3D-printing

Creo Direct free version does have a window/dialog for .stl resolution - doesn't seem to remember your changed settings though and you have to find, reset resolution for every export/save

mechanical interface flatness/acoustic pressure seal is nice but I don't think waveguide surface resolution is in general a acoustic radiation problem:

speed of sound is about a Million times slower than light
very good reflectors like telescope mirrors are finished to ~1/10 wavelength

for sound 1/10 wave in air @ 20 kHz ~ 1/16", 1.5 mm

intentional sound diffuser surfaces are lumpy on human fingers, handspan to arm length size scale


in the phase plug and throat 1/16" may not be to good - but thats more about the relative tolerance, smoothness of the impedance of the narrow passages


for the walls of the horn itself I don't think even 1/16" stepping/roughness would be a practical problem, but diffraction grating effects may occur with very uniform period of the steps as in 3D printer layer thickness or a CNC program path stepping

but very little finishing effort is needed in filling, sanding, solvent "etching", ect. to get below 1/100 wave at 20 kHz

still a coarse grain finish by eye or touch

diffraction grating math for efficiency/power in the modes for fractional wave modulation period, grazing incidence looks non trivial but I suspect it is a non issue at audio
 
Last edited:
I am a bit baffled by it looking fine. You cannot see the creases every 8 degrees or so in the waveguide, or that you consider them unimportant, or that you believed they come from printing in layers, or something else?

An STL file is simply a list of triangles. Triangles have a flat surface. In order to approximate the curved waveguide with triangles the triangles need to be small enough for their flat surface not to show. Using too many triangles on a flat surface only makes the file size bigger than necessary but not using enough triangles on a curved surface will degrade the surface finish.

There will be a tolerance somewhere in whatever software you used to transform your smooth solid model into a bunch of triangles for the STL file. Reducing that tolerance substantially and rerunning the print is likely to be a fairer test of the surface finish of the printer.
I didn't want the file to become too large to fit here so I settled for what you can see. But sure, for a fair test of surface finish from a printer it should be a higher resolution stl, I'll work on it :)

Creo Direct free version does have a window/dialog for .stl resolution - doesn't seem to remember your changed settings though and you have to find, reset resolution for every export/save

mechanical interface flatness/acoustic pressure seal is nice but I don't think waveguide surface resolution is in general a acoustic radiation problem:

speed of sound is about a Million times slower than light
This is true even in the non-free version I'm using (Creo Parametric 2.0). The settings I used was about a factor 3 smaller than the default settings. It's good that we don't need a "smooth" surface as that takes a lot of work, thanks for the info :)

/Anton
 
Last edited:
Measurements of OS waveguide

I've made some fairly crude measurements today (didn't have a lot of time), hopefully they will tell me something :)

This was the setup:
Dn4NPNnl.jpg

A 43x53 cm (3 mm thick) HDF sheet with tweeter mounted offset to reduce diffraction from different edges to combine. There was no seal between tweeter and HDF, shouldn't be necessary at these frequencies.

The raw tweeter was measured with its original faceplate but not countersunk. For the waveguide measurements the faceplate was removed and a larger hole was made in the HDF sheet. I left a 2 mm high edge as I felt it was more important that there was no large hole than a perfect flush mount:
AlqhajOl.jpg

I'm using a UMIK-1 with REW (5.01) and all measurements are gated (8 ms). The distance (1 m) was kept between mic and R2604 membrane. The precision in angle is +-3 degrees, probably the reason why on axis doesn't have the characterstic dip.

With faceplate and baffle on axis:
qJh6rWUl.png


With waveguide and baffle on axis:
CTPUUdLl.png


Compared:
vF2XQa1l.png


Comparison of 10 degrees off axis:
ra5URGSl.png


Comparison of 30 degrees off axis:
yEJQi9Vl.png


Comparison of 45 degrees off axis:
fhiIYpNl.png


All measurements without waveguide:
rcR8hXzl.png


All measurements with waveguide:
bpujiJHl.png


Background noise:
uujHSQ2l.png


And this is how the waveguide looks when mounted to the R2604:
wkA14NZl.jpg


/Anton
 
Better stl file?

I've attached something that resembles what I think I want to print. I'll need to do some thinking though, cause it needs a lot of support-material and time (14 h). The first stl I posted (and printed) was specifically made to use very little support-material.

dSQA37ll.png


/Anton
 

Attachments

  • waveguide_ellip_220x140x85.zip
    253.3 KB · Views: 94
Founder of XSA-Labs
Joined 2012
Paid Member
In Simplify3d, the software that Onni is using to translate the stl file to g code which are the actual printer vectors, there may be some additional processing going on there that makes the print smoother. It depends on the resolution setting. But the stl file is not wysiwyg - it really depends on the quality and accuracy of the post processing software. I look at my typical stl files (generated by SW) and it looks grainy like big triangles but when previewing toolpath, it obviously makes it smoother.
 
