This is very cool. I saw the large A for Ara multiple entry horns. I need to build some one day, too. I think ATH generated round waveguide would be the way to go.
I really don't know. I have never thought much about flare rates. Can you propose an experiment that would reveal some more detail in a systematic manner?
Last edited:
The idea came from here:
https://www.google.com/url?sa=t&sou...gQFnoECBMQAQ&usg=AOvVaw2w9SFJXhe7Li3VAlsWtewE
Equation #8 and the text that the least reflection occurs for an exponential expansion of the cross section area of the wave front.
An exponential horn assumes flat wave fronts, so maybe try a pure exponential horn and a spherical horn for comparison.
https://www.google.com/url?sa=t&sou...gQFnoECBMQAQ&usg=AOvVaw2w9SFJXhe7Li3VAlsWtewE
Equation #8 and the text that the least reflection occurs for an exponential expansion of the cross section area of the wave front.
An exponential horn assumes flat wave fronts, so maybe try a pure exponential horn and a spherical horn for comparison.
In these waveguides the wavefront gets close to spherical very quickly, so I'm not sure that this has a merit - for a spherical wave a simple conical horn is obvisously the one with the least reflection. I'm still not sure what to take from this.
Last edited:
Hi guys! First post here 😀!
Just managed to start wrapping my head around this software yesterday, alongside ABEC and essentially everything else...
I wanted to share what I've been working on, which is a fairly wide axisymmetric beamwidth waveguide to couple a 1" CD to a 12 inch woofer. CD is planned to be a BMS4552ND with exit angle of 24deg. Believe it or not, my plan was all along to use something axisymmetric for this design, which really helps speed up my workflow re compute times.
My best outcome is as shown below, and I'd like feedback if anyone spots anything particularly erroneous. Dimensions are 314mmx90mm. I've also included the script I used (the attached .txt file) so people can check my work to see if I screwed up something without realizing.
Overall, I'm pretty happy with this outcome so far, and it's definitely taken a lot of tweaking to get to something I'm mostly happy with, at least in theory. I am aware that a baffled design is inherently screwed up in a few ways that an IB simulation can't show, but I'm hoping a wide roundover will take care of me... fingers crossed? I am still a bit of a newbie to all this technical stuff so simulating it all in baffle is somewhat beyond my capabilities.
Cheers!
Just managed to start wrapping my head around this software yesterday, alongside ABEC and essentially everything else...
I wanted to share what I've been working on, which is a fairly wide axisymmetric beamwidth waveguide to couple a 1" CD to a 12 inch woofer. CD is planned to be a BMS4552ND with exit angle of 24deg. Believe it or not, my plan was all along to use something axisymmetric for this design, which really helps speed up my workflow re compute times.
My best outcome is as shown below, and I'd like feedback if anyone spots anything particularly erroneous. Dimensions are 314mmx90mm. I've also included the script I used (the attached .txt file) so people can check my work to see if I screwed up something without realizing.
Overall, I'm pretty happy with this outcome so far, and it's definitely taken a lot of tweaking to get to something I'm mostly happy with, at least in theory. I am aware that a baffled design is inherently screwed up in a few ways that an IB simulation can't show, but I'm hoping a wide roundover will take care of me... fingers crossed? I am still a bit of a newbie to all this technical stuff so simulating it all in baffle is somewhat beyond my capabilities.
Cheers!
All waveguides have design constraints.
William Cowan's baffled waveguides have been recently featured in Stereophile magazine.
https://www.diyaudio.com/community/...y-featured-in-stereophile.395906/post-7271042
Oh, I meant less so that baffled designs are specifically inferior or anything like that, but rather that my simulations would not be representative of the real world use case on a real world baffle. Mostly due to my lack of experience in simulation. ABEC and ATH is intimidating! If only we could cheat the laws of physics...
Those speakers look beyond awesome. I always have had a soft spot for absurd monolith designs with smooth geometry. I think the psychological effect of them alone would elevate the experience for me, even beyond actual sound quality.
Those speakers look beyond awesome. I always have had a soft spot for absurd monolith designs with smooth geometry. I think the psychological effect of them alone would elevate the experience for me, even beyond actual sound quality.
