3D printed waveguides

[Wolf]

Not sure, but there is a thread somewhere.
Wolf

[TN Allen]

Pete Schumacher sent me a cross section of one of their early profiles, perhaps from the MiniByzi(?). I may have the name wrong. As I recall Dan P. described the shape as a modified torus. I checked the cross section and from what I could find for an approximate 8" diameter the cross section radius was close to 3.95", which was also quite close to the cross section radius on a wave guide used by Zaph? My recollection of names is not so good, but no doubt someone will correct these. I can find the thread and cross sections if anyone is interested in these. In my experience, ani_101 is correct. There is nothing I've found either regarding waveguide design. There may be some design criteria somewhere, but much of what is available is very academic, and heavy on math, and not particularly practical. I would like to find I am wrong about this. The Elliot guide did not test well as I recall. Ani_101 is also correct indicating that x and y coordinates can easily be imported into CAD programs to generate NURBSplines. This requires an Excel file formatted and exported for a given CAD program. From these a person can then produce surface and solid models and subsequently tool paths and or files for 3D printing. There are threads on this forum where data and cross sections are discussed. Ryan Bouma, rpb and others provided quite a few test results and insightful discussion regarding wg contour and efficacy. In a way, the cumulative information in the threads probably is a good place to start. I think there are also other forums on which extensive wg discussions have been recorded. In the past I have offered to work with anyone interested in wave guides, and would do so at present, within reason. I can provide the CAD files and mill guides on the Haas where I teach. There are 3D printers available, although I prefer the Haas and stacked MDF for guides. If and when it is finally completely assembled and running, I'll also have access to a 4'X8' CNC router. The offer is for research, rather than commercial projects. If anyone wants to discuss this, send me a phone number in a PM.

[DDG]

You can 3d print at UPS stores dirt cheap.

https://www.theupsstore.com/print/3d-printing

[DanP]

I've got access to a real 3d printer (EOS P760) and would be willing to produce some test units for you, Brandon. But I will say that there only seem to be a few factors that influence the performance of a waveguide as some of the previous testing has shown (at least those designed to load a dome tweeter): diameter, depth, initial throat angle, and throat transition geometry. The profile of the flare should have as close as possible to a constant radius which combined with the 4 factors just mentioned, basically locks in the whole WG design by picking one value.

There is some wiggle room, but you basically pick the diameter and depth you want for directivity, boost, and AC alignment, and that determines the potential performance of your WG. In order to realize this performance, however, it needs to mate to the tweeter properly. This is why I always thought the better direction would be to create a few generic WGs (small, medium, and large, in shallow and deep) that could accommodate a tweeter adapter. That would make it reasonable for a company like PE to have them produced and sell them. 6 WGs and probably 20 adapters could probably cover 90% of the available tweeters.

If that's not the direction you want to go in, I'm down for whatever. I've got limited CAD skills/software, but as was mentioned before, I designed the WG that Pete and Ryan used in the Vapor Audio Aurora. PM me what you're wanting to do and we'll see where it goes. I really would like to see waveguides turn into a reasonable option for a new DIYer to take advantage of.

Dan

[Adam_G]

I was going to have this one printed. It is a Tractrix horn design based on a calculator I found online which used excel. My local place wanted too much money to print it for me to go ahead with the plan unfortunately. I believe I designed it to have proper loading down to 750hz for a smoother low-end roll off when crossing low. I've got different CAD files available if anyone wants them.

[johnnyrichards]

I'm in - if there is anything I can do to help, I have a few hours free-time a month.

[augerpro]

Thanks for chiming in Dan. I know you said PM, but I'll just post here for now to get everyone's input, I don't see this as "my" project, but a community project who's fruit any DIY'er in the future can take advantage of.

I like the adapter, but since we are not commercializing this and don't have to worry about inventory costs, wouldn't it be just as easy personalize the throat for each wg since we're not making them, just making files for the end user to print? Although I can see the advantage while we're prototyping, would definitely cut down on costs and time to have many adapters for one set of wg's. That's a bridge we can probably cross when we come to it?

