Another measured piece: A460D + B&C DE360
0-deg adapter used; roughly 0/10/20/30 deg off axis, as always; not an actual SPL
0-deg adapter used; roughly 0/10/20/30 deg off axis, as always; not an actual SPL
And back to something serious 🙂
Continuing with the BMS 4554, finishing the Gen2 adapter -
There may still be some imperfections at the joins, they are not sealed properly yet.
I think I already saw it a bit smoother around 9 kHz but overall this is what I like a lot.
- These were the initial experiments:
Continuing with the BMS 4554, finishing the Gen2 adapter -
There may still be some imperfections at the joins, they are not sealed properly yet.
I think I already saw it a bit smoother around 9 kHz but overall this is what I like a lot.
- These were the initial experiments:
BMS 4554 extended, with an adapter starting right at the exit of the (ring) diaphragm, which is roughly ⌀16 mm.
Still keeping the directivity of Gen2, I think this is actually pretty cool.
Shaping the response is no issue with a DSP. Perhaps it could be further optimized in the adpater, hard to say.
(Measured with A520G2; at two different distances from the mouth.)
- Did a quick sweep for harmonics, at two different levels, just to have an idea:
2nd, 3rd, 4th:
Seems to me like 700 Hz would still be no...
Still keeping the directivity of Gen2, I think this is actually pretty cool.
Shaping the response is no issue with a DSP. Perhaps it could be further optimized in the adpater, hard to say.
(Measured with A520G2; at two different distances from the mouth.)
- Did a quick sweep for harmonics, at two different levels, just to have an idea:
2nd, 3rd, 4th:
Seems to me like 700 Hz would still be no...
With no smoothing, other than a ~5ms IR window:
Below with 10 and 100 ms window. Reflections but also an evidence that there's no problem near the low end:
5 ms vs 100 ms window, smoothed:
Below with 10 and 100 ms window. Reflections but also an evidence that there's no problem near the low end:
5 ms vs 100 ms window, smoothed:
Last edited:
Where would you actually cross this now? Still >700hz as in your initial testing?With no smoothing, other than a ~5ms IR window:
Have you had a look at harmonics yet?
600hz would be an absolutely stellar result with that smoothness. And it’d make choosing a bass/mid bass so much easier.
600hz would be an absolutely stellar result with that smoothness. And it’d make choosing a bass/mid bass so much easier.
Could you show your measurement setup? Your measurement is really clean. I'm jealous...
Follow-up for any who are interested. It turns out that the effect I measured has been known for some time. Here's a particularly relevant paper which shows that porous absorbers do indeed exhibit significant odd-order nonlinearity at high SPL: D. A. Nelson, "Interaction of finite-amplitude sound with air-filled porous materials" (1985).
Should I be concerned about these holes?
My 3D optimizations aren't going as well as my 2D ones, and unfortunately they don't seem to translate well from 2D to 3D either.
Does it have an impact on the simulation or is it just bug in the depiction of ABEC?
Also any input is welcome how to simulate waveguides in 3D reliably. This is my current config, but the results aren't very good. 2D I can get excellent results, 3D seems really random.
Also any input is welcome how to simulate waveguides in 3D reliably. This is my current config, but the results aren't very good. 2D I can get excellent results, 3D seems really random.