But it will be worse in a flat and finite baffle. I actually think that such designs are inferior, as has been demonstrated in this thread pretty extensively.
It's simple to add an enclosure to the model, even in 1/4 symmetry it's probably worth it to include it, compared to an infinite baffle model. At least horizontally it should give a lot more realistic results. 1/2 symmetry (for a WG above a woofer) would be better but of course calculation times rise steeply.
Last edited:
I mean, yeah, turning out worse in a real baffle is kinda what I’ve been saying… I just am really not into the whole freestanding waveguide look. Personal preference, I’m aware that it’s technically superior.
I may do a half symmetry simulation, just because the box will have a lot of verticality. Do you have any tips for how to script this, beyond the user guide? Thinking a simple box, 30mm roundover. Specifically, looking for tips on how to optimise sim time because while my 5600X is decently quick, any iterations I eventually end up doing will add up time quickly.
Afaik circsym mode is very different to regular 3D when it comes to sim parameters, so I need to unlearn what I have learnt…
Cheers!
I may do a half symmetry simulation, just because the box will have a lot of verticality. Do you have any tips for how to script this, beyond the user guide? Thinking a simple box, 30mm roundover. Specifically, looking for tips on how to optimise sim time because while my 5600X is decently quick, any iterations I eventually end up doing will add up time quickly.
Afaik circsym mode is very different to regular 3D when it comes to sim parameters, so I need to unlearn what I have learnt…
Cheers!
Last edited:
Alas! The hubris of man! My computer weeps in torment, as it is forced to calculate far beyond its means!
But hey, I figured out enclosure modelling.
Only one quadrant will have to do as I tried doing half symmetry and that essentially killed my system, with no progress being made in over an hour. Things now seem SIGNIFICANTLY better with less to calculate, I’ll get some rough preliminary results sometime between now and the heat death of the universe, preferably the former.
Who would’ve thought something like this could be so fun?
But hey, I figured out enclosure modelling.
Only one quadrant will have to do as I tried doing half symmetry and that essentially killed my system, with no progress being made in over an hour. Things now seem SIGNIFICANTLY better with less to calculate, I’ll get some rough preliminary results sometime between now and the heat death of the universe, preferably the former.
Who would’ve thought something like this could be so fun?
It is quite normal when calculating models with lots of elements for things to slow down considerably. The calculation will get to to 25 to 40% reasonably quickly and then appear to halt progress for quite a long time in terms of percentage shown. This is normal and after a while it will jump to 60 percent slow down jump and the last 5% will seem to take a very long time. If the model is below 10,000 elements it will solve eventually but the solve time goes up exponentially as the elements rise.
Some ways to help when iterating is to concentrate the frequencies in the area of interest and not simulate high frequencies if you don't need to. At lower frequencies the mesh can be much coarser and still solve without error. The box width and depth will have the greatest effect on the woofer and lower end of the waveguide so only simulating up to 2K will make things much faster if the mesh is reduced. Keep the number of frequencies down to start with.
An infinite baffle or circsym will always look cleaner than a 3D mesh in free air will, it is the nature of the calculation, look through the noise a little. Avoid significant areas of flat baffle around the waveguide so the edge termination starts as soon as possible, this will give you the result most like a freestanding version would. Making the box shorter in depth than width will increase the low frequency directivity behind and to the sides from the nulls that form from the back corners. Making it deeper will reduce the directivity effects of the box somewhat (really just move them to a lower frequency) but if the baffle is quite narrow compared to the depth the result is not that smooth. The best way depends on what kind of directivity target you have in mind.
Have Paraline lenses been simulated yet? I saw ribbon to horn simulations. I'm curious about Paralines because they shift vertical pattern based on where you locate the compression driver entry. Danley's SBH-10 paraline horn says it can control 140W x 10V using 8 paralines in a 60 inch tall horn/array. I think all he does is shift the entry port to achieve the vertical control but I'm not sure.
