I don't think so. Two things: the wiggles can be reduced with more extended termination (parameter s). Two graphs attached where I did just that. I tried to avoid it because s also waistbands the low frequencies and sets the maximum width of the pattern, if it ought to be linear between 1-10k at -3 dB. With more termination, 'wiggles' get less. That is why I stand with my interpretation that what the graphs show are actual resonances from design issues. Perhaps it can be optimized, maybe there are limitations if the size is set (~ ⌀315mm). The graphs mabat showed where not looking high resolution, the lines looked fairly straight. I could be wrong. The recent graphs I showed had ABEC.NumFrequencies = 100.
edit: Just saw the resolution of the graphs seem not to match, maybe I would have had to increase another variable like angular segments. Maybe, when mabat tomorrow discloses which frequency number he used, we can take a better judgement of the issue.
edit: Just saw the resolution of the graphs seem not to match, maybe I would have had to increase another variable like angular segments. Maybe, when mabat tomorrow discloses which frequency number he used, we can take a better judgement of the issue.
Attachments
Last edited:
When you have made thousands of these simulations and created your own meshes from scratch you start to see things which may not have been apparent at the start or you understand better what causes issues and how to recognise them. Increasing the number of simulation frequencies does not automatically give you higher resolution data. Where there are numerical instabilities all it does is confuse the data by generating more random junk. In a full 3D simulation it takes effort, trial and error to get meaningful data above 10 to 12K and it's even less likely to be representative of the reality for any given driver.
I understand, high frequencies tend to be a tedious task. I do not know yet what causes the ripples, but found them to be a feature/bug of the JBL PT-waveguides, why I thought the simulation might really be correct. I would like to find out how to get a clean view, let’s see if that is possible with more trials.
As there are no resonances in the axisymmetric case I would be quite surprised if there were any in this case, because there's really nothing more abrupt, only displaced relative to the origin - one would almost expect that this will be actually less prone to resonate. But as we know, intuition can be wrong. Maybe in the direct comparison without a conical exit section of the driver it can be worse. And maybe it will be worse for drivers that don't have any exit section (i.e. where the phase plug slots terminate the driver).
Last edited:
BTW, this is where the cos() function is, if anyone wondered - up to that point it's all conical and an OSSE profile starts from there (in every direction a little different one).
That's just for an explanation what the "Slot" feature does.
Slot.Length = 2 + 20*cos(2*p)^2
That's just for an explanation what the "Slot" feature does.
Slot.Length = 2 + 20*cos(2*p)^2
Attachments
Last edited:
Changed the throat element size from 3 to 2 mm (three different inclinations) -
Attachments
fluid, once again, your oservation ̶o̶f̶ ̶d̶i̶f̶f̶r̶a̶c̶t̶i̶o̶n̶ ̶ remains was right. With a smaller interface offset as opposed to none or an offset that made the high frequencies stray wildly, the response of the device from #8019 looks like this now.
Attachments
Last edited:
Without any warranty, you may try this - the executable I'm currently using: https://at-horns.eu/release/ath-beta-5.zip
The meshing of the throat looks much better with ath5. VACS output looks smooth, apart from something happening at 12k, which might actually and finally be a property of the design now? There is something happening in the impedance curve from the output I posted before. The new beta looks like another advancement. It calls for another reporting script, so no ath report.
Attachments
Here are the new scripts required (under "bin\lib") but be aware that the reporting still works correctly only for axisymmetric horns, mainly regarding the DI. For a general 3D case the surface around the source would have to be sampled differently - that's still one of the many things that are on the "TODO" list....
No, I don't believe so.
Last edited:
I probably wasn't totally in what I wrote, when something looks like diffractive ripple but it's in a place that doesn't make sense and it doesn't follow elsewhere in the response then it is probably a phantom. I don't simulate much above 12k with 3D meshes because it takes longer and the results are always somewhat suspect.
Maybe but probably not
This is with higher throat resolution (3mm) and I will start another run with 2mm later.
Another nice thing with the new version: the limitations I experienced with ath 4.7.1 are gone: I can choose a lower OS.k and also increase length segments again.
Another nice thing with the new version: the limitations I experienced with ath 4.7.1 are gone: I can choose a lower OS.k and also increase length segments again.
Attachments
Last edited:
Because I could not counter the tightening of the pattern above 10k if the slots took too much surface area. The pattern got narrow then. Now that I can choose a lower OS.k, I need to see where that gets me, maybe the slots can be used more aggressive without loss of pattern width.
One question: why do you add a fixed number before you start the cosine fuction? I omitted this completely because I saw no use, but maybe I missed something.
One question: why do you add a fixed number before you start the cosine fuction? I omitted this completely because I saw no use, but maybe I missed something.
Generate the curves for Fusion and have a look to see if the throat slots are nice and smooth there. I think what is left as wiggles in your simulation is not enough mesh resolution to properly describe the slots. You are very much into the weeds here so I would say the simulations (for this version at least) are done
Did you want a flatter DI and leave in the on and near axis ripples?