How to build a spherical speaker?

How diffraction acts has everything to do with the frequency and sizes we are taking about
Those distinctions are lost by calling everything diffraction. Baffle step is the gain increase when wavelengths are shorter than the baffle dimension. Radiation shifting from 4pi to 2pi. With infinite baffle the entire spectrum operates in 2pi. Does that mean the entire spectrum is diffracted with edgeless infinite baffles?
 
  • Like
Reactions: TNT
Semantics. It my be convienient for you to pigeon hole the ends nto 2 different bins, you do have differant kinds of behaviour if you ignore the size/frequency relationship.

What do you do with the stuff in the middle that has a transitional response. Ignore it because it doesn’t fit into one of your bins?

It is diffraction, but what a diffraction looks like is very size/frequncy dependent.

dave
 
BTW, you shouldn't necessarily judge diffraction based on response ripple which you might see on a simulation like this. Diffraction can be an audible problem and also be difficult to identify by looking at response.)
I agree Allen. To my ears, the primary audible effect of diffraction ripple is frequency response anomalies. However, I believe there are audible aspects that go beyond frequency response. With program material that has naturally recorded spatial cues, diffraction off of a sharp cabinet edge or unsmooth baffle surface can damage the three dimensional presentation. I have always assumed that frequency response ripple is the most obvious way of identifying it.

To me the second and third graphs show baffle step with and without diffraction. Baffle step is a 2pi - 4pi transition and involves the entire baffle. Diffraction is edge scattering when the baffle is significantly larger than the wavelength and only occurs at the baffle boundaries. I don't the benefit of conflating the two.

The baffle step is caused by diffraction. When the wavelength is significantly larger than the baffle, the sound diffracts completely around the cabinet. This makes the speaker behave omnidirectionally. Without diffraction, the speaker would never transition to 4-pi radiation.

But your point is relevant. Dealing with baffle step (which is caused by diffraction) is a very different process than dealing with high frequency baffle edge diffraction.
 
I have always assumed that frequency response ripple is the most obvious way of identifying it.
Diffraction is not primarily a frequency domain issue even though it can have an effect, there's no clear connection. Not only don't the geometric properties behind the creation of response variations through diffraction necessarily line up with the amount of diffraction, but also not with the audibility regardless.
 
Diffraction is not primarily a frequency domain issue even though it can have an effect, there's no clear connection. Not only don't the geometric properties behind the creation of response variations through diffraction necessarily line up with the amount of diffraction, but also not with the audibility regardless.
So how would one know it really is diffraction creating this problem? We'd like to take the voodoo layer out of it, you know 😉.
 
  • Like
Reactions: TNT and tmuikku
Yeah, there is no voodoo although AllenB has suggested so for many times without giving any info about it. Post #267 has more content as there has ever been for the argument he has been having.

If there is no diffraction created secondary sound source at the edge then there is no interference with direct sound, wiggle in frequency response. When there is secondary sound source, then there is wiggle in the frequency response simple as that. Sounds superimpose, interfere, there is no way around, and is perfectly valid way to evaluate if there is diffraction happening in a way that the secondary sound source forms. Diffraction always happens when object and wavelength are similar size, its only the secondary sound source forming at sharp edge which might cause audible issues. If there is no secondary sound source we don't hear the diffraction, I mean diffraction is then sound going away from us, around the box, and thats it. It goes around the box anyway, the only difference is the secondary sound source at the edge, be it rounded or not.

Hes right though, that the frequency response doesn't necessarily tell about audibility, only about existence of the secondary sound source, the edge. Diffraction usually happens very soon after direct sound, I don't know how brain processes it exactly. It might be just the change in timbre, or something else, or both. For example think a small two way bookshelf speaker, measure edge distance from tweeter and then compare that to distance of the woofer, they are about similar distance and would be similarly audible in this regard, for tweeters response. I suspect the frequency response is audible directly and then there might be some audible effects by the temporal aspect of it. We gotta remember the interference is not just for on-axis, or listening axis but all directions like towards early reflections, which might now be much more different than direct sound, diffraction related interference made it worse.