In Simplify3d, the software that Onni is using to translate the stl file to g code which are the actual printer vectors, there may be some additional processing going on there that makes the print smoother. It depends on the resolution setting. But the stl file is not wysiwyg - it really depends on the quality and accuracy of the post processing software. I look at my typical stl files (generated by SW) and it looks grainy like big triangles but when previewing toolpath, it obviously makes it smoother.
I understand what you are saying but I don't agree. Simplify3D doesn't seem to do any smoothing (at least not how I have it set up):
OTzFjBbl.png

STL

e0JdhJtl.png

Path

EfEdo2Xl.jpg

Print

The same pattern can be seen in all images.

Now when I'm looking at the print (especially the throat) I agree that the resolution is too low. It's so easy to just set a lower value when exporting.

/Anton
 
The STL file needs to be a lot finer in order for it not to show up in the print. You now appreciate this and so can find the relevant tolerances and change them by ten or whatever. However there is another thing can be done to improve the situation using the same number of triangles which is not to generate the vertices of the triangles in a regular pattern. It is this that is bringing out the edges. There are different methods for tessellating smooth surface and some inject vertices in irregular patterns and some like the one you used inject a regular pattern. Have look to see if you can change the tessellation method. Delaunay would be a good option to see but there are others.
 
Your response will be smoother if you create a spline that joins the surface of the tweeter to the surface of the baffle.

The QSC and JBL waveguides measure very very smooth because of this.

It also improves the polar response.

car-waveguide-61.jpg

Here's an example. Make the spline in two dimensions, and then rotate the spline around the Z axis.


If you want to get really crazy, try cloning the M2 waveguide. I have some posts where I did that; it's an interesting design. The QSC is smoother but the M2 does some interesting things that aren't possible with a conventional waveguide.

car-waveguide-54.jpg

car-waveguide-47.jpg

Here's an M2 type of waveguide I printed

Check out my post history for more pics
 
Last edited:
I'll need to do some thinking though, cause it needs a lot of support-material and time (14 h).

dSQA37ll.png
1. The reason for thee need of support material is the huge quare surface of the front flange. Therefor: Print the part "upside down". You won't be able to print that front radius as perfectly, You'll have to add a little "kink".
2. Simplify the part's design.

I photoshopped both into Your drawing, to illustrate these items. The first item will reduce the need for support material to almost zero, and inevitably leads to beautiful front flange surfaces without the need for manual finishing. But no matter which way You print the part in the end, the second item will reduce print time and material, and make the part stronger.

r
 

Attachments

  • Uantitled-1.jpg
    Uantitled-1.jpg
    74.7 KB · Views: 68
With a 3D printer you can also do some bizarre things that aren't practical with a mold.

For instance, you could add a thousand 2mm cross braces.

Picture006-1.jpg


Taken to the extreme, you might even be able to print a Geddes style foam plug right into the waveguide itself. (Since the foam itself has a shape like a web.)

12457_53.jpg
 
Founder of XSA-Labs
Joined 2012
Paid Member
Taken to the extreme, you might even be able to print a Geddes style foam plug right into the waveguide itself. (Since the foam itself has a shape like a web.)

You could use the "infill" feature and set the fraction and "skip" infill layers etc to get a criss-cross weave of filament to be porous, it won't have a random feature but I don't think that matters.

The cross braces and ability to print concentric phase plugs with expanding channels is very useful as well. I think 3d printing will revolutionize waveguide and CD horn diy concepts. No longer does someone need access to fancy tooling to produce accurate waveguides and phase plugs.

Probably ~15% infill seems ok...
infill-understanding.jpg
 
Last edited:
I'm using a UMIK-1 with REW (5.01) and all measurements are gated (8 ms). The distance (1 m) was kept between mic and R2604 membrane. The precision in angle is +-3 degrees, probably the reason why on axis doesn't have the characterstic dip.
Comparing the flat baffle with the simulation and the Scan-Speak measurements suggests something is going on that needs to be understood. Similarly for the waveguide measurements. As is often, perhaps usually, the case first time something doesn't look quite right.

When performing comparisons lining up the humps and dips to see how they have grown, shrunk, shifted or whatever is important. Can you get the measurements and various simulations onto the same plot?
 
Your response will be smoother if you create a spline that joins the surface of the tweeter to the surface of the baffle.

Check out my post history for more pics
I've looked at most of your stuff, very interesting :) Makes me want to build a synergy... I'm not sure what you mean with the spline part though. This is what the profile looks like from the side:
37qz4prl.png


1. The reason for thee need of support material is the huge quare surface of the front flange. Therefor: Print the part "upside down". You won't be able to print that front radius as perfectly, You'll have to add a little "kink".
2. Simplify the part's design.

I photoshopped both into Your drawing, to illustrate these items. The first item will reduce the need for support material to almost zero, and inevitably leads to beautiful front flange surfaces without the need for manual finishing. But no matter which way You print the part in the end, the second item will reduce print time and material, and make the part stronger.

r
This is actually how I printed the first waveguide (in the first post). No rounding and upside down. But hadn't thought about the possibility to have a roundover and a kink, thanks! There will still be some problems with the print quality when the angle is larger than 45 degrees, but it's worth a shot.

/Anton
 
Last edited:
With a 3D printer you can also do some bizarre things that aren't practical with a mold.

For instance, you could add a thousand 2mm cross braces.

Picture006-1.jpg


Taken to the extreme, you might even be able to print a Geddes style foam plug right into the waveguide itself. (Since the foam itself has a shape like a web.)

12457_53.jpg
This is indeed an interesting thought! Now I just need to figure out how to make Simplify3D print it...

You could use the "infill" feature and set the fraction and "skip" infill layers etc to get a criss-cross weave of filament to be porous, it won't have a random feature but I don't think that matters.

The cross braces and ability to print concentric phase plugs with expanding channels is very useful as well. I think 3d printing will revolutionize waveguide and CD horn diy concepts. No longer does someone need access to fancy tooling to produce accurate waveguides and phase plugs.

Probably ~15% infill seems ok...
infill-understanding.jpg
I think there needs to be 2 geometries (one for waveguide and one for foam) to be able to set different infill for the waveguide (30 % currently) and the foam.

I checked the settings in Simplify3D and found Use random infill placement for each layer. This should be useful! It looks like this:
1 layer:
wNNplJel.png


2 layers:
a6ark9nl.png


4 layers:
xUFC69Ll.png


8 layers:
XAJjP2zl.png


20 layers:
IJhW6Iql.png


It's like a rectilinear foam.

Comparing the flat baffle with the simulation and the Scan-Speak measurements suggests something is going on that needs to be understood. Similarly for the waveguide measurements. As is often, perhaps usually, the case first time something doesn't look quite right.

When performing comparisons lining up the humps and dips to see how they have grown, shrunk, shifted or whatever is important. Can you get the measurements and various simulations onto the same plot?
I haven't simulated the measured waveguide yet, I'll post comparisons as soon as that's done!

/Anton
 
Simplify3D lets you choose directions for the layers, the pictures above was with the default +45 and -45 degrees. Here I've added 0 and 90 degrees as well:

20 layers:
VA33Ch2l.png


One option is of course to make another STL for the foam plug and print that separately and attach it to the waveguide post printing.

/Anton
 
1. The reason for thee need of support material is the huge quare surface of the front flange. Therefor: Print the part "upside down". You won't be able to print that front radius as perfectly, You'll have to add a little "kink".
2. Simplify the part's design.

I photoshopped both into Your drawing, to illustrate these items. The first item will reduce the need for support material to almost zero, and inevitably leads to beautiful front flange surfaces without the need for manual finishing. But no matter which way You print the part in the end, the second item will reduce print time and material, and make the part stronger.

r
I'm not sure how the second part will reduce print time. Complexity is not a problem, volume is. Here is a model that should print decently upside down:

vAZf178l.png

The kink is set to 30 degrees (i.e. an overhang angle of 60 degrees). This will require slow printing (of outmost layer) for good result.

vFOJDzxl.png

I added chamfers to minimize bridging but without adding a lot of volume.

Simplify3D calculates the print time to 5 h and material usage of 125 g (about 4 USD). This is with 0 support material and a 210x130x85 mm waveguide for the R2604.

/Anton
 
Founder of XSA-Labs
Joined 2012
Paid Member
P Bateman was referring to a spline as the primitive entity in 3d models that completely describes a curve. All curves and curved surfaces, if they can be approximated as splines, make the compute time faster. I don't think it is an issue for 3d printing to have to use splines. I was able to approximate a tractrix curve by overlaying a bitmap image of a tractrix on the face of a solid in SolidWorks, then by overlaying a single spline segment, use the 3 point adjustments of inflection point, and end point angle and steepness, I modeled the tractrix curve. Normally, an equation would need to be used. If more complicated shapes, use a series of splines. I got the inspiration from this video of how to make a Batman emblem with a few splines.

https://www.youtube.com/watch?v=XIIcbUI0tyw

You can certainly do this for an OS or SEOS profile.
 
Founder of XSA-Labs
Joined 2012
Paid Member
I'm not sure how the second part will reduce print time. Complexity is not a problem, volume is. Here is a model that should print decently upside down:

vAZf178l.png

The kink is set to 30 degrees (i.e. an overhang angle of 60 degrees). This will require slow printing (of outmost layer) for good result.

vFOJDzxl.png

I added chamfers to minimize bridging but without adding a lot of volume.

Simplify3D calculates the print time to 5 h and material usage of 125 g (about 4 USD). This is with 0 support material and a 210x130x85 mm waveguide for the R2604.

/Anton

Would you mind making something like this but with a 1in dia throat and a flat mounting boss plate for a 3/2 hole bolt pattern compression driver? That would be a real useful .stl upload as many of us could immediately go to press and have a nifty SEOS horn for $5!
Thanks in advance! :)

Oh, and have the option of an infill with random angles to get the HOM busting reticulated foam. I believe this printed "foam" may actually be a novel new invention that is clearly different than conventional reticulated foam.

I like what you did here:

VA33Ch2l.png


This is golden - a one step foam loaded printable OS waveguide! :cheers:
 
Last edited: