I'm trying to design a set of speakers that my wife and I both like (thread). The direction I have for the moment is integrating two subwoofers in the TV-console and using an on-wall solution for the rest of the frequency range (>100 Hz). The cornu spiral is one idea, it's a little wide for the room. So now I want to try to design an on-wall speaker that looks and sounds nice. This thread is mainly to try to simulate frequency response (using FEM-software) for an on wall-speaker, trying different geometries.
Let's start with something that is easy to build and simulate - a half sphere with a full-range driver. I'll use the MarkAudio Alpair 7.3 as that is one of my candidates (reasonable price and performance, good experience with the CHR-70).
I started with a reference, in-wall speaker. The response was first simulated with Basta!, in a 4.8 l closed box.
Basta! uses a flat piston approximation (as most softwares) and gives me the cone acceleration for this combination of driver and box volume. I use this as input to COMSOL Multiphysics (the FEM-software I use). In an infinite baffle I get very similar response, the only difference comes from that the (approximated) cone shape is added in the FEM-model.
Quite flat, as expected.
Let's try the driver in a 280 mm half sphere on-wall, this is the model:
And this the response:
That doesn't look so good... Let's try a larger half-sphere.
Ok, half-spheres on-wall is probably not a good idea. There is some large fluctuations depending on both frequency and off-axis angle. If the response would be made flat on-axis it would still be terrible off-axis.
How about a thinner version? 600 mm radius, but with a height of 120 mm (instead of 190 mm for the 400 mm half-sphere).
Better, but nah. I'm pretty sure I need to go onto something that isn't axi-symmetric, so I can place the driver offset from the center. This means going over to a 3D-model... I'll try tomorrow 🙂
Assumptions used in the models:
Infinitely stiff cone.
Inside of enclosure only modeled by stiffness (air spring), not geometry - no internal reflections.
The wall is infinite and infinitely stiff.
No other walls/ceilings/...
The cone and half the surround has the set acceleration. The rest does not move at all.
/Anton
Let's start with something that is easy to build and simulate - a half sphere with a full-range driver. I'll use the MarkAudio Alpair 7.3 as that is one of my candidates (reasonable price and performance, good experience with the CHR-70).
I started with a reference, in-wall speaker. The response was first simulated with Basta!, in a 4.8 l closed box.
Basta! uses a flat piston approximation (as most softwares) and gives me the cone acceleration for this combination of driver and box volume. I use this as input to COMSOL Multiphysics (the FEM-software I use). In an infinite baffle I get very similar response, the only difference comes from that the (approximated) cone shape is added in the FEM-model.
Quite flat, as expected.
Let's try the driver in a 280 mm half sphere on-wall, this is the model:
And this the response:
That doesn't look so good... Let's try a larger half-sphere.
Ok, half-spheres on-wall is probably not a good idea. There is some large fluctuations depending on both frequency and off-axis angle. If the response would be made flat on-axis it would still be terrible off-axis.
How about a thinner version? 600 mm radius, but with a height of 120 mm (instead of 190 mm for the 400 mm half-sphere).
Better, but nah. I'm pretty sure I need to go onto something that isn't axi-symmetric, so I can place the driver offset from the center. This means going over to a 3D-model... I'll try tomorrow 🙂
Assumptions used in the models:
Infinitely stiff cone.
Inside of enclosure only modeled by stiffness (air spring), not geometry - no internal reflections.
The wall is infinite and infinitely stiff.
No other walls/ceilings/...
The cone and half the surround has the set acceleration. The rest does not move at all.
/Anton
Anton - this next sentence will exhaust my technical acumen when it comes to computer simulations:
The first assumption on your list is incorrect - "full range" drivers in general derive their HF bandwidth specifically by not being infinitely stiff.
Internal shapes, and certainly in the case of drivers with very lightweight cones such as any of the Mark Audio metals, very shallow depths, will very likely cause issues with internal reflections that I'm not sure would be revealed in any modeling
The first assumption on your list is incorrect - "full range" drivers in general derive their HF bandwidth specifically by not being infinitely stiff.
Internal shapes, and certainly in the case of drivers with very lightweight cones such as any of the Mark Audio metals, very shallow depths, will very likely cause issues with internal reflections that I'm not sure would be revealed in any modeling
You are absolutely correct, it's mainly the LF and MF area that is possible to simulate by this method. I do however have a model that takes the Thiele & Small parameters and uses these to model the behavior with an arbitrary box geometry (and yes, even ported/horns/TL...). This is still with an infinitely stiff cone. It can however be expanded to include the complete cone, if the geometry and material is known.
For the optimization of the geometry for on-wall placement that shouldn't be necessary. As the behavior in the HF area is hardly affected by box shape and placement.
/Anton
what you actually hear will be affected by enclosure location and listening position relative to direct axis - fixed on / in wall placement certainly adds some constraints
Disclaimer - based on my recommendation, a dozen or so years ago my sister bought a pair of small wall mounted 2-ways made by a local hi-fi shop. They sounded fine in the showroom - if you sat in exactly the right spot - but when installed in short walls on either side of a wide archway, with "primary" listening positions consisting of 2 couches parallel to their broadcast axis, they were considerably underwhelming. I've been rather disinclined towards wall mounts ever since.
Sorry, forgot to mention the dependence on placement of listener and wall/ceiling reflections. But the HF area is less dependent on the shape of the box (except in extreme cases).
I base this more on math and simulation than on listening impressions. I have heard too few on-wall speakers and don't really trust my ears that much (yet). 🙂
/Anton
This software sounds perfect for simulating a "U" shaped cutout beneath the full range driver in a FAST OB to improve directivity and imaging, see topic here http://www.diyaudio.com/forums/full-range/249984-cheap-fast-ob-literally-20.html#post3863096 . However, I could not find anything related to Basta! for audio sims. Can you please post to FEM software you are using?
I have not seen that thread. But yes, you should be able to simulate any geometry. Only drawback is that you need som serious computer power for larger simulations. A simple 2D simulation takes a couple of minutes on a 3570K, 3D simulations take at least 20 minutes. For 3D-simulations of geometries without symmetries and high frequencies (above 10 kHz) require many hours of simulation time.
As I said I'm using COMSOL Multiphysics. I use this mainly for structural dynamics and heat transfer simulations at work.
/Anton
Also there's the Planet10 CHR-70 on-wall & in-wall designs to look, [FONT=Verdana, Helvetica, Arial, san-serif]Home Theatre Appendix[/FONT] ... Chris could clarify this, but the A7.3 should work in those enclosures.
Last edited:
- Status
- Not open for further replies.