Hi everyone! excuse my lack of knowledge on this software and horn design workflow. I'm a broke college engineering student and i have a cnc and want to make some horns for a pair of speakers that i have already made. they are a 2 way with an 8" faitalPRO 8fe200 woofer and a MorelCAT308 1-1/8" soft dome tweeter in a Joseph crower biradial horn design, however this horn is not specifically designed for my morel tweeter. what I'm really trying to know is, what exit angle should I use for a horn attached to a soft dome tweeter? is a soft dome tweeter even a good place to start if i'm trying to horn load it? are there any interesting designs that have been made with ATH4 that would work that are open source?
I've not really sim'd really large devices before. What Mesh settings would you recommend for an overnight run / final check of a design approx 650 x 650 x 240 deep for 1" CD upto about 15k.
Ie,
Code:
Horn.Adapter = {
Segments = 14 ;
ZMap = 0.5,0.4,0.5,0.5
}
Horn.Part:1 = {
Segments = 32 ;
ZMap = 0.5,0.3,0.5,0.9
}
Mesh.AngularSegments = 32
Mesh.LengthSegments = 10
Mesh.ThroatResolution = 12.0 ; [mm]
Mesh.MouthResolution = 24.0 ; (mm)
Last edited:
After spending some time down the dark path of waveguide research i came across some recent articles authored by Mark Dodd
& Jack Oclee-Brown that i have not seen mentioned anywhere on this site yet. They describe a new method of horn path length correction using corrugations/ripples to create N1P wavefronts which appear to have some pretty reasonable advantages. The AES paper attached (and patent EP3806086A1) gives a pretty good explanation on on approach but does mention "The manual design process used for these examples is time-consuming and intricate" which does initially make it sound fairly intimidating. Luckily some more searching found a recent academic thesis by Lewis MacDonald who has since optimised the design more and even discussed a pathway for automating the design process. The thesis on University Of Salford site is not public but there is a version on ioa.org.uk which was recently presented at a conference that appears to have all the important information but much a smaller than the file on the uni site so maybe doesn't have the fully analysis. File size is too big for the last one too attach so here is the link: https://www.ioa.org.uk/catalogue/paper/metric-based-development-acoustic-lenses-and-waveguides
What are the chances this could be added to Ath4 at some point in the future as drawing one of these up by hand does not appear realistic.
& Jack Oclee-Brown that i have not seen mentioned anywhere on this site yet. They describe a new method of horn path length correction using corrugations/ripples to create N1P wavefronts which appear to have some pretty reasonable advantages. The AES paper attached (and patent EP3806086A1) gives a pretty good explanation on on approach but does mention "The manual design process used for these examples is time-consuming and intricate" which does initially make it sound fairly intimidating. Luckily some more searching found a recent academic thesis by Lewis MacDonald who has since optimised the design more and even discussed a pathway for automating the design process. The thesis on University Of Salford site is not public but there is a version on ioa.org.uk which was recently presented at a conference that appears to have all the important information but much a smaller than the file on the uni site so maybe doesn't have the fully analysis. File size is too big for the last one too attach so here is the link: https://www.ioa.org.uk/catalogue/paper/metric-based-development-acoustic-lenses-and-waveguides
What are the chances this could be added to Ath4 at some point in the future as drawing one of these up by hand does not appear realistic.
It looks like this works for flat/thin waveguides that use the 3rd dimension to make the path length equal across the waveguide.
Kinda like how a piece of cloth behaves:
I think we would need a fourth dimension to achieve this for the horns we use here.
Full brainstorming mode:
It needs the path length to be the same, so if we use a different (meta) material with different sound speeds as a function of the radius... That could be the fourth dimension.
Kinda like how a piece of cloth behaves:
I think we would need a fourth dimension to achieve this for the horns we use here.
Full brainstorming mode:
It needs the path length to be the same, so if we use a different (meta) material with different sound speeds as a function of the radius... That could be the fourth dimension.
I could imagine a radial array of those thin ducts, in a kind of a multicell approach. But something tells me that probably more new problems would arise than would be solved.
Last edited:
- Currently I'm in a process of printing a mold for a "regular" R-OSSE, a pretty big one (⌀540 mm) for a 1.4" driver.
If it works as predicted (which I already have almost no doubts about), I'd be more than happy:
If it works as predicted (which I already have almost no doubts about), I'd be more than happy: