- Thread Starter
- #341
These types of models where you enter in driver size and mounting and baffle dimensions model two things, diffraction and driver radiation off axis (dispersion). The error you mentioned is mostly due to the model of driver dispersion, not the diffraction. It assumes the driver radiation is the solution to the wave equation for a rigid pistonic disc. A real speaker cone enters into break up and emits like a smaller cone at higher frequencies, broadening radiation there. There are no physical parameters specified for speaker drivers that lets anyone model this break up except the driver's designers that have access to the FEM model.
To have a hope of an off axis simulation useful for DIY design, the actual driver radiation on an IEC baffle needs to be used (available from the manufacturer or retailer), the IEC baffle's diffraction "backed out" using the tool's diffraction modeler, and then the diffraction of it mounted in the target box modeled and added back in. This has to be repeated at every observation/measurement location. I wrote this up here: http://audio.claub.net/software.html -> http://audio.claub.net/software/DaveDalFarra/Simple Loudspeaker Design ver2.pdf I wrote this up quickly for a DIY meet and Charlie and Jeff then went one better and later created their splice tool.
What I'm curious about is the second part, how well Vituix Cad models diffraction. It doesn't need to get this "wrong" by much to create audibly misleading results because the error will exist over a wide frequency range, so even small dB errors will be noticeable. Here's how the other tools fared compared with anechoic measurements. The Edge vs Paul's BDS vs Jeff Bagby's DBS (based off Paul's BDS so no surprise those two are close):
View attachment 68871
That's really great data there. Ultimately, with this speaker at least, the diffraction simulation is just a way to get an idea of what diffraction will look like. Naturally, full polar measurements at 10 degrees will be made, which will show diffraction effects pretty clearly.
The reason I'm going to the trouble of making a nice, accurate, tall platform is that I realize that quality of measurement is now the 'hard part' of speaker design. If you have really good data you can design a crossover with all the graphs in the cea 2034 right under your nose.