It would be good to share this on the diyaudio Dml as a full range forum. People with more knowledge than me could have some helpful input. I have been making panel speakers and have enjoyed the results with great sounding ones for home and this would make it so much easier than the trial and error many people go through.
Congratulations! I have been very interested in this topic for several years. It would be interesting to read the documents you have published on the subject. In the meantime, thank you, and Happy New Year!
Thanks! There are way too many publications to list, but my Google scholar account lists them all: scholar.google.com/citations?user=Dpw7u9IAAAAJ&hl=en&authuser=1
This is a really cool tool. I just binged all three videos. I can't say how excited I am too see something like this. Here are some thoughts about possible improvements: 1. Damping. Both within the panel itself and at the boundaries. An undamped (say FFFF) aluminum (or glass!) panel is truly horrible in real life, while the same in acrylic is much, much better. But in your video the aluminum didn't look nearly as bad as it should, nor as different as it really is from acrylic. Also, using damping materials at the supports changes the response dramatically, having some capability in the model for varying that seems critical. 2. Material Properties. Do you envision the ability to input self definable materials? Especially orthotropic? Composite panels with say, carbon fiber skins and lightweight cores are commonly used. But their properties vary with the specific design, so the ability to input the elastic properties directly (rather than choose a generic material type) would be critical. 3. Any chance of modelling the impulse response? Or a wavelet spectrogram? Frequency response is important, but the response in the time domain is equally important. 4. What about sensitivity/efficiency? An big challenge in flat panel design is getting good sensitivity. Acrylic has great internal damping but horrible sensitivity, due to is relatively low stiffness and high density, while skin/core/skin composites are much better in that regard. It would be great if you model could predict some measure of sensitivity. 5. Less importantly (to me anyway), but still nice, would be the ability to model shapes other than rectangular, including rounded edges. And to model mixed boundary conditions (FSFS) for example. Great work, I can't wait to see what's next!
Thanks for the comment - these are all really great points. I'll try to respond in order... 1. Yes, damping is (inaccurately) set to the same value for all materials right now. You're right that acrylic has super high internal damping and aluminum/glass/etc will have very low damping (and high Q modes). A lot of these material-dependent parameters will have to be calculated empirically, which I'm hoping to do in my lab for common materials. I can also just create a user input box for quality factor as a temporary solution! The edge damping is something I can model for sure, but I'll have to run experiments to determine how accurate the model is... 2. Yeah, that wouldn't be a problem for isotropic materials. These simulations tend to be tricky for non-isotropic materials like cardboard or posterboard. For example, I published some comparisons between measurement and simulation here (acoustics.org/how-to-find-the-best-material-for-making-exciter-based-plate-speakers/) showing how weird posterboard acts. Posterboard simulations are very accurate at low frequencies, and very inaccurate at high frequencies. I'll have to do more research into how to best model that - maybe a frequency-dependent Young's modulus or something along those lines. 3. Yeah, I can generate impulse response or waterfall plots in this software. The weird part about doing that is that they are very, very location dependent - 1m away on-axis is very different from 1m away at 45 degrees. It's really different in that way from a traditional speaker. I can put that in, though, and maybe allow the user to select a measurement point. 4. Right now the output is being modeled with respect to a reference input force (0.1 J). So sensitivity is *kinda* being modeled, but just not with respect to the normal 1W electrical power input. You'd definitely see that posterboard has a higher average output than acrylic. It does make more sense to translate everything back through the exciters to a sensitivity, so I'll work on adding that in. 5. FSFS and FCFC are in there already! I've been thinking about how to model rounded edges, so I'll see if I can get that in there eventually as well.
Brillant work Prof. Anderson! I look forward to seeing more of your findings in this fascinating field of engineering.
Wow this looks like it will be so helpful when making DML panels speakers!
It would be good to share this on the diyaudio Dml as a full range forum. People with more knowledge than me could have some helpful input.
I have been making panel speakers and have enjoyed the results with great sounding ones for home and this would make it so much easier than the trial and error many people go through.
Congratulations! I have been very interested in this topic for several years. It would be interesting to read the documents you have published on the subject. In the meantime, thank you, and Happy New Year!
Thanks! There are way too many publications to list, but my Google scholar account lists them all: scholar.google.com/citations?user=Dpw7u9IAAAAJ&hl=en&authuser=1
This is a really cool tool. I just binged all three videos. I can't say how excited I am too see something like this. Here are some thoughts about possible improvements:
1. Damping. Both within the panel itself and at the boundaries. An undamped (say FFFF) aluminum (or glass!) panel is truly horrible in real life, while the same in acrylic is much, much better. But in your video the aluminum didn't look nearly as bad as it should, nor as different as it really is from acrylic. Also, using damping materials at the supports changes the response dramatically, having some capability in the model for varying that seems critical.
2. Material Properties. Do you envision the ability to input self definable materials? Especially orthotropic? Composite panels with say, carbon fiber skins and lightweight cores are commonly used. But their properties vary with the specific design, so the ability to input the elastic properties directly (rather than choose a generic material type) would be critical.
3. Any chance of modelling the impulse response? Or a wavelet spectrogram? Frequency response is important, but the response in the time domain is equally important.
4. What about sensitivity/efficiency? An big challenge in flat panel design is getting good sensitivity. Acrylic has great internal damping but horrible sensitivity, due to is relatively low stiffness and high density, while skin/core/skin composites are much better in that regard. It would be great if you model could predict some measure of sensitivity.
5. Less importantly (to me anyway), but still nice, would be the ability to model shapes other than rectangular, including rounded edges. And to model mixed boundary conditions (FSFS) for example.
Great work, I can't wait to see what's next!
Thanks for the comment - these are all really great points. I'll try to respond in order...
1. Yes, damping is (inaccurately) set to the same value for all materials right now. You're right that acrylic has super high internal damping and aluminum/glass/etc will have very low damping (and high Q modes). A lot of these material-dependent parameters will have to be calculated empirically, which I'm hoping to do in my lab for common materials. I can also just create a user input box for quality factor as a temporary solution! The edge damping is something I can model for sure, but I'll have to run experiments to determine how accurate the model is...
2. Yeah, that wouldn't be a problem for isotropic materials. These simulations tend to be tricky for non-isotropic materials like cardboard or posterboard. For example, I published some comparisons between measurement and simulation here (acoustics.org/how-to-find-the-best-material-for-making-exciter-based-plate-speakers/) showing how weird posterboard acts. Posterboard simulations are very accurate at low frequencies, and very inaccurate at high frequencies. I'll have to do more research into how to best model that - maybe a frequency-dependent Young's modulus or something along those lines.
3. Yeah, I can generate impulse response or waterfall plots in this software. The weird part about doing that is that they are very, very location dependent - 1m away on-axis is very different from 1m away at 45 degrees. It's really different in that way from a traditional speaker. I can put that in, though, and maybe allow the user to select a measurement point.
4. Right now the output is being modeled with respect to a reference input force (0.1 J). So sensitivity is *kinda* being modeled, but just not with respect to the normal 1W electrical power input. You'd definitely see that posterboard has a higher average output than acrylic. It does make more sense to translate everything back through the exciters to a sensitivity, so I'll work on adding that in.
5. FSFS and FCFC are in there already! I've been thinking about how to model rounded edges, so I'll see if I can get that in there eventually as well.