OSSE = {
r0 = 14.0
a0 = 51.15
a = 61.2
k = 10
L = 19.2
s = 1.8
n = 3.18
q = 0.995
}
Mesh.ZMapElementSize = 0.1,0.3,0.25,0.85
_Source.Contours = {
dome WG0 22 5 3 -1 3 1
}
Source.Contours = {
zoff 0
point p1 4.9 0 4
point p2 0.75 10.5 0.75
point p3 0 10.5 1
point p4 -0.25 10.75 1
point p5 0 11 1
point p6 1 12 1
point p7 0 13 2
point p8 0 14.5 5
cpoint c1 -15.36 0
cpoint c2 0 10.75
cpoint c3 0 12
arc p1 c1 p2 1
line p2 p3 1
arc p3 c2 p4 1
arc p4 c2 p5 1
arc p5 c3 p6 0.75
arc p6 c3 p7 0.25
line p7 p8 0
}
Morph.TargetShape = 1
Morph.FixedPart = 0.1
Morph.Rate = 3
Morph.AllowShrinkage = 0
_Morph.TargetWidth = 130
_Morph.TargetHeight = 130
Morph.CornerRadius = 1
Source.Velocity = 2 ; axial motion
Mesh.ZMapPoints = 0.3,0.2,0.7,0.9
Mesh.Enclosure = {
Spacing = 15,15,15,15
Depth = 150
EdgeRadius = 15
EdgeType = 1
FrontResolution = 6,6,16,16,
BackResolution = 24,24,24,24
}
Mesh.Quadrants = 1 ; =1 for 1/4 symmetry,
;Mesh.VerticalOffset = 80
Mesh.AngularSegments = 100
Mesh.LengthSegments = 20
;Mesh.CornerSegments = 2
Mesh.SubdomainSlices =
Mesh.ThroatResolution = 4
Mesh.MouthResolution = 12
_Mesh.InterfaceResolution = 8
Mesh.RearResolution = 20
ABEC.SimType = 2 ; 1 = Infinite Baffle, 2 = Free Standing
ABEC.f1 = 1000 ; [Hz]
ABEC.f2 = 20000 ; [Hz]
ABEC.NumFrequencies = 63
ABEC.MeshFrequency = 1000 ; [Hz]
ABEC.Polars:SPL = {
MapAngleRange = 0,90,19
NormAngle = 10 ; [deg]
Distance = 3 ; [m]
Offset = 95 ; [mm]
Inclination = 90
}
Output.STL = 0
Output.ABECProject = 1
Report = {
Title = "ATH report. All praise mabat!"
NormAngle = 10
Width = 1200
Height = 800
SPL_Range = 50
MaxRadius = 90
PolarData = "SPL"
}
r0 = 14.0
a0 = 51.15
a = 61.2
k = 10
L = 19.2
s = 1.8
n = 3.18
q = 0.995
}
Mesh.ZMapElementSize = 0.1,0.3,0.25,0.85
_Source.Contours = {
dome WG0 22 5 3 -1 3 1
}
Source.Contours = {
zoff 0
point p1 4.9 0 4
point p2 0.75 10.5 0.75
point p3 0 10.5 1
point p4 -0.25 10.75 1
point p5 0 11 1
point p6 1 12 1
point p7 0 13 2
point p8 0 14.5 5
cpoint c1 -15.36 0
cpoint c2 0 10.75
cpoint c3 0 12
arc p1 c1 p2 1
line p2 p3 1
arc p3 c2 p4 1
arc p4 c2 p5 1
arc p5 c3 p6 0.75
arc p6 c3 p7 0.25
line p7 p8 0
}
Morph.TargetShape = 1
Morph.FixedPart = 0.1
Morph.Rate = 3
Morph.AllowShrinkage = 0
_Morph.TargetWidth = 130
_Morph.TargetHeight = 130
Morph.CornerRadius = 1
Source.Velocity = 2 ; axial motion
Mesh.ZMapPoints = 0.3,0.2,0.7,0.9
Mesh.Enclosure = {
Spacing = 15,15,15,15
Depth = 150
EdgeRadius = 15
EdgeType = 1
FrontResolution = 6,6,16,16,
BackResolution = 24,24,24,24
}
Mesh.Quadrants = 1 ; =1 for 1/4 symmetry,
;Mesh.VerticalOffset = 80
Mesh.AngularSegments = 100
Mesh.LengthSegments = 20
;Mesh.CornerSegments = 2
Mesh.SubdomainSlices =
Mesh.ThroatResolution = 4
Mesh.MouthResolution = 12
_Mesh.InterfaceResolution = 8
Mesh.RearResolution = 20
ABEC.SimType = 2 ; 1 = Infinite Baffle, 2 = Free Standing
ABEC.f1 = 1000 ; [Hz]
ABEC.f2 = 20000 ; [Hz]
ABEC.NumFrequencies = 63
ABEC.MeshFrequency = 1000 ; [Hz]
ABEC.Polars:SPL = {
MapAngleRange = 0,90,19
NormAngle = 10 ; [deg]
Distance = 3 ; [m]
Offset = 95 ; [mm]
Inclination = 90
}
Output.STL = 0
Output.ABECProject = 1
Report = {
Title = "ATH report. All praise mabat!"
NormAngle = 10
Width = 1200
Height = 800
SPL_Range = 50
MaxRadius = 90
PolarData = "SPL"
}
It’s not a bug in ABEC the mesh has been generated with the normals reversed on some elements. It will simulate as if you have a hole in the waveguide. The effect depends on the size of the hole. For a one off fix you change the normals of the elements that are reversed but it is a bit of a pain to do so.
