Acoustic Horn Design – The Easy Way (Ath4)

So, now what about the original ATH460G2, which I have ready for some time but haven't released it yet? 🙂
But it's a bit different overall.

1727975710993.png
1727975747312.png
 
@heijnsva - if it's of any value, I did 4 layers of wall and 25% infill on the throat and it came out very strong/stiff. Not sure why anyone would need it even more dense than that. On the petals, I did 3 layers for walls and 15% infill, and again, those came out really stiff. I can't remember, but I believe I used either Line or Rectilinear infill pattern.
 
@heijnsva - if it's of any value, I did 4 layers of wall and 25% infill on the throat and it came out very strong/stiff. Not sure why anyone would need it even more dense than that. On the petals, I did 3 layers for walls and 15% infill, and again, those came out really stiff. I can't remember, but I believe I used either Line or Rectilinear infill pattern.
Hmm, that would take forever to print on my printer. 2 walls no infill takes 3 hours for one petal. Maybe I'm impatient, or I need to tweak the printer setting a bit.
 
Yes, and it's in fact very very close to the A460G2 + STD-3 (A520G2 + STD-3 measured here, which is also virtually the same). It was reoptimized for that particular throat diameter and opening angle, releasing all other constraints, but it came out very similar (which only illustrates how good G2 already is 🙂). It's just a tiny bit smoother, perhaps, and a bit less narrowing towards HF. If someone already has the A460G2, there's very little to be gained, but it should be better (for that driver).

BTW, the DI of the A460 is basically at tilted line between 500 Hz and 15 kHz. Hardly this can be any better.
 
Last edited:
I would appreciate some advice on infill and wall thickness especially for the throat part.
I tend to not comment on something I have no strong opinion about, and can imagine various settings leading to perfectly usable results, it can depend on the filament used, but typically one wants to minimize the printing time. So in the end I think it's all about what you find appropriate. Modern slicers allow to reinforce some chosen regions (around bolt holes, flanges, etc.) with parameter modifiers, etc. Do it how you feel it's right 🙂
 
I see it's not always a feasible option, but a lot can be gained using a bigger nozzle. I use 0.6mm and would not go back for the larger pieces. There's just no good reason, it's all better with the larger one (IMO) - you basically get stronger prints and faster.

But I still use an older printer (modified Ender 6), not the most recent hyper-fast ones. Maybe it's different with those, there can be different limits.
 
Last edited:
  • Like
Reactions: heijnsva
Seems like it works 🙂
Now it's just a matter of EQ, that's the simple part. You can do anything, the waveguide will obey, staying as clean as it gets.

- Do you have the means to show it normalized, say @5-10deg?
520g2 with 36EXT1 - LaVoce DF10.171K

normalized to 0db @ 1khz. 5ms gating, 1m mic distance. 0 to 60 degrees in increments of 2 to 3 degrees each. 1/24 smoothing

filter:
200hz LR4 (for protection)
eq:
q:2.6 g:-2.7 @542hz
q:3.1 g:3.2 @734hz
q:3.4 g:-1.5 @1065hz
q:1 g:-3 @3195hz
q:1 g:-2.2 @ 7716hz

ath520normalized.jpg


I somehow misunderstood what you wanted and shared a different plot with you, I have just found out 🙂 I have the 30ppi reticulated foam now and I will be testing with it.

Meanwhile I have also created a file for making this out of wood via cnc. I put it alltogether as one and then cut it into two horizontal slices. The throat will be mjf printed (I want to be able to change the adapters) but the rest will be walnut wood. The cnc can dive 140mm so the wooden horn will be in just two parts. I am getting solid walnut in 50mm and I will laminate them first and then carve them via cnc.
 
Last edited:
It's still not what is typically meant by "normalized polars". 🙂
Here's an example provided by @Dkalsi:

These are the raw (actual) polars, as measured:
1728044928694.png


And this is the same data normalized, i.e. all the curves are shown relative to a chosen one (@10 deg here):
1728045081849.png

(Basically you subtract the reference curve from all the other.)

The normalized polars show what's -theoretically- possible if you EQed the reference curve perfectly flat.
The directivity and diffracions may be easier to see in the second graph sometimes.
 
Last edited:
  • Like
Reactions: kodomo
Well I don't think it's really necesary, it's obvious by now that there's absolutely no problem and it behaves as predicted.

- So, how does it sound? 🙂 It looks terrific.

You may want to check the distortion down low. It will be interesting to see how low you can really push that driver in such device.
 
  • Like
Reactions: kodomo
Here's another idea. Let's make the waveguide itself as light as possible, here it's only 5 mm thin wall, possibly using something like PLA LW (foaming filament). The waveguide is then only pushed against the driver by the threaded rods, attached close to its center of mass. I actually find this pretty elegant.

1728050739058.png
1728050536333.png


Or like this -
1728052138916.png
 
Last edited: