 2kHz, and corresponds to the target-on-axis frequency response of the midrange driver (and also of the whole midrange of the whole speaker).
The gray curve roughly models the off-axis drop in the sound pressure of the midrange driver towards high frequencies.
The green curve selectively models the baffle diffraction. The frequency curve of the baffle diffraction corresponds to that of the midrange driver, which was convolved with its off-axis characteristics: Since high frequencies are bundled, they have less of an effect on the edge of the baffle. Since the room geometry changes from half-space to full-space at the edge of the baffle, the diffraction level is reduced by -6dB compared to the useful signal. Furthermore, in the time domain window, the impulse appears inverted, as it is a pressure loss when the wave drops into full space at the baffle edge.
The blue curve is the result of adding the two impulse responses, the ideal midrange (red) and the diffraction (green), into a single impulse.
The black curve corresponds to the difference between the ideal midrange and the diffraction.
Comments:
There is agreement that a well-controlled frequency response is of the utmost importance. This is mainly and traditionally based on tonal considerations. Frequency response and impulse response, however, have a clear, 1: 1 relationship and can be converted into one another. Thus, a pulse curve in the time domain corresponds exactly to a sound pressure curve across the frequency scale and vice versa.
The present simulation, based on a synthesized pulse course and frequency response specifications, assumes a worst-case, circular baffle with a correspondingly problematic, worst-case baffle diffraction. The black curve shows their frequency response anomaly compared to a perfect, diffraction-free system. This lies in the pass range of the bandpass between 200Hz and 2kHz within delta 7dB. That means in the graph in the range of -4dB… + 3dB.
Since this is a worst-case scenario, it can be assumed that even only moderately better baffles are likely to have significantly smaller on-axis level irregularities. This means that even moderate, tonally well-tolerated irregularities in the frequency response also have the potential to impair spatial locatability. For this reason it seems right to aim for “ruler flat” frequency responses in relation to the listening position.
Corresponding meticulous equalization by means of DSP appears to be quite sensible. This ensures that at least one point in the room, the sweet spot for the best possible solo listening pleasure, has optimal conditions for both tonal balance and spatial location. This correction by means of DSP refers exclusively to the direct sound, which in the time domain only applies until the first room interference occurs.
In this very example and in a digital multiway system, you simply could apply perfect DSP by convolving the midrange's driver input signal with an impulse corresponding to the inversed amplitude of the black curve. This would eliminate all (or at least nearly all) of the effects induced by the baffle diffraction. You would end up with a near-perfect impulse corresponding to the initial one from the red data points = the theoretically perfect LR 4 Bandpass from 200Hz to 2kHz.
For non-DSP corrected and therefore less well filtered systems it might prove useful to optimize the baffle diffraction not only within an ITD of 0.66ms (which is a must). But also to take special care of tit’s behavior within 0.33ms. Since this value corresponds to the azimuthal acoustic events within the conventional 60 ° stereo setup.
Whether with or without DSP, the aim for a psychoacoustic diffraction behavior that appears as minimal as possible by means of constructive methods is part of the functional aesthetics of loudspeaker construction. DSP can do practically everything, even if it can be used as a successful but exclusive “brute force” measure to correct problematic diffraction behavior. Enzo Ferrari once remarked the following about the then young aerodynamics: "L’aerodinamica si addice a gente che non sa costruire motori". He was wrong and right in his vanity and his pride. In this sense, the baffle diffraction should be optimized as a primary designer goal. With and in spite of the possibilities of the DSP.
Nota Bene: These statements are based on assumptions. Well-founded contradictions, better values, theories and models are therefore always welcome.
Attachments
Last edited:
Yes, quite weird, indeed. Form Follows Function ... You don't need to go 3d along with these kind of countours. Try it yourself in 2d and you will see. You can add such a shape to your regular rectangular speaker box front for first experiences and measurements. Really nothing 3d. How goes the saying? Flat is beautyful? Or so?