Try to decrease the value of Mesh.ThroatResolution (to avoid a large difference of neighbouring element sizes).
Last edited:
This certainly explains why my 3D optimizations are going nowhere.
I found example from you that looks like it has way higher resolution than my simulations: https://www.diyaudio.com/community/...he-easy-way-ath4.338806/page-628#post-7357260
And also an example how to set an enclosure by defining lines, which I asked about a few days ago: https://www.diyaudio.com/community/...he-easy-way-ath4.338806/page-481#post-6977513
I'll try all of that the next few days.
For now I have one more question: Does merging two good working circular geometries with the same length work well? Circ. symmetric simulations are so much quicker, so combining a good circular waveguide on the horizontal axis with a narrower waveguide on the vertical axis would save a lot of time simulating. Since my 3D simulations often had these holes I can't tell yet myself, yet, but I'll add more variables to my optimizer script and let it run over the weekend - if no one tells me otherwise. 🙂
I'll have to set up my rating scripts to read out horizontal and vertical SPL first, too. Quite a bit of work, but this is a fun project. Thanks for making that possible with ath and thanks to anyone helping here!
I found example from you that looks like it has way higher resolution than my simulations: https://www.diyaudio.com/community/...he-easy-way-ath4.338806/page-628#post-7357260
And also an example how to set an enclosure by defining lines, which I asked about a few days ago: https://www.diyaudio.com/community/...he-easy-way-ath4.338806/page-481#post-6977513
I'll try all of that the next few days.
For now I have one more question: Does merging two good working circular geometries with the same length work well? Circ. symmetric simulations are so much quicker, so combining a good circular waveguide on the horizontal axis with a narrower waveguide on the vertical axis would save a lot of time simulating. Since my 3D simulations often had these holes I can't tell yet myself, yet, but I'll add more variables to my optimizer script and let it run over the weekend - if no one tells me otherwise. 🙂
I'll have to set up my rating scripts to read out horizontal and vertical SPL first, too. Quite a bit of work, but this is a fun project. Thanks for making that possible with ath and thanks to anyone helping here!
Short of the fact that narrower may require different length for best results, I think it mostly works and to me it would make sense to start with such profiles (i.e. which work well as axisymmetric alone). Be ready for surprises though.
Smolen and Halley's paper on optimization of 3D horns has some interesting insights in that respect: https://secure.aes.org/forum/pubs/conventions/?elib=20684
They started with 2D simulations of the vertical and horisontal, did an iterative (automated) optimization of those, and then combined them. They found that the vertical and horisontal profiles had some mutual influence, such that further optimization in 3D would be required to fully optimize the design. It's been a while since I read it, so I don't remember the full details, but it's fairly interesting.
They started with 2D simulations of the vertical and horisontal, did an iterative (automated) optimization of those, and then combined them. They found that the vertical and horisontal profiles had some mutual influence, such that further optimization in 3D would be required to fully optimize the design. It's been a while since I read it, so I don't remember the full details, but it's fairly interesting.
That's the A520G2. As for the BMS4554, I don't know, at least a week or two, it's still a lot of work.
BTW, the measurements on the A520G2 I use are still the cleanest I have seen. Even my A460 (which should be very close) is not that clean in the midrange, and I still don't know why. I'm only starting to suspect it's because the A520G2 is warped around the mouth due to a "sunlight accident" (it's visible in some of the photos). This is still something to explore. It would be easy to add some intentional "warping" if this prooved to be true. It's only the surface, one wouldn't expect this to have a strong effect, but who knows. I'm about to assemble an A460G2. Can try a warped version afterwards.
This is one such comparison - A520G2 vs A460 (thick line):
BTW, the measurements on the A520G2 I use are still the cleanest I have seen. Even my A460 (which should be very close) is not that clean in the midrange, and I still don't know why. I'm only starting to suspect it's because the A520G2 is warped around the mouth due to a "sunlight accident" (it's visible in some of the photos). This is still something to explore. It would be easy to add some intentional "warping" if this prooved to be true. It's only the surface, one wouldn't expect this to have a strong effect, but who knows. I'm about to assemble an A460G2. Can try a warped version afterwards.
This is one such comparison - A520G2 vs A460 (thick line):
Last edited:
