I use FreeCad to create any additional parts for the horn models. Assuming you have a point cloud or a meshed horn surface, you can import that into FreeCad. The driver membrane is created using [Part WB]->[Create Shape]->[Wire] and you select the throat edges to create the wire (ie. polyline as @fluid stated). You can then create a face (ie. surface) for the driver membrane from that wire and export it as a step file. GMSH can mesh the step file to any resolution you need while preserving the polygon edges and nodes required to match the horn throat (ie. conformal). It's always an exact edge match that's independent of mesh resolution, which is important for the simulation.
Last edited:
Thank you guys for the help. The solution makes sense to use polylines to capture the edges. The problem was caused because both meshes were created based on a circle (created with the revolve tool in FreeCad).
For mesh size it is suggested that elements need to be less than half-wavelength of the highest frequency under investigation. Also, the elements should be small enough to capture finer details. DonVK suggested the following mesh sizes for a 400Hz Horn up to 1000Hz:
"For examples a 400Hz horn simulated up to 10Khz, needs the driver membrane meshed at 5mm (150 elements), the inner horn at 10mm (2800 elements), outer horn at 30mm (1000 elements) mouth interface at 10mm (1000 elements), and edge at 20mm (100 elements)."
Half wavelength of 10kHz is approximately 17mm, so how does one decide to use a smaller mesh size for the driver (flat disc) than for example the curved inner horn section? Or are these choices based on experimentation?
For mesh size it is suggested that elements need to be less than half-wavelength of the highest frequency under investigation. Also, the elements should be small enough to capture finer details. DonVK suggested the following mesh sizes for a 400Hz Horn up to 1000Hz:
"For examples a 400Hz horn simulated up to 10Khz, needs the driver membrane meshed at 5mm (150 elements), the inner horn at 10mm (2800 elements), outer horn at 30mm (1000 elements) mouth interface at 10mm (1000 elements), and edge at 20mm (100 elements)."
Half wavelength of 10kHz is approximately 17mm, so how does one decide to use a smaller mesh size for the driver (flat disc) than for example the curved inner horn section? Or are these choices based on experimentation?
There are many guidelines for setting the mesh resolution (edge length) that vary from 1/2 to 1/6 lambda. I try to balance the #elements against the simulation time which will increase rapidly (cubic) with #elements in the model. The guideline I used is from many experiments trying to get clean results with moderate simulation times. A good way to judge your model is by looking at the polar plots. They should be well behaved (no abrupt changes or discontinuities). When you see a "shattered" polar plot with lots of noise your model is broken starting from the frequency the effect occurs. The problem could be a non-conformal mesh (holes) or mesh resolution that's too low.
If your horn is relatively small you could mesh the entire model together at the same time. However, if the horn is large, or you want to mount it in a cabinet with other drivers you need to consider the #elements for each component in order to have reasonable simulation times.
If your horn is relatively small you could mesh the entire model together at the same time. However, if the horn is large, or you want to mount it in a cabinet with other drivers you need to consider the #elements for each component in order to have reasonable simulation times.
If you use an interface between subdomains then that also needs to match the vertex points in the same way as the driver. If it doesn't it is like having a hole in the mesh which will just produce garbage results.
If your waveguide is axisymmetric then look at using circsym in ABEC, the mesh frequency can be up to 80KHz and still take only a few minutes to solve.
I've just started to play about with 3D printing horns. I have used the linked spreadsheet for a couple so far:
First I made a round 850Hz horn with a simple hypex profile
Then I made an elliptical one:
The workflow is somewhat cumbersome. I have found that the best way is as follows:
The system is a 5 way active horn setup. I'm looking to improve results for the JBL 2461 (1", about 1.6k-8k) and 2441 (2", 500Hz-1.6k). So far the round horn is a real improvement over the 800Hz tractrix horn I was using. The second elliptical one is currently printing!
I also have the JM Le Cléac'h quasi elliptical spreadsheet and have written a macro to extract a useable point cloud. They will be printed next, although even the 800Hz one will need to be printed in two pieces. The ones for the 2441 will be a jigsaw!
I'd like to offer hearty thanks to DrBA/docali for creating and sharing his work, it's really cool. Without it I don't think I would have ventured out onto this path.
First I made a round 850Hz horn with a simple hypex profile
Then I made an elliptical one:
The workflow is somewhat cumbersome. I have found that the best way is as follows:
- Design in the DrBA spreadsheet and export point cloud
- Open in Meshlab to make into surfaces
- Open in InstantMeshes to convert to a quad mesh
- Open in Fusion 360 to convert to t-splines, then into a body and create the model
The system is a 5 way active horn setup. I'm looking to improve results for the JBL 2461 (1", about 1.6k-8k) and 2441 (2", 500Hz-1.6k). So far the round horn is a real improvement over the 800Hz tractrix horn I was using. The second elliptical one is currently printing!
I also have the JM Le Cléac'h quasi elliptical spreadsheet and have written a macro to extract a useable point cloud. They will be printed next, although even the 800Hz one will need to be printed in two pieces. The ones for the 2441 will be a jigsaw!
I'd like to offer hearty thanks to DrBA/docali for creating and sharing his work, it's really cool. Without it I don't think I would have ventured out onto this path.
I used the JMLC spreadsheet. I'm not sure why tbh - it looks like you have developed the SWH one further, not least with documentation!