Last edited:
Sorry for coming late, but edit time has expired: This green curve not only has reversed polarity in the time domain, but also it is delayed by 0.33ms, of course. To simulate for the time lag for the 11.32cm baffle radius.
While I follow the rest of your post, I fail on this part. It’s clear by itself, but what assumption lies under it?
Sorry... but a DSP is a line-level tool. It can only change the signal shape, level and delay/phase. But the resulting signal is still sent to the same ONE driver unit. And since you only have one driver unit - you only have one source. Which still leave us with the same problem in the room - no matter what you do with a DSP. Off course, you can make everything seem perfect in a single spot in the room.... but this will all fall apart when moving just a tiny bit.
As soon as the sound leaves the drivers cone/dome.... everything is more complex and converts from a 2D signal to a 3D sound field. Again - the DSP lose all control - except in one single spot in the sound field - exactly where you put the microphone.
You can average an area and get a way less precise result that you can aim for with the DSP - and be happy with the average resulting curve.... but then it's kinda like smoothing a FR to make it look nice - to hide the actual errors.
It all depends on how deep we want to go. We can't have it all... but making a smoothed/averaged image of the basic issue - is not the way to go - then I'd rather have a more smooth directivity.
Geddes also points out that you can't have perfect imaging and perfect spaciousness at the same time. Typically a narrow dispersion loudspeaker like hi's - will give you a more precise imaging in a well dampened room - whereas a more wide dispersion loudspeaker like a Revel type design, will give you more spaciousness in more reverberant room at the expense of som imaging loss. I think - that Geddes goes for imaging, because that is his preference. If try myself to go for the middle ground, by using a smaller waveguide - so that I have some level of imaging, but still retain a typical spaciousness level like in most typical speakers. It's all a compromise.
But still.... a 3D problem - which a room is - can only be properly solved with 3D solutions. A DSP is still a 2D tool - unless you have extra sources like in multi-sub and like the trickery that Beolab 90 does
The (only) assumption is related to the psychoacoustic decoding of a (any!) two-peaked impulse into a (virtual and ambiguous) perception of azimutal informations. I have searched for studies about this, but didn't find anything in the literature so far in order to verify or falsify this assumption. I would very much like to know wheter this is plain BS or not.
If this assumption would be true, then this would establish a relation between the direct signal from the center of the driver and the delayed signal from the baffle diffraction on one side, and the potential blurring of azimutal perception on the other side: T_center_driver_to_baffle_edge would get a similar meaning as the TID by the specific psychoacoustic processing.
Yes. And I don’t see much contradiction in what you write later on. I agree on most of it.digitalthor said:
But I do not quite understand this story about Geddes and “spaciousness” and “imaging”. Geometrically, I understand an “image” as one specific 2D representation of “space” (3D in these terms)? So “space” would be “image” plus the dimension of “depth” along with the psychoacoustic interpretation of ths perceived "soundfield"? Oh my, these terms ... isn't "field" 2D again ... And again much psychoacoustic processing we still do not unterstand well. Diving into the scull, we unfortunately do not have any longer these precise models as we have for the acoustic world outside the scull.
Anyway, maybe I should have better written “ … DSP can do practically everything, even if it can be used as a successful but exclusive “brute force” measure to compensate or minimize problematic diffraction behavior, at least for one point within the auditorium.” And I think it is quite evident from my writing that I strongly advocate the correct physical design of baffles in order to smooth the diffraction effects. Before applying any DSP in order to correct for an eventually misbehaving diffraction. By the way, I also opt for aerodynamics before motor power, but certainly do appreciate and welcome both options.
Last edited:
I think we must not underestimate the power of masking. Recognizing a second, almost identical signal, weaker than the first and several hundreds of microseconds later is highly problematic for us.
Also, the diffraction signal reaches both our ears too. The ‘comparator’ in our brains that receives the neural pulses from our ears might be fooled, we can’t rule that out completely, but it is not very likely.
Also, the diffraction signal reaches both our ears too. The ‘comparator’ in our brains that receives the neural pulses from our ears might be fooled, we can’t rule that out completely, but it is not very likely.