Hi PierreQuiRoule,
I have the false impression that I understand the calculation described in the first two paragraphs. I also understand, with reference to your post # 168, the elliptic profile described in the above-cited paragraph. But, how did you determine the shape, height, and length? Was it by amnual adjustments or did you figure out a mathematical description?
Kindest regards,
M
I have the false impression that I understand the calculation described in the first two paragraphs. I also understand, with reference to your post # 168, the elliptic profile described in the above-cited paragraph. But, how did you determine the shape, height, and length? Was it by amnual adjustments or did you figure out a mathematical description?
Kindest regards,
M
Interesting. Yes, I would be happy to assist. All I have to do is change a few parameters to re-generate the throat profile and have the same company that printed mine print yours. Let me know!
Hi M
The elliptic profile is for the roof/floor height, not the side walls shown in post #168. The automatically generated ellipse specifies the curve of the adapter roof and floor when it is looked at sideways.
In post #168 the walls were obtained by subtracting a "donut-like shape" from the side wall. The donut's parameters were mathematically derived, not manually fitted. The choice of a "donut" was arbitrary, and I just wanted to convince myself the concept was feasible. I like that it is an easy to visualize shape and it might perform well.
I have since removed the "donut shape" constraint. So Prototype #2 is not a print of post #168. Instead the new throat has side walls that are mathematically derived across their whole surface. They are wavy and reminiscent of JBL's progressive transition shapes. I find this interesting as I did not intend this, but this is where logic and math took the profile.
Pierre
PS: Thanks to the help of many fellow DIYers, I recently came to realize something that should have been obvious from the onset. A throat adapter is nothing but a horn: a horn that embeds into another, larger horn. This is good inspiration for the next leg of this experiment.
Hi PierreQuiRoule,
thank you for the reply.
I used a 3D drafting to morph circular area (S1) to a rectangular area (S2) over a length (L), but unless I made some mistake, the expansion coefficient calculated back from thusly obtained shape does not correspond to that when S1 ans S2 are circular. Thus, it is my understanding that the need for the wall modification stems form the need to transform the circular throat to the rectangular horn entry.
At this point I do not have any idea how to automatically generate the necessary correction as you have done.
Kindest regards,
M
thank you for the reply.
I used a 3D drafting to morph circular area (S1) to a rectangular area (S2) over a length (L), but unless I made some mistake, the expansion coefficient calculated back from thusly obtained shape does not correspond to that when S1 ans S2 are circular. Thus, it is my understanding that the need for the wall modification stems form the need to transform the circular throat to the rectangular horn entry.
At this point I do not have any idea how to automatically generate the necessary correction as you have done.
Kindest regards,
M
Many different things at play here when converting circular to rectangular.
1. getting the correct surface area profile for the objective hypex T and Fc.
One can achieve this easily by adjusting the width between the walls all along the profile to get there. The walls can be vertical (no convexity). Works like glue. IMO this is already better than an off-the-shelf linear adapter.
2. shaping the spherical wavefront to a cylindrical wavefront.
This is where I think the walls should not be perfectly vertical, because speed of sound being constant, the synchronized wave edge on the circular entrance would then arrive at different times on the vertical exit wall (and thus not produce a cylindrical wave front at the exit). This is more tedious to solve and involves tradeoffs.
1. getting the correct surface area profile for the objective hypex T and Fc.
One can achieve this easily by adjusting the width between the walls all along the profile to get there. The walls can be vertical (no convexity). Works like glue. IMO this is already better than an off-the-shelf linear adapter.
2. shaping the spherical wavefront to a cylindrical wavefront.
This is where I think the walls should not be perfectly vertical, because speed of sound being constant, the synchronized wave edge on the circular entrance would then arrive at different times on the vertical exit wall (and thus not produce a cylindrical wave front at the exit). This is more tedious to solve and involves tradeoffs.
Hi PierreQuiRoule,
sorry to bother you again, but after thinking about it overnight, I believe that I understand 1. in your post. The width may be adjusted in a similar fashion as you did in post # 168, or JBL does with some of their wave-guides that have non-flat walls, e.g., JBL 5006815 Horn Lens or Klipsh "mumps". I still do not know how to do it rigorously, i.e., mathematically.
But, I cannot picture point 2. The entrance to the horn is rectangular, which implies parallel walls. Could you please provide additional explanation?
Kindest regards,
M
sorry to bother you again, but after thinking about it overnight, I believe that I understand 1. in your post. The width may be adjusted in a similar fashion as you did in post # 168, or JBL does with some of their wave-guides that have non-flat walls, e.g., JBL 5006815 Horn Lens or Klipsh "mumps". I still do not know how to do it rigorously, i.e., mathematically.
But, I cannot picture point 2. The entrance to the horn is rectangular, which implies parallel walls. Could you please provide additional explanation?
Kindest regards,
M
It's all there, although granted it can be hard to visualize in 3D. I also stop and think and stare in blank space trying to picture in my head what is going on.
Point 1 is met with profiles of posts #90 and #127 (Prototype #1). A cross section at distance x would show straight vertical walls. The exit wavefront is spherical-ish...
Point 2 is not met by those two profiles. One solution that satisfies #2 is walls that are straight vertical only at the adapter exit distance x, as shown in #168.
Me, after a couple good nights sleep I didn't see it anymore, then I did. It's confusing I know.
Point 1 is met with profiles of posts #90 and #127 (Prototype #1). A cross section at distance x would show straight vertical walls. The exit wavefront is spherical-ish...
Point 2 is not met by those two profiles. One solution that satisfies #2 is walls that are straight vertical only at the adapter exit distance x, as shown in #168.
Me, after a couple good nights sleep I didn't see it anymore, then I did. It's confusing I know.
Joseph Crowe has a very similar design;
https://josephcrowe.com/products/custom-yuichi-a-290-biradial-cad-model-make-any-size
Thank you for sharing this (I somehow had missed those drawings). It is hard to see from Troy's drawings if in a plane (constant x) view the sidewalls bulge inside like what I proposed above. They do when viewed from the top, which is very common place. However they don't seem to do when looking on axis, but this is hard to see. In any case, Troy is a true inspiration and paying due attention to throat adapters which is a good thing. And if I reinvented someone else's idea, which is very possible, then I am happy too.
A quick update... Prototype #2 (spherical to cylindrical) printed well, and is waiting for my return home to test and measure. A sheet view of a quarter of the throat is shown below: entrance at Botton, exit at the Top, side wall to the right, green is inside, and yellow is outside. One can see the exterior of the side wall (yellow) bulging deeper inside the throat in comparison to the arc following the corner of the rectangular exit. In this instance, the software computed 35 side wall arcs from entrance to exit, and equalized their lengths so they are all the same.
Now, Marco's posts in Pano's thread have led me to consider prioritizing flare rate profile in the next adaptor prototype (#3), which is still on the drawing board.
A quick update... Prototype #2 (spherical to cylindrical) printed well, and is waiting for my return home to test and measure. A sheet view of a quarter of the throat is shown below: entrance at Botton, exit at the Top, side wall to the right, green is inside, and yellow is outside. One can see the exterior of the side wall (yellow) bulging deeper inside the throat in comparison to the arc following the corner of the rectangular exit. In this instance, the software computed 35 side wall arcs from entrance to exit, and equalized their lengths so they are all the same.
Now, Marco's posts in Pano's thread have led me to consider prioritizing flare rate profile in the next adaptor prototype (#3), which is still on the drawing board.
That design has a much more significant narrowing in the horizontal dimension. This can certainly have a positive effect as the smaller dimension pushes up the frequency where the driver naturally narrows based on it's size. Even with smooth curves this is a version of a diffraction slot.
Be cautious of what things look to you versus how the wavefront sees it. Abrupt changes in curvature are problematic, but things are different in the case of a fin horn as the wavefront is split into the different channels, the wavefront shape at the entrance to the fins has a big effect on whether an abrupt change is seen.
You have to be careful not to introduce a compression or decrease of the surface area otherwise you get a bump in the radiation impedance. I do not how the surface propagation is handled here. Maybe one could use zero flare rate for the first section when the round is transforming to the narrow region. But generally by doing this like in this picture you violate the underlying flare rate of the horn and this makes the approach questionable.
The proper expansion can't represent the right radiation impedance if the wavefront integrity does not support it, sometimes that needs to be fixed. In other cases the wavefront may be good from the source, but needs to change to suit the horn in question.
In any case, the change in radiation impedance can be equalised. If the change in expansion has rectified the wavefront, the result will be good.
Hmm ... would it be possible to build two adapters, one satisfying Condition 1 but not Condition 2, and the second with some convexity and/or ripples that satisfy both conditions?
I only mention this because we might be at the outer boundary of what modeling can do, and a real-world set of measurements and/or direct auditioning might be useful. Sure, equalization would compensate for throat-pinch effects, but what of increases in IM distortion that would not appear is simulations (which assume linear drivers with perfect phase plugs and diaphragms).
@Lynn Olson Something like what you describe, yes. I am still working on satisfying both conditions and am getting close. Prototype #2 (Post #191) satisfies the second condition, and comes close to satisfying the first. Its flare rate monotonically increases from 6.9 to 12.7 (entrance to exit). Not bad, but the 745NEO flare rate is somewhere between 8 and 10, and the throat flare rate of the Athos TH4001 clone might be around 8, but in reality I don't think it has been precisely measured yet. So I intend to DIY a calliper to measure width/thickness of the fins and use a laser to measure tangential angle at the throat. As a side note, I have just changed my model again, and it can now produce any flare rate expansion while satisfying the second condition - as long as the entrance and exit dimensions allow it. Pending measurements of the horn throat, it could for example generate a constant m=9 rate expansion and feature the convexity/ripple for cylindrical wavefront shaping.
Measurements will come. In particular, I intend on measuring horn mouth response on a diagonal (corner from corner) as this is where I expect to measure some difference as a result of satisfying the second condition. All this takes time
Measurements will come. In particular, I intend on measuring horn mouth response on a diagonal (corner from corner) as this is where I expect to measure some difference as a result of satisfying the second condition. All this takes time
Last edited: