Acoustic Horn Design – The Easy Way (Ath4)

"Protruding" the walls inward really seems to help. Not quite there yet - some form of knuckles will be needed I guess :) The diagonals start to suffer however.
(Still the same 33 mm throat.)

- This horn I would be satisfied with.


  • pro.PNG
    228.6 KB · Views: 412
  • pro-h.png
    18 KB · Views: 415
  • pro-v.png
    17.9 KB · Views: 396
  • pro-d.png
    17.4 KB · Views: 402
  • pro-pmap.png
    29.5 KB · Views: 401
Last edited:
Now let me show you something:

- the plots below are for horizontal, vertical and diagonal polars (a 7.5 [email protected])
- on the left is the M2 clone (shown before) labeled "X", on the right an OS waveguide made squarish, aka "ATH"
- these two have the same throats (33 mm) and are about the same size

ATH has nothing you would call a diffraction slot, I assume. I just can't see where is this M2 clone actually better, having one. I see a tighter pattern control with the ATH overall, it is virtually as good diagonally as horizontally or vertically. What I am missing? I'm sorry but this "optimized diffraction slot" thing seems like a fairy tale to me. At least for a pattern control. I can't comment on HOMs.






Fundamentally, the way that horns and waveguides work is that they take the same amount of energy and they focus that energy into a narrower beam.

A round oblate spheroidal waveguide will 'illuminate' the room consistently. IE, the sound on the vertical axis is identical to the horizontal axis and is identical to the diagonal axis.

The JBL "high definition imaging" waveguides have a diffraction slot that's shaped like an "X."

The diffraction slot does three things:

1) it narrows the directivity on the diagonal axis

2) it widens the directivity of the horizontal and the vertical axis. I like to describe this as "robbing Peter to pay Paul." Basically the HDI waveguides improve the response on one axis by degrading the response on another.

3) When sound radiates into a narrower beam, the output goes UP. Because the volume of an HDI waveguide is smaller than an oblate spheroidal waveguide, the HDI waveguides are more efficient.

These three things are illustrated in your simulations. Note that the HDI waveguide has similar beamwidth to the oblate spheroidal waveguide, but the HDI waveguide is 3dB more efficient at 2khz.
It doesn't, just look at the data. The same beamwidths can be achieved without it so how can you say that it widens the directivity?

That's a good point:

A big part of the reason that I'm fascinated by the HDI waveguides is because I've found that wide beamwidth waveguides don't behave as simulated in the real world.

For instance, I've 3D printed waveguides that were 33cm wide, which performed no better than a 16.5cm waveguide.

From simulations that I've done in ABEC, I believe that what's happening is that the wavefront radiated by the compression driver can't expand fast enough. Due to this, the waveguide behaves as if the walls of the waveguide aren't there.

My 'hunch' is that this is happening because some waveguides produce a converging wavefront, some produce a diverging wavefront, and some produce a flat wavefront.

The JBL HDI waveguides sidestep a lot of this, because the walls are 'pinched.'

In addition to this, there's a couple other reasons to use the HDI waveguides:

1) they look cool

2) they produce higher SPL than an O.S. waveguide

I'm not saying they're the greatest thing on earth, just saying that I'm a fan. If you prefer oblate spheroidal, that's perfectly reasonable, OS performs really well.
Hey, this looks fine (374 x 336 mm) -

ThroatDiameter = 33 ; [mm]
ThroatAngle = 0 ; [deg]
Coverage_Horizontal = 110; [deg]
Depth = 120 ; [mm]
SE_s = 0.7
SE_n = 4.0
SE_q = 0.995
Depth.ConicSectionPart = 1.0
Shape = raw2rect
Shape.FixedPart = 0.01
Shape.CornerRadius = 40 ; [mm]
Shape.StretchExp = 3
ShapeCurve = SF
SF = 1.65,2.3,4,2,5.1,2.9
SF.Rot = 0

; Mesh Setting
Mesh.AngularSegments = 100
Mesh.DepthSegments = 28
Mesh.CornerSegments = 10


  • spl_d.png
    17.1 KB · Views: 85
  • spl_v.png
    17.7 KB · Views: 100
  • spl_h.png
    17.8 KB · Views: 148
  • stl.PNG
    382.9 KB · Views: 126
Last edited:
So this is the updated comparison - the second take on the "ATH" (on the right). This one looks like a "nude" version of the M2 clone.

OK, the knuckles add this bump in efficiency (?) around 1,7 kHz. Is that all?


  • bck3.PNG
    295 KB · Views: 126
  • ath_3.PNG
    304.1 KB · Views: 350
  • cmp-d.png
    42.9 KB · Views: 388
  • cmp-v.png
    42.8 KB · Views: 386
  • cmp-h.png
    43.9 KB · Views: 395
Last edited:


2009-01-24 2:20 pm
OK, the knuckles add this bump in efficiency (?) around 1,7 kHz. Is that all?
Seems the over 10K response is smoother without the big dip in the M2 clone, could you extend the sim to go to 20K to see it better? This might well be one of those things that turns out to be different in reality unless you model the exact exit of the driver to horn interface, but interesting as augerpro was putting a lot of effort into minimizing that issue with his shallow waveguides and sim to reality was quite close.
With a 33mm throat the compression driver determines the polar pattern above 10KHz, not the waveguide. I would argue the plots are going too high already.


Ideally we should be doing the ABEC sims all the way down to the diaphragm. (including the phase plug in the sim.)


In the JBL 2409H the phase plug is barely there, just a few centimeters deep.

And in the 2410H, it appears to be sticking out of the waveguide itself! Seems to imply that the ring radiator diaphragm may be less than a centimeter from the waveguide.

Basically the phase plug is disappearing.


  • s-l1600 (1).jpg
    s-l1600 (1).jpg
    156.8 KB · Views: 57
With a 33mm throat the compression driver determines the polar pattern above 10KHz, not the waveguide.

This is not correct. A waveguide done properly can go well above what is considered the region where it is assumed that the driver dominates.

So why does the simulation change between waveguides when it's being driven by an ideal device?

Because of what I said above, it's incorrect. Some drivers with poorly coherent exit wavefronts may have more trouble than others, but there is no theoretical reason why the above should be true, nor was it true in system that I made.