Zitat:There is no benefit to bringing stl into GMSH because its a mesh-mesh conversion.
a) use Freecad Std mesh stl to ABEC and remesh if required
b) use Freecad Mefisto mesh stl to ABEC and remesh if required
b) use Freecad stp to GMSH msh then to ABEC and remesh if required

There was also the issue of specifying ABEC3 "MeshFrequency" or "EdgeLength" as required in the control solver. It's related to the quality of the simulation's upper freq and can result in an ABEC remesh if it's insufficient. In the ABEC3 help document, the relationship is specified as "EdgeLength=0.5*c/MeshFrequency" where c~344m/s. The FreeCad mefisto stl algorithm can specify edge length, so 20mm edge -> 8600Hz and 30mm edge -> 5733Hz before triggering ABEC remeshing.

I set the mesh parameters to provide approx 2000 element meshes to allow a fair comparison between the algorithms. What is interesting, in the table below, is the dramatic jump in #elements when the imported mesh is too coarse for the specified MeshFrequency and a remesh is triggered. The HAT is aimed at the test disc and the diffraction patterns should be very similar if the mesh surfaces are adequate. My preferred design path will be FreeCad stl direct to ABEC as I can get as good results (as GMSH) with fewer steps. I will also correct a statement I made earlier in post #102, you actually do need to create the 1/4 symmetry mesh before importing into ABEC.

The pictures are :
1) freecad mefisto stl 20mm edge ABEC meshfreq 8Khz and obs 3.6Khz
2) freecad mefisto stl 30mm edge ABEC meshfreq 4Khz and obs 3.6Khz
3) freecad std stl 0.94mm and 30deg dev ABEC meshfreq 4Khz and obs 3.6Khz
4) freecad stp GMSH delaunay 0.33 size ABEC meshfreq 4Khz and obs 3.6Khz
5) mesh comparisons

Zitat:Mir geht es um die Dämpfung des Gehäuses selbst, ohne zusätzliches Material.

Zitat:Mir geht es speziell und die Dämpfung der Gehäusewände und des Ports.

Zitat:Ql is the Q-value due to unwanted enclosure and woofer leakage losses
Qa is the Q-value due to enclosure damping material (or absorption losses)
Qp is the Q-value due to port losses

In normal cases Ql is in the range 5-20. Ql is 7 in most cases, but 15 to 30 if the enclosure has very little leakage. Ql may be 7 even with an enclosure with no leakage because the driver unit itself may have leakages.

A box with moderate lining has a Qa of about 30. A heavily stuffed box can have a Qa of 3. Qa is 7 for about a half-filled enclosure.

A perfect port will have absolutely no losses in which case Qp is very large. A good port, one which does not significantly restrict air flow, has about 100 of Q-value.

Zitat:I have developed this knowledge during developing new enclosure models for Juha Hartikainen’sexcellent freeware speaker design program, WinISD Pro

Zitat:Unter der Schaltfläche < Advanced > können zudem die Gehäuseverluste Ql, Absorbtionsverluste Qa sowie die Verluste der Reflexöffnung Qp eingegeben werden. Die voreingestellten Werte passen schon ganz gut und brauchen nicht verändert zu werden. Nur erfahrene Nutzer sollten hier die Verluste editieren.

Zitat:
; -------------------------------------------------------
; Horn Geometry Definition
; -------------------------------------------------------

ThroatDiameter = 25.4 ; [mm]
ThroatAngle = 0 ; [deg]
Coverage_Horizontal = 100 ; [deg]
Coverage_Vertical = 60 ; [deg]
Depth = 80 ; [mm]
Depth.ConicSectionPart = 0.5 ; 0.5
Shape = raw ; raw | raw2rect
SEExp = 2.3 ; superellipse exponent

Termination = baffle ; baffle | free-air

Zitat:In the OS waveguide a wall profile (as seen from a longitudinal cut) is a conical section, namely hyperbola.Because of its axial symmetry, this curve is the same everywhere. What would happen if we changed the mouth shape arbitrarily and kept the wall profile to be hyperbola at each longitudinal cut? This is the technique employed in ath4 and it proved to work quite well.

Zitat:Of course this is not the whole story because the horn needs to be terminated into free space finally. How smoothly it can be done? It seems that Euler spiral, also known as clothoid, would be the shape to try for its gradual change of curvature. This is also the shape employed in ath4. In the end a moderately sized horn with pretty good beamwidth control and frequency response smoothness everywhere can be designed. For a more detailed treatise and also for the terminology used see the appendix A.

Zitat:What about a rectangle? Superellipse can approach rectangular shape but never actually reach it. Moreover, for high exponents the area near the throat is also quickly shaped to the corresponding s-curve. That might not always be what you would want. For rectangular mouths a different approach was chosen and a second stage of the algorithm devised.Once the regular horn shape is calculated with some lower superellipse exponent, the entire horn is morphed(or better said stretched) into a new mouth shape. In this case it can be a real rectangle, possibly rounded, or anything else.