Simulation Analysis - Screws in Waveguide

[StuartC]

This is probably the single best thread I've seen so far on this (excellent) forum. This is AudioScienceReview at its best.

 

[René - Acculution.com]

This is probably the single best thread I've seen so far on this (excellent) forum. This is AudioScienceReview at its best.

Much appreciated Stuart, thanks a lot. I have many other topics that I can talk about, so I will try to find the time.

 

[Rottmannash]

It's not tricky at all, it just adds to the cost. For example, here is the old Thiel CS 2.3:

View attachment 127466View attachment 127468

Those are beautiful.

 

[sarumbear]

One I like to see are diffractions around the edges of a rectangular speaker.

I second that. I would love to see a modern analysis of Olsen’s empirical one.

 

[hardisj]

Rene,
Thank you for taking the time to do this for me. It's good to see some numbers instead of just "I think". As I go forward, when I encounter speakers with waveguide assemblies I suspect may be a problem I will reach out to you with physical measurements and see if you have the time to model them. I'd like to test this myself at some point as well and maybe we can see how the prediction through simulation lines up with measurements.

I do wonder if the symmetry of the placement will also effect things. For example, maybe the response would have more of an on-axis issue but off-axis the symmetry is OK. And vice versa. ??

 

[Rick Sykora]

Audioholics has measurements of different grilles:
https://www.audioholics.com/loudspeaker-design/grilles

Some are definitely worse than others. I bought the Infinity R-series towers and center and while initially thought the grilles were very cheap (plastic, not wood) I came to think they did so to minimize interference (well, and shipping cost).

Just to pile on...

No need to waste compute time to simulate grille diffraction. It has been measured a lot and almost every speaker has shown to be problematic. Most grilles are for protection. If you want the best sound and cannot measure the difference for your own speakers, simply take the grilles off.

 

[MarkS]

Or buy speakers designed and engineered to be used with grills on, which is what I do.

 

[Rick Sykora]

Or buy speakers designed and engineered to be used with grills on, which is what I do.

Sounds like a good plan, but unless the manufacturer can provide trustworthy proof, you would still need to be able to do test it yourself.

Too many claim transparent grilles today and testing has shown that they are not.

 

[fluid]

Thanks. I think Geddes mentioned this thread in an interview I saw recently but not sure.

Yes this is the thread Earl Geddes mentioned, the tool Marcel has created there is fantastic. Here is an example of the sort of directivity control that can be achieved. A number of them have been built and measured and when you take out the differences between the different drivers the simulated results are very close to actual.

Comsol is a great tool for those with access to it. Ath will generate ABEC BEM scripts. ABEC runs in a demo mode without being able to save the results or a full version for non commercial use can also be had for free by requesting it from the developer R&DTeam.

 

[richard12511]

It's not tricky at all, it just adds to the cost. For example, here is the old Thiel CS 2.3:

View attachment 127466View attachment 127468

If you account for the grill in your design, though, you mess up the sound for those who like exposed drivers. I guess if you have an active speaker, you can ship the speaker with different DSP setting to account for the treble loss of the grill.

 

[MarkS]

Sounds like a good plan, but unless the manufacturer can provide trustworthy proof, you would still need to be able to do test it yourself.

Too many claim transparent grilles today and testing has shown that they are not.

It's easy enough to tell from the design whether or not there is extra diffraction from the grille frame. The cloth itself almost always has a minimal impact.

 

[Tks]

This is great and all but inconsiquential the moment I fire up an MQA track (guaranteed to hear what the producer and/or artist heard at the studio).


sorry I couldnt help myself

 

[richard12511]

This is great and all but inconsiquential the moment I fire up an MQA track (guaranteed to hear what the producer and/or artist heard at the studio).


sorry I couldnt help myself

MQA has solved the "circle of confusion".

 

[JustAnandaDourEyedDude]

Software: COMSOL Multiphysics
Method: Finite Element
Geometry: Realistic 1" tweeter geometry, arbitrary waveguide (5 cm radius, 2 cm depth), 4 screws (0.67 cm diameter, protuding 1.2 mm), in baffle
Mesh: Quadratic Lagrangian elements (6 elements per wavelength @20 kHz)
DOFs: ≈ 262,000
Other: The exact same mesh was used for both cases; with and without screws. Calculation time on laptop around 1 hour per case. This is pure acoustics case, so there is no explicit solid mechanics; instead an acceleration is put on the tweeter surface.

Conclusions -
For this particular case the influence of the screws in fairly minimal, yet it is there. The wavelength even at 20 kHz is 1.7 cm, which is larger than the screws themselves, and so it is to be expected that the sound field will not be affected much. It should be noted that such an analysis should be carried out on a case by case basis, as waveguides (and screws) differ from product to product, and if we get near 1 dB difference it could be relevant.