Could you expound a little more on the different factors you listed above? Also, what throat angle did you find to work well? It seems like it would be quite high, just as a guess my first instinct would be to make the angle normal to the tangent of the dome at the point where it would hypothetical intersect. If we assume the dome radiates spherically, which I believe is the general assumption. Did you have a spreadsheet to calculate these dimensions?

[augerpro]

So I can see going two paths at the same time: one using the waveguide math and spreadsheets developed by others, and a more iterative process that will probably start with aping known designs like Revel, JBL, Genelec, Mackie, etc.

I've attached some of the spreadsheets and Geddes' white papers, including one doc that is a compilation by someone of relevant quotes from the big Geddes thread in diyaudio: https://www.dropbox.com/s/q9wcxxqdd1...uides.rar?dl=0 Rereading the material, I realized this whole thing is a lot simpler if we have a spherical source. Then a simple conical is best (ignoring the mouth to baffle interface for the moment). Then the only issue worth combing the data for are the mouth roundover. And with the desire to use a constant large radius roundover I see how this devolves into really setting a throat angle, and picking either depth or mouth diameter, the one setting the other, and ending up with a constant flare waveguide. OR, you ignore the mouth roundover and just stick with the simple conical. This appears to be the route Revel is going down now with their newest designs: http://www.revelspeakers.com/product...oduct/m16.html

[randyohoh]

Autodesk offers their Fusion 3D design software for a one year free trial. The download is fast but the software is slow to start on my 2.8 ghz Pentium I7 with 8 gb of DDR3 ram. Supposedly tool paths can be generated right in the Fusion CAD software which could sent to a cnc router.

[ani_101]

As far as I know, fusion 3d is free for hobbyist. You do need to sign in and it periodically updates, but it is free for non commercial use. After one year you can reselect the startup option.

From the Adobe site:
After you download your free trial, click on the trial countdown number (X days) next to your user name on the upper right navigation. You’ll then see an option to register for a free 1-year startup license.
A free 1-year startup license is available for hobbyists, enthusiasts, makers, and emerging businesses that make less than US$100,000 in revenue per year. At the end of 1 year, you can reselect the startup entitlement or transition to a commercial entitlement.

[andy19191]

If you haven't seen it there is some information on typical issues in this thread.

I would echo Dan P that for typical home audio waveguides what requires detailed attention is the throat geometry to get a good high frequency response. The optimum shape is almost certainly not an initial throat angle but one that takes account of the shape of the surround and how it moves. A phase plug of some sort it also likely to be beneficial. The typical Geddes concerns are more relevant for deep waveguides/horns and might be a bit misleading.

Most CAD programs are poor at creating mathematically specified curved surfaces with many simply not being able to do it. However, they usually have no problems performing the full set of CAD operations on the surface if it is first imported as a NURBS surface patch. It might be wise to look to use different software for the two functions. In the past, including Dan and Pete's exercise at one point, I have written a script to generate the basic waveguide surface and could offer some guidance or help along these lines if required.

[augerpro]

If you haven't seen it there is some information on typical issues in this thread.

Reading through it now. Comsol seems like the way to go to simulate horn behavior.

[andy19191]

Reading through it now. Comsol seems like the way to go to simulate horn behavior.

COMSOL is a good way to develop a geometry but it is very expensive and so likely to be available only to those that can access it at work. It also takes a fair bit of time to learn how to use it effectively as can be seen in the thread. There are cheaper codes and even some freely available ones but they tend to be more difficult to use, more unreliable, more inefficient in terms of computer resources, may require writing glue code, etc...

Does anyone here have access to a copy of COMSOL with the Acoustic Module installed? Or an equivalent?

[DanP]

Thanks for chiming in Dan. I know you said PM, but I'll just post here for now to get everyone's input, I don't see this as "my" project, but a community project who's fruit any DIY'er in the future can take advantage of.