AllenB, if one looks at single frequency response graph like on-axis it might hide the diffraction related interference under other issues the speaker is having, like cone issues, but if you have normalized response 0-180 degrees existence of edge diffraction, interference due to secondary sound source, is quite easy to see as narrowing at the main hump and narrowing after that, alternating up in frequency the bigger the baffle to the transducer.
diffraction-interference.pngno-diffraction-interference.png
Here is real speaker measurements, some interference happening due to secondary sound source, easy to identify.
diffraction.png
 
Last edited:
In the end all this is just to try make better sounding speakers, at least avoid making poor sounding ones. It doesn't matter who is wrong or right but it would be very helpful to all if we could discuss and this stuff with out bad feelings and unnecessary debate and to get better understanding on it. I'm willing to change my views on all this if AllenB or someone else gives good explanation for the stuff, all I want to do is find out how to detect diffraction on the measurements as well as by ear to determine what is bad and then avoid it before building and listening, from simulations. Its easy to spot it from frequency response, only audibility is a question. Its said to be audible, but its hard to hear, without knowing specifically what to listen to.

While diffraction is one main aspects of a spherical speaker compared to a box its way of topic, how to build one 🙂
 
Last edited:
....

Hes right though, that the frequency response doesn't necessarily tell about audibility, only about existence of the secondary sound source, the edge. Diffraction usually happens very soon after direct sound, I don't know how brain processes it exactly ...
Maybe its here where the waterfall should be useful. The WF is hard to analyse - a lot of forest in from of the trees... I think it is in there but current presentation and interpretation is not on par perhaps...

//
 
Here is how you can think about it, use ripple tank to visualize and get insight how it plays out.
http://www.falstad.com/ripple/Ripple.html?rol=$+3+512+64+0+0+936+0.048828125 s+2+197+264+2+0.1707277097902098+0+44.16285076324422+971.582716791373+1+0 w+0+0+265+278+265 w+0+275+265+276+473

ripple-tank.png

Think like this: if you have perfect sound source, ideal piston driver it emits flat frequency response, ok on-axis but lets just take one wavelength as example, such that emits towards the edge as well. Image above approximates such situation except used point source as was lazy. The secondary sound source happens at the edge, subtracts from the sound that goes around the edge and is delayed due to geometry from the frontal and opposite side directions so arrives after the direct sound. One can think it just attenuated opposite phase source at the edge. In the example its approximation of loudspeaker looked from top down, listener would be up on the image, backside or the front wall would be at bottom of the image somewhere.

Alright, the secondary source propagates with the direct sound for the part that diffracted around the corner towards bottom and right on this image. But the part that propagates upwards (on-axis) and to left in this image is delayed from the direct sound. The direct sound is still pristine but if one were to take impulse response and then window it including the edge diffraction interference appears in frequency response, depending on the wavelength and delay to the edge. When delay is shorter than about 1/4wl there is no difference as they are in phase, when delay is 1/2wl there is dip and so on. As one can see from the image, the delay is more to the left between direct sound and the back wave, and less as we go more on-axis (upward in the image). Delay depends on distance from source to edge and at which angle one observes at. Most delay on opposite side (round trip from the edge), least on same side (zero). On-axis has roughly delay from center to edge, because we listen a lot farther than the speaker size so there is not mucho added path length other than what it is from transducer to edge. Hense, we get strong interference on both sides of the speaker due to the opposite side edge back wave coming late and making interference, also very strong interference on axis because both (or all) edges might have similar distance from transducer and arrive exactly same time and make stronger interference.

When transducer is off-center on the baffle, or edge geometry is so that length varies from source to any point at the edge, the delays are spread out inspecting from on-axis and interference spreads out in frequency. If baffle was circular, constant length from edge to transducer center, then we would see worst effect, on-axis.

From all this, we see wiggle in frequency response because every frequency has different wavelength and its relationship to the edge changes, where other frequency has 1/2wl difference and destructive interference happens while some other frequency has multiple of wavelength and constructive interference happens, wavy shape on frequency response plot which varies by direction, on-axis might have constructive interference and off-axis destructive, path length difference is about two times for the off-axis position as towards on-axis.
 
Last edited:
The baffle step is caused by diffraction.
A source emitting wavelengths significantly smaller than an adjacent boundary is diffraction: so that should also cover room boundaries? The definition is expanding beyond utiility with little demonstration beyond repetition. If you want a slam dunk, demonstrate how a linear doubling or tripling in baffle size tames what are normally called diffaction ripples or how rounding corners significantly shifts the baffle step transition frequency.
BTW, your first graph has a minimum approximate transition frequency of 17 Hz but since you chose a square baffle instead of round that will shift lower.
 
If you want a slam dunk, demonstrate how a linear doubling or tripling in baffle size tames what are normally called diffaction ripples or how rounding corners significantly shifts the baffle step transition frequency.
These don't happen, opposite happens, more ripple with wider baffle until its part of the room unless bigger and bigger roundovers. Bafflestep just turns lower in frequency with increasing baffle size. Rounding the edge removes the secondary sound source and resulting ripple in frequency response but does not significantly affect the baffle step other than the part that was caused by diffraction created secondary sound source, main diffraction hump gets lowered which is right at up on the baffle step. Baffle steps stays about the same but secondary sound source was removed that makes extra hump for the baffle step (top) if inspected on-axis. Assuming edgeles enclosure, a sphere, diffraction would still happen where the bafflestep slope is there just isn't secondary sound source. Frequencies above baffle step only reflect forward hence they appear elevated in frequency response, and below baffle step the wavelength is so long the box is invisible to it, not much interaction with the enclosure and sound doesn't reflect or diffract or anything, propagates to all directions.

I suspect Hifijim knows this stuff though and its just semantics, words, easily contributing to misunderstanding. I'm hoping my writings would increase understanding, not create any more confusion. Its feels complicated subject but reallys is not complicated as phenomenon if thought through simplified concepts like ripple tank, very simple phenomenon in the end, just wave propagating and interacting but it just looks like complicated because we cannot see sound and its 3D and affected by the 3D world we see. Imagination to rescue.

edit.
Everyone trying to get hold on all this please remember the frequency response graph is not a snapshot at some particular time but a short film averaged out over time and made into a snapshot photo. Unless you are looking at real time analyzer. This means that every sound that arrives at microphone within certain time window contributes to the frequency response graph magnitude. If windowing is too short we didn't get all the low frequencies which take longer time to fully come through. If its long enough to include low frequencies we include all kinds of secondary sound sources like diffraction as well arriving to a mic within the time window. (if edge is crazy 34cm away from transducer its delay is one wavelength of 1kHz to give some perspective). Sound arriving later might be at different phase than the direct sound that came through first, out of phase is reducing SPL and in phase is increasing SPL at any given frequency one looks at. All depends on distance from transducer to edge in case of edge diffraction.
 
Last edited:
These don't happen
That's my core point which your clear explantion expands. The baffle width defines a transition point of acoustical behaviours, like nominal 0 and 100 degrees celsius define behavioural transition points for water. Sure glaciers, oceans and clouds are essential all 'really' water but no utility follows from insisting this fact 'really' determines how we approach the different manifestations. It's a pointless semantic simplification.
 
Yes, when there is understanding of the phenomenon (any) then it can be utilized in a (loudspeaker) design. If there is not then the outcome is random in a way, not necessarily worse or better, just something perhaps unexpected or something that was ignored and is what it happens to be. Understand diffraction enough to make it cause no problems, at least addressed some if not completely, and you (any of you) are golden.

Everyone trying to make their minds about diffraction and figuring out all the complicated detail please consider this, everything can be thought to be relative and approachable from philosophical point of view. If one has mental set that diffraction doesn't matter much then there is no point going to deep end just enjoy what ever it is that feels important. But if one wants to figure out how to make better and better sounding system then one cannot ignore things like diffraction. Still, its important to try and first figure out if it is (as phenomenon) something that can be benefitical, or is it always detrimental for sound, or neutral? If its always detrimental its enough to try and minimize it as well as possible keeping all the other perhaps more important design aspects of loudspeaker, and no need to go any further on it. If its something neutral, perhaps there is no need to address it anyway. Perhaps just do some roundovers to address it for the part that is easy to do. Personally I wouldn't do spherical speaker because it doesn't feel practical enough for my manufacturing capabilities. You get the twist. Hope it helps.
 
Last edited:
  • Like
Reactions: TNT and hifijim