You have my admiration for an excellent and rare contribution of an acoustics simulation to ASR. Very impressive that you are working in acoustics and microacoustics using FEM and BEM.

The quadratic Lagrangian solution representation within each element presumably means the solution will be third-order accurate in the mesh spacing (element size), so the 6 elements per min. wavelength should be enough to resolve the volumetric acoustics. However, my question is whether you used mesh clustering (local refinement) at the parts of the boundary representing the protrusions of the screw heads. The average element diameter of 2.9mm exceeds the protrusion height of 1.8mm and is about half of the screw-head diameter. If you say the exact same mesh was used for the cases both with and without the screws, I wonder how the presence of the screws is accounted for in the simulation if clustering is not used to resolve the geometric detail. Also, I have not come across this acceleration condition you mention on the tweeter surface. From what I recall, in most codes, one specifies a sinusoidally oscillating (at the specified frequency) normal velocity component (i.e., along the tweeter axis) at the relevant boundary. Without understanding the resolution of the acoustic interaction at the screw-head boundary, and doing a mesh-refinement study, I would lack confidence to say that influence was a 0.5dB or 1dB or 0.25dB deviation in the SPL.

 

[René - Acculution.com]

I second that. I would love to see a modern analysis of Olsen’s empirical one.

That is next on the list then

 

[René - Acculution.com]

Yes this is the thread Earl Geddes mentioned, the tool Marcel has created there is fantastic. Here is an example of the sort of directivity control that can be achieved. A number of them have been built and measured and when you take out the differences between the different drivers the simulated results are very close to actual.

View attachment 127571

Comsol is a great tool for those with access to it. Ath will generate ABEC BEM scripts. ABEC runs in a demo mode without being able to save the results or a full version for non commercial use can also be had for free by requesting it from the developer R&DTeam.

Thanks, I will have a look. I think that general tweeter geometries, there could be an advantage in starting from analytical work, but then applying shape optimization on top.

 

[René - Acculution.com]

This is great and all but inconsiquential the moment I fire up an MQA track (guaranteed to hear what the producer and/or artist heard at the studio).


sorry I couldnt help myself

Shots fired.

 

[René - Acculution.com]

You have my admiration for an excellent and rare contribution of an acoustics simulation to ASR. Very impressive that you are working in acoustics and microacoustics using FEM and BEM.

The quadratic Lagrangian solution representation within each element presumably means the solution will be third-order accurate in the mesh spacing (element size), so the 6 elements per min. wavelength should be enough to resolve the volumetric acoustics. However, my question is whether you used mesh clustering (local refinement) at the parts of the boundary representing the protrusions of the screw heads. The average element diameter of 2.9mm exceeds the protrusion height of 1.8mm and is about half of the screw-head diameter. If you say the exact same mesh was used for the cases both with and without the screws, I wonder how the presence of the screws is accounted for in the simulation if clustering is not used to resolve the geometric detail. Also, I have not come across this acceleration condition you mention on the tweeter surface. From what I recall, in most codes, one specifies a sinusoidally oscillating (at the specified frequency) normal velocity component (i.e., along the tweeter axis) at the relevant boundary. Without understanding the resolution of the acoustic interaction at the screw-head boundary, and doing a mesh-refinement study, I would lack confidence to say that influence was a 0.5dB or 1dB or 0.25dB deviation in the SPL.

The isoparametric quadratic setup ensures that both geometry and sound field are resolved to second-order. Both the tweeter surface and screw surfaces are resolved with 24(!) elements per wavelength. A convergence study is beyond the scope of this analysis, but for a client I would probably dive a little deeper into certain aspects of this. The same mesh can be used, because I apply internal hard walls on the screws and turn that on and off, respectively. Otherwise the meshes will differ a little, so this is the best that can be done.

Regarding acceleration vs velocity (both are sinusoidal, no issue there), sound pressure level is generally flat across frequency for a constant acceleration. So whatever input I have will give the relative difference between screws and no screws.

Great to hear your experience in numerical work, very impressive. The N-S equation are the basis for the microacoustics that I work with, but linearised. COMSOL has indeed come a long way, but luckily you can still input your own equations, which I also take advantage of. I have used BEM a lot for exterior acoustics, and it is now built into COMSOL (used OpenBEM a lot in the past).

 

[DualTriode]

Hello All,

There is nothing here.

This is as it says a Saturday afternoon musing with some simulation graphics, a graphic what if.

I am more impressed with @amirm ‘s measurements, with and without a “transparent” grille cloth.

Thanks DT

 

[vkvedam]

Excellent work! Thanks @René - Acculution.com