I like the adapter, but since we are not commercializing this and don't have to worry about inventory costs, wouldn't it be just as easy personalize the throat for each wg since we're not making them, just making files for the end user to print? Although I can see the advantage while we're prototyping, would definitely cut down on costs and time to have many adapters for one set of wg's. That's a bridge we can probably cross when we come to it?

Could you expound a little more on the different factors you listed above? Also, what throat angle did you find to work well? It seems like it would be quite high, just as a guess my first instinct would be to make the angle normal to the tangent of the dome at the point where it would hypothetical intersect. If we assume the dome radiates spherically, which I believe is the general assumption. Did you have a spreadsheet to calculate these dimensions?

To expound on the factors, I think it's important to note that the profile shape has very little to do with its performance other than being the most consistent transition possible from the initial throat angle to the flat plane of the speaker baffle. Pete and I tested this with 3 different profiles in 2 sizes each: 1) straight flare profile with small roundover to baffle, 2) straight flare profile with large roundover to baffle, and 3) variable radius from throat (beginning at 90 degrees). After seeing the smoothness of the radiused vs straight profiles, we went a step further at the advice of Phil Bamburg and made that radius consistent to get a more shallow throat angle, to reduce the possibility of throat reflections. The response from this WG was smooth as silk and it seemed to make sense as it is the least restrictive and least disruptive in terms of wave propagation. It's dimensions were: 8.5" dia, 2" deep, 1.625" throat diameter, 57 degree throat angle, 3.95" constant radius profile.

So now, a larger diameter WG will boost lower frequencies and a deeper WG will boost it more. Both diameter and depth affect directivity proprtionally. So you choose your diameter and depth based on your design goals and your throat diameter based on your tweeter, then your constant radius and throat angle are almost predetermined - a larger radius will begin to approximate a straight line, so you just make the radius as small as you can to span the gap between mouth and throat. So there is no magic here, and in my opinion, no need for further testing. The magic is in the transition area between the dome and the WG throat. This is where your very predictable WG can turn very ugly and where I don't think we've come up with any novel, repeatable solutions for mating tweeters and WGs. Dave Pellegrene did some really good work and I believe he tried to get the WG throat as close as possible to the actual dome and carved out working space for the surround and faceplate geometry where applicable. He achieved very good results like this, but I think there are dangers of creating resonant cavites. In contrast, the WG I designed sat right on the RS28's flat faceplate and continued the transition already in place between the faceplate and the grill/surround area of the tweeter. This also provided great results, but I think there are dangers of creating reflective ridges.

I'm estimating 6 WG sizes and 20 tweeter mounting geometries would give us what we needed to make waveguide loading a possibility on 90% of available dome tweeters. The above factors, combined with the amount of work required to generate 120 individual CAD models and the amount of prototype support for several iterations (we'll need to test, tweak, and repeat before they are approved, of course) of 120 full sized waveguides seems impractical to me. However, generating 26 models and providing prototype support for 6 full size WGs and several iterations of 20 small tweeter adapters seems reasonable. I also think some anti-reflection geometries in the throat area could serve useful as well - and that is completely possible with 3D printing.

So as someone who has had access to this for years, here's where I'm at with it: I know if I follow the above rules, I'll get a solid performing waveguide, so I only ever need to figure out the transition area if I want to use a specific tweeter. This is where my personal interest lies. However, if you or anyone else has a different aspect that they'd like to explore, I'm down to help with CAD and prototypes.

Dan

[TN Allen]

Round guides with a cross section defined by a radius, are more efficiently turned on a lathe using a simple jig to control the cutter movement along an arc. The surface finish can easily be much smoother than that produced by a 3D printer, and requires considerably less time sanding than the milled surface from a CNC machine, even with a small step down or across.

Producing an appropriate fit around the dome, or a ring radiator, usually requires removing the face plate because the throat transition on many face plates is not consistent with a throat area contour based upon a radius cross section. Removing the face plate can also result in further mounting problems, and in the case of the RS28, or Usher dome tweeter, can allow the diaphragm to slip slightly out of alignment. An easy way around the problems that can result in removing face plates, is to use tweeters that do not have face plates.