Acoustic Horn Design – The Easy Way (Ath4)

[sheeple]

I don't know what you man by a "shorter termination" but it's no problem -

A lower factor q, making the transition between the curved waveguide and the flat baffle more abrupt.

 

[mabat]

That doesn't play any role. The front baffle simply continues from the last outward curve (i.e. "slice") of the waveguide - it can be anything and the algorithm doesn't care or even know what's "inside" that curve.

 

[sheeple]

Good to know. So the meshing algorithm would probably also integrate a waveguide which is slightly wider than the flat area of the baffle, and connect it to the curvature.

 

[mabat]

No, the waveguide must fit inside the flat area of the front baffle. That's a limitation.

 

[mabat]

BTW, this looks surprisingly good (just a quick and crude try / solving time 6 minutes) -

 

[sheeple]

No, the waveguide must fit inside the flat area of the front baffle. That's a limitation.

I understand, the curvature is attached to a box definition, not subtracted. So it needs to be downsized a bit for simulation, by about 2.5 cm. This device was 26.5 cm total, intended to fit a baffle width of 32 cm, with 4 cm radius on each side (8 cm total). A downsized version will probably still give a good idea of its performance.

 

[limacon]

BTW, this looks surprisingly good (just a quick and crude try / solving time 6 minutes) -

View attachment 1038219

Woah!

 

[sheeple]

I think the resolution is a bit charming, as the diffraction effect between 2k and 3k are smoothed. Nice to see though.

 

[mabat]

Higher resolution far-field data, horizontally, vertically, diagonally -

There's obviously a lot of diffraction, yet it's still pretty balanced - I guess this is due to the overall dimension ratios (307 x 236 x 179 mm).
It will be interesting to see this with all the edges rounded.

 

[sheeple]

Yes, squishing the vertical introduced some ripples, but it seemed to me the more significant bumps are located opposite to where there is a ditch on the other axis. This is also apparent from the earlier infinite baffle simulation. If diagonal is a mean (it isn’t strictly) of both axes, this explains why this is the smoothest response. Should sum very good. Interested to see results for rounded edges.

 

[mabat]

Diagonals are often the smoothest. Typically those are the least symmetric situations.

 

[mabat]

I understand, the curvature is attached to a box definition, not subtracted. So it needs to be downsized a bit for simulation, by about 2.5 cm. This device was 26.5 cm total, intended to fit a baffle width of 32 cm, with 4 cm radius on each side (8 cm total). A downsized version will probably still give a good idea of its performance.

Thinking about it, it should not be that difficult to implement a bending of the sides of a waveguide so that it became a part of a rounded edge.

 

[sheeple]

For ath modeling or printed?

For print, I thought this had probably more influence on termination than diffraction of a 1-2 mm protrusion of the outermost horizontal areas, which happens when the rounding is taking away some of the flat baffle.

 

[mabat]

For the modeling (and then making) but since it would be probably difficult to build into a cabinet there's really no point I guess.

 

[mabat]

OK, I've got the rounded box modeling ready, more later...

 

[mabat]

Radius 15 mm (as on the picture above) -

That's not bad.

 

[sheeple]

Wow, this is nice!

I had already seen with fluids sim that for horizontal, maybe parameter s, which can create a narrow banded waist-banding, could be improved. Will try this out when it’s our turn.

 

[fluid]

I don't know what you mean by a "shorter termination" but it's no problem -

He means the final flare is over a shorter length to make the overall waveguide smaller. augerpro's waveguides were similar and it made getting the Ath mesh to CAD mesh tricky. Doing it all in one in a gmsh script is much better and will give a better and easier result than the cut and fit that was necessary before to see the effect of a finite baffle.

One has always to be careful about mesh quality but that's not difficult.

When it had to be done in two stages avoiding these long thin triangles was quite difficult and they caused problems. Interestingly this solved quite nicely up to 10K

but it needed to be this to solve without any real problem the Ath generated version above would solve better still.

augerpro's waveguide solved nicely like this, but there it's too many elements to expand it out to a whole baffle in half symmetry.

 

[fluid]

Somehow I missed a lot lot of posts above where some of it had already been covered.

This was the horizontal from the good model, the sharp edges on the cabinet have more impact on a smaller waveguide with a more abrupt termination.

With non axisymmetric waveguides having more vertical baffle dimension and more gradual rounding can help to somewhat offset the mouth being smaller in that dimension. Something that will be much easier to experiment with straight from Ath.

 

[mabat]

Important:

The long thin triangles are a consequence of a sub-optimal setting of Mesh.ZMapPoints and it's easily correctable. The default setting works fine in most cases but sometimes it needs to be tweaked a bit, just as in the cases above. This holds not only for the mouth but also at the throat.

It's still not properly documented in the User Guide but see this post:
https://www.diyaudio.com/community/...-design-the-easy-way-ath4.338806/post-6102810

Updated Desmos tool for easier setting of the four Mesh.ZMapPoints numbers - it should be intuitive:
https://www.desmos.com/calculator/grihwchoyw

What is a "slice" in Ath is described here (page 15 of the UG):
http://www.at-horns.eu/release/ath-4.8.0.pdf#page=15