Acoustic Horn Design – The Easy Way (Ath4)

[maiky76]

Just for the record, 2" throat, ⌀400 x 250 mm:

View attachment 1048279

H, V (0 - 90/5)
View attachment 1048280 View attachment 1048281 View attachment 1048282

R-OSSE = {
  R = 200
  r0 = 25
  a0 = 0
  a = 33
  k = 3 + 5*sin(p)^2
  r = 0.35
  b = 0.12
  m = 0.83
  q = 4.4
}

I gave that one a try too.

R-OSSE = {
  R  = 204.4463 ; [mm]
  r0 = 25 ; [mm]
  a0 = 0 ; [deg]
  a  = 33.3064 ; [deg]
  k  = 2.5606 + 3.2955*sin(1*p)^5.0417; []
  r  = 0.28838 ; []
  b  = 0.11572 ; []
  m  = 0.86488 ; []
  q  = 5.8055 ; []
         }

Note: the dynamic range of the normalized polars is 30dB vs 45dB
Again, trying to find solutions in the vicinity (size and coverage) of the starting point.

I am curious, what the expected advantages of the asymmetrical profile?

-C-C for X0ver in not one of them (still circular mouth)
-Smoothness does not seem to be as good both from simulation POW and from the direct comparison showed by @Dkalsi
-On-axis smoothness by breaking the symmetry? But smoothness ON can be made rather aceptable through optimization.
-Different H and V coverage angles but at the expense of the “constant” characteristic

 

[Antoine8082]

I thought they made interface to simplify ADAU+Sigma Studio.

Hello,

I can share or made to/for you some Sigma config for ADAU1707 if you want.

 

[mabat]

Here are horizontal polar measurements of the ST260 vs ST260-KVAR

Do you still have the raw non-normalized data?

 

[mabat]

Fusion is very picky about rails they must intersect each profile/slice precisely.

That's guaranteed with the curves exported from Ath. The last time I did it I was surprised how smoothly it went.

 

[fluid]

That's guaranteed with the curves exported from Ath. The last time I did it I was surprised how smoothly it went.

As you said before sometimes it works when you loft a section at a time but not all in one go. I've had the same trouble with curves I've made myself where they do intersect but the order in which they are selected seems important it getting the loft to function. Then other times there is no problem at all.

 

[Olombo]

@Dkalsi: would you mind to make a waterfall plot from your st260 or st260-kvar? I would be interested if there the correlation between simulated acoustic impedance and resonances can be drawn or not. Regarding the discussion 2-3pages before.

 

[mabat]

-On-axis smoothness by breaking the symmetry?

Actually, the notion that an asymmetry is needed to improve on-axis smoothness has been proved false a long time ago. There's no problem getting an exemplary smooth response even with an axisymmetric device. That has been settled I think.

 

[mabat]

I would be interested if there the correlation between simulated acoustic impedance and resonances can be drawn or not.

If there was a sharp peak in the impedance (or any abrupt change), the frequency response would be affected by that - there would be a sharp peak as well. And a sharp peak in FR has a longer decay in time response/CSD. The same holds for the response of the driver (and the response combined). There's really no resonance induced by the horn acoustic impedance that you wouldn't see in the amplitude frequency response (if measured properly).

 

[mabat]

Just a remark -

sin(p)^5.0417 (in maiky's code) is not a legitimate mathematical operation - it would need to be fabs(sin(p))^5.0417.

 

[maiky76]

Just a remark -

sin(p)^5.0417 (in maiky's code) is not a legitimate mathematical operation - it would need to be fabs(sin(p))^5.0417.

Thanks for pointing that out although I am quite sure what you mean.
What issue is it causing? sin and x^y are both defined, aren't they?
From the numerical optimization POW the only important point is that one input always yields the same output.

I am not familiar with the LUA language you are using, I checked the doc before making the optimizer but I could not find any details on the way the functions are behaving and C99 should work or throw an error.
I would assume that in Ath either NaN or +/-∞ would throw an error or cause the design to be nonsensical, I might be wrong.
When I tried I did not get any error and the design looks "normal" so I assumed that the function would drop the imaginary part or ignore what it did not like. Is that the case?

 

[mabat]

As the k is supposed to be real, you just can't raise a negative number to a general fractional power. For 5.0417 it just won't work for p between π and 2π.
Dropping the imaginary part is not an answer

 

[sheeple]

I gave that one a try too.

Maiky, could you try an improvement with your optimization script on this baffled waveguide in quarter symmetry? My goal was a 'wide' pattern of about 50 degrees at HF, beginning even a little wider in the lower frequencies, and a slightly sloping DI. The latter I realized through the vertical axis.

Results for quarter symmetry / half symmetry:



The current dimension is 265 x 200 mm, where the width is defined by the intended baffle. The vertical dimension I had found to be a good trade off for the intended goal, tho reduce vertical dimension, yet retain vertical pattern control. The term q was used to set the vertical dimension, this is why it has so many digets. I wonder if the dip between 1.000 and 2.000 Hz could be solved better, that shows up in the horizontal plane, and if the ripple could be further smoothened.

Quarter symmetry results translate quite well into hald symmetry, final box, so quarter sym can be used for optimization (quarter symmetry <-> hald symmetry):

hor


ver

; ____ Parameters ____

Throat.Angle = 10.08
Throat.Diameter = 25.4
Throat.Profile = 1
Coverage.Angle = 50             -9*sin(p)^2

Length = 66.5
Term.s = 2.3                         -0.05*sin(p)^2
Term.n = 3.4                         +0.4*sin(p)^2
OS.k = 0.45                         +0.3*sin(p)^2 
Term.q = 0.87674                -0.00425*sin(p)^2

; ____ Source Mode ____

Source.Shape = 1
Source.Curv = 0
Source.Radius = -1
Source.Velocity = 1

; ____ Enclosure ____

Mesh.Enclosure = {
Spacing = 27,30,27,338 ; edge distances (left,top,right,bottom) [mm]
Depth = 270 ; enclosure depth [mm]
EdgeRadius = 25 ; radius of the edge treatment [mm]
EdgeType = 1 ; 1=rounded, 2=chamfered
FrontResolution = 8,8,16,16 ; front side mesh element size (q1,q2,q3,q4) [mm]
BackResolution = 20,20,20,20 ; back side mesh element size (q1,q2,q3,q4) [mm]
}

; ____ 3D Mesh Settings ____

Mesh.InterfaceResolution = 8.0
Mesh.LengthSegments = 30
Mesh.AngularSegments = 48
Mesh.ThroatResolution = 3
Mesh.SubdomainSlices =
Mesh.ZMapPoints = 0.5,0.1,0.76,0.733

 

[Dkalsi]

Do you still have the raw non-normalized data?

Sorry, I missed this. Yes - please find attached non-normalized data (as well as other random measurements: (i) CSD of ST-260KVAR (ii) CSD of ST-260, (iii) H-Polars of B&C DE360 on ST-260-KVAR, and (iv) H-Polars of the Peerless on Joseph Crowe ES-800 Radial Horn).

 

[TNT]

Has anyone compared CSD vs. polar? I never seen that... does the CSD look different between 0 and say 30 deg?

//

 

[mabat]

Yes - please find attached non-normalized data

Thank you, that's a nice comparison. I quite like the KVAR mutation.
What is a bit surprising to me is that at HF both the non-KVAR and the KVAR ST260 produce virtually the same SPL on-axis, including the slope of the response, although the KVAR has (should have) more beaming pattern vertically. So I would expect the KVAR to be a little bit more flat but that's not what's happening - curious. Maybe the vertical beamimg is still too mild to have a substantial effect.

 

[mabat]

Has anyone compared CSD vs. polar? I never seen that... does the CSD look different between 0 and say 30 deg?

As the CSD is a different look at a (single) impulse respone, it will be different in different directions, correspondingly to how different are the IRs. Basically, all resonances should be present at the same frequencies in all directions - that's the typical behaviour and also how we recognize resonances from a set of polars.

 

[MadEconomist]

Apologies if there's a dedicated bug reporting space where this belongs, but I think I've found a bug with 4.8.2. when modeling complex sources, ATH throws the error "the 1d mesh seems not to be forming a closed loop", and the middle of my dome doesn't get meshed - this applies even to the example dome file from page 40 of the manual, and to custom dome configurations that work fine in 4.8.0.

 

[mabat]

I think this has been fixed already: https://www.diyaudio.com/community/...-design-the-easy-way-ath4.338806/post-6996527

 

[MadEconomist]

I think this has been fixed already: https://www.diyaudio.com/community/...-design-the-easy-way-ath4.338806/post-6996527

Ahh, so it has! Sorry, I was just using the latest version from https://at-horns.eu/, so I didn't see the beta version.

 

[Ro808]

2 weeks ago I told another member that it won't be long before we see ATH4(inspired) loudspeaker systems and,
lo and behold...