Meshlab does have a tri to quad filter, but it's not terribly useful. It either joins two triangles into a quad, or subdivides them. This loses what little there may have been of the sense that the surface is made up of smooth curves. Converting this into splines in Fusion 360 resulted in a rather bumpy surface and poor performance. What is needed is a resampling into quads, which can then be converted into much longer splines that yield lovely smooth curves. The InstantMeshes step achieves exactly that. (And it's free and open source.)
Meshlab does have a tri to quad filter, but it's not terribly useful. It either joins two triangles into a quad, or subdivides them. This loses what little there may have been of the sense that the surface is made up of smooth curves. Converting this into splines in Fusion 360 resulted in a rather bumpy surface and poor performance. What is needed is a resampling into quads, which can then be converted into much longer splines that yield lovely smooth curves. The InstantMeshes step achieves exactly that. (And it's free and open source.)
Hi Docali, I went to your website but like accepting the terms for cookies. Don’t know why every website needs that agreement to see the website.
Sorry, but I didn't come up with that, but in my country, unfortunately, every website needs a reference to cookies, otherwise you are vulnerable to being warned and this can cost a lot of money. Sad but true. As far as I know, my website does not set any cookies, but I have to provide such a banner. Crazy, right?
So I have to excuse for that.
And they will track you anyway, just without cookies. Its just that its been traditionally done with cookies and now cookieless methods are replacing the old ones because the old ones don't provide proper data anymore as only fraction of visitors allow cookies.
Tracking is usually quite harmless by the site owner perspective, often anonymous and genuine data harvesting for making the site better / measure if there is return for advertisement campaigns and perhaps collecting some general info about demographics of visitors if at all possible. Its problematic when someone starts to aggregate the data across multiple sites to make profile of you as a person and sell it for profit, your digital fingerprint and persona. This is what the big tech does, its Facebook main source of income to profile people and sell the knowledge for advertisement as far as I have understood it. Its probably what makes money for Google as well, many others I presume.
Its those companies that make money with marketing or advertisement that would benefit, collect and use the data. Regular small businesses working on other sectors don't care too much, owners of the websites, as long as their business runs selling what ever that is, its not your data or privacy they are selling.
Wait until central bank digital currency rolls in, its no longer a feeling that you are tracked and controlled but actually are and there is no opt-out button
Lets hope alternative future wins, a truly decentralized digital currency.
Tracking is usually quite harmless by the site owner perspective, often anonymous and genuine data harvesting for making the site better / measure if there is return for advertisement campaigns and perhaps collecting some general info about demographics of visitors if at all possible. Its problematic when someone starts to aggregate the data across multiple sites to make profile of you as a person and sell it for profit, your digital fingerprint and persona. This is what the big tech does, its Facebook main source of income to profile people and sell the knowledge for advertisement as far as I have understood it. Its probably what makes money for Google as well, many others I presume.
Its those companies that make money with marketing or advertisement that would benefit, collect and use the data. Regular small businesses working on other sectors don't care too much, owners of the websites, as long as their business runs selling what ever that is, its not your data or privacy they are selling.
Wait until central bank digital currency rolls in, its no longer a feeling that you are tracked and controlled but actually are and there is no opt-out button
Lets hope alternative future wins, a truly decentralized digital currency.
Last edited:
Nice!
You seem to be a horn-aholic.
Did you construct the Multicell yourself?
And did you also measure the axissymmetric expo horn?
Wait until central bank digital currency rolls in, its no longer a feeling that you are tracked and controlled but actually are and there is no opt-out button
The plans for this are already on the shelves, just like more sinister 'track & trace' tools.
Programmable money, unlimited power if centralized, scary stuff.
if $consumer->has_suspicious_forum_comments() & $consumer->has_bough( FORBIDDEN_THINGS )
then $consumer->account()->freeze_all( YEAR_IN_SECONDS )->allow_only ( GOV_GROCERY, GOV_ENERGY )->add_message(" for your own good! ")
unless $consumer->already_punished()
then $consumer->get_friends_and_family_members()->set_status( MONITOR_CLOSELY )
Too much off topic though
if $consumer->has_suspicious_forum_comments() & $consumer->has_bough( FORBIDDEN_THINGS )
then $consumer->account()->freeze_all( YEAR_IN_SECONDS )->allow_only ( GOV_GROCERY, GOV_ENERGY )->add_message(" for your own good! ")
unless $consumer->already_punished()
then $consumer->get_friends_and_family_members()->set_status( MONITOR_CLOSELY )
Too much off topic though
Last edited:
Would it be possible to open a separate thread about cookies madness?
I would suggest that everyone complaining about such things should read about the duties by law every person has who drives a webpage and how many money you have to invest to drive the page and how many time is needed to make it compliant and how many time is needed to get it almost safe against attackers...
I would suggest that everyone complaining about such things should read about the duties by law every person has who drives a webpage and how many money you have to invest to drive the page and how many time is needed to make it compliant and how many time is needed to get it almost safe against attackers...
Yeah I'm printing loads of horns at the moment! I've worked out a great way to extract the shape of e-jmlc horns, so that's on the printer at the moment.
I didn't make the multicells myself, but will probably sell them soon now I can print my own. This is much more fun.
As for measuring, not yet... I'm going to build a few with different topologies, have a good listen and then might measure of I can work up the energy to do so.
Unfortunately, it's precisely the (true) exploiters who try to implement these rules.
After a reunion with this unsmoothed response plot of a Klangfilm Eurodyn horn in combination with a (GIP) Western Electric 594A (mid driver)... I'm into spherical wave horns again
(Red is on-axis, yellow is about 45° off-axis)