There is an explanation of the ray-tracing algorithm used for baffle analysis in PSD-Lite. PSD-Lite lets you model a baffle that has up to 20 corners, and you can move the corners around to create some exotic shapes. Once the baffle response is calculated, you can add this response to the overall response for designing the crossover. To view the ray-tracing explanation, select the "Design Baffle" module and then select "Help" and then "Diffraction Model" from the pull-down menu.
The ray-tracing algorithm for baffle diffraction is used in PSD-Lite, Edge, and the Baffle Diffraction and Boundary Simulator. This approach provides a fairly accurate approximation of the response at a given summation point. It is also possible to calculate the response "directly" by solving the Helmholtz equation in 3D. Jeff Candy has recently published an article in the AES Journal describing that approach, but it's not for casual reading.
PSD-Lite can be downloaded here: http://www.audiodevelopers.com/Software/PSD_Lite/setup.exe
The ray-tracing algorithm for baffle diffraction is used in PSD-Lite, Edge, and the Baffle Diffraction and Boundary Simulator. This approach provides a fairly accurate approximation of the response at a given summation point. It is also possible to calculate the response "directly" by solving the Helmholtz equation in 3D. Jeff Candy has recently published an article in the AES Journal describing that approach, but it's not for casual reading.
PSD-Lite can be downloaded here: http://www.audiodevelopers.com/Software/PSD_Lite/setup.exe
Very nice program Neil. I hadn't tried it yet because I thought it was just a different version of PCD which I was (previously) happy with but now I see that that you've included so much more. Very nice.
But to the point here, unless I am mistaken, I still cannot model a sphere in your baffle diffraction program. I was hoping that it might be able to model a round baffle (like the Edge) the size of a driver but then also add in a very large roundover (like 3 or 4" - unlike the Edge) that would essentially turn it into the front half of a sphere. But I can't get a baffle shape other than a 4 sided box. (Not complaining of course - having worked a tiny bit with some Excel programming, I well know how much work you must have put into it. Thank you most kindly.)
But to the point here, unless I am mistaken, I still cannot model a sphere in your baffle diffraction program. I was hoping that it might be able to model a round baffle (like the Edge) the size of a driver but then also add in a very large roundover (like 3 or 4" - unlike the Edge) that would essentially turn it into the front half of a sphere. But I can't get a baffle shape other than a 4 sided box. (Not complaining of course - having worked a tiny bit with some Excel programming, I well know how much work you must have put into it. Thank you most kindly.)
Sorry, no spheres. It will do rounded edges, but there aren't enough samples along the rounded edges to accurately model a sphere. It's possible to modify the code to model the baffle response of a sphere, but it would take some significant effort, and it's not on the current "to-do" list.
You can approximate a round baffle by adding more corners--just right-click on a line and select "add corner". A 20-sided figure will be pretty close to a circular baffle, but the GUI doesn't make it easy to get the corners lined up. Also, the program currently doesn't implement a very accurate roundover model. That roundover feature was done before I came up with the algorithm to calculate the internal path based on wall thickness. By using the "exact" interior path I will be able to implement a better roundover algorithm. That improved roundover model is on the "to-do" list.
You can approximate a round baffle by adding more corners--just right-click on a line and select "add corner". A 20-sided figure will be pretty close to a circular baffle, but the GUI doesn't make it easy to get the corners lined up. Also, the program currently doesn't implement a very accurate roundover model. That roundover feature was done before I came up with the algorithm to calculate the internal path based on wall thickness. By using the "exact" interior path I will be able to implement a better roundover algorithm. That improved roundover model is on the "to-do" list.
Last edited:
Neil, hey I just think that you guys who are doing this kind of work are fantastic. No complaints from me.
Thanks for the corners tip. I would have got to that eventually. In fact I think the rest of my day is shot now cause I'm about to open it up again and start trying it out.
The way that you've combined so many different things into 1 program is very sweet.
Cheers
Thanks for the corners tip. I would have got to that eventually. In fact I think the rest of my day is shot now cause I'm about to open it up again and start trying it out.
The way that you've combined so many different things into 1 program is very sweet.Cheers
So, very interesting Neil. By using all 20 corner points, it's fairly easy to model a round baffle and then to add a large radius roundover to approximate what would appear to be the front half of a sphere with your program.
First graph below is the diffraction result with a 16" diameter round baffle with an 8" roundover with a 7" driver in the middle. Just for comparison purposes for the OP, the 2nd graph is the same 7" driver on a 16" x 22" baffle with the driver centered 8" from the bottom. Results look like what you'd expect according to Olson's research, although there is just a little extra peaking in the sphere model here.
Just curious whether you might have any idea as to the degree of accuracy of what I've done? I was a little surprised at the fairly large difference in the baffle step transition point between the 2 graphs but it would seem to agree with what sreten has been saying and to also jive well with lucadelcarlo's subjective experience with his 2 trial conditions.
lucadelcarlo, if you haven't downloaded this program for yourself yet (which I think you should), if you tell me the dimensions of what you are working with I can do a quick simulation for you and post the results.
First graph below is the diffraction result with a 16" diameter round baffle with an 8" roundover with a 7" driver in the middle. Just for comparison purposes for the OP, the 2nd graph is the same 7" driver on a 16" x 22" baffle with the driver centered 8" from the bottom. Results look like what you'd expect according to Olson's research, although there is just a little extra peaking in the sphere model here.
Just curious whether you might have any idea as to the degree of accuracy of what I've done? I was a little surprised at the fairly large difference in the baffle step transition point between the 2 graphs but it would seem to agree with what sreten has been saying and to also jive well with lucadelcarlo's subjective experience with his 2 trial conditions.
lucadelcarlo, if you haven't downloaded this program for yourself yet (which I think you should), if you tell me the dimensions of what you are working with I can do a quick simulation for you and post the results.
Attachments
I could try making it more accurate but it would take some time. Right now the roundover algorithm only uses 4 points. That is, it divides the roundover radius into 4 reflection points and then applies the weights to each point, as Paul Verdone describes in the user manual for his BDS program: http://www.audiodevelopers.com/misc/BDS_Manual.pdf. Paul actually generated weights for 8 points, and I could try using those values to see if the results better approximate the expected results for a sphere.
Those graphs do indeed reflect what I seem to be hearing, consistent with what others have said and of course Olson. For now, thank you for this great offer to do a simulation, Jreave, and Neil for the original programming work you're now commenting on.
Dimensions:
Sphere: 17.7 cm
Box: H 17.5 cm x W 15.5 cm x D 21.5 cm
12 cm driver
Note: the driver is top mounted and with a 4mm thick basket frame which itself sits atop a 4mm mounting ring, so in total, the front of the speaker flange is 8mm away from the sphere's surface. Does the program offer a possibility to take that into account, and were the graphs supplied earlier depicting a mounting style? I'm very keen to examine what effect this may be having on diffraction. Physically remodeling the spheres is a difficult process, so some degree of CAD-based modeling predicting response as guide would indeed be an excellent tool.
Dimensions:
Sphere: 17.7 cm
Box: H 17.5 cm x W 15.5 cm x D 21.5 cm
12 cm driver
Note: the driver is top mounted and with a 4mm thick basket frame which itself sits atop a 4mm mounting ring, so in total, the front of the speaker flange is 8mm away from the sphere's surface. Does the program offer a possibility to take that into account, and were the graphs supplied earlier depicting a mounting style? I'm very keen to examine what effect this may be having on diffraction. Physically remodeling the spheres is a difficult process, so some degree of CAD-based modeling predicting response as guide would indeed be an excellent tool.
Last edited:
The program has a built-in drawing tool to outline the baffle shape, but that tool is way too simple to address mounting rings, re-radiation from driver openings, absorption from cone materials, reflections from interior walls through the cone and other such implementation details that can affect diffraction. In order to model mounting rings and other details like that, I would have to use a 3D CAD model that provides much more detail than my simple drawing tool, and the model would need to be considerably more complex. I actually considered using an embedded 3D drawing capability and importing SketchUp files, but decided I already had plenty to do
. For complex geometries and modeling absorption, you are probably better off just building a physical model and taking measurements.Also, the model currently only provides a single user-specified summation point, and baffle diffraction can vary quite a bit depending on where the measurements are taken. So at this point I would rather spend time creating a 3D picture of the baffle (and crossover) effects than adding more fidelity or additional features to the current diffraction model.
Thanks Neil for your feedback.
So lucadelcarlo, especially considering that the program cannot account for your mounting method (which is going to add some extra rippling to the diffraction pattern) and that the program may not be totally accurate with the large roundover size that I have to use, I wouldn't take the simulation for your sphere beyond a ballpark estimation. Here are the results. The sphere is the 1st one.
The 0dB points end up quite different - it looks like about 2000Hz for the sphere and about 700Hz in the box. No wonder you are hearing such a difference. It would be interesting to see how these graphs compare if you actually get around to measuring them.
So lucadelcarlo, especially considering that the program cannot account for your mounting method (which is going to add some extra rippling to the diffraction pattern) and that the program may not be totally accurate with the large roundover size that I have to use, I wouldn't take the simulation for your sphere beyond a ballpark estimation. Here are the results. The sphere is the 1st one.
The 0dB points end up quite different - it looks like about 2000Hz for the sphere and about 700Hz in the box. No wonder you are hearing such a difference. It would be interesting to see how these graphs compare if you actually get around to measuring them.
Attachments
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.