Yeah, I agree this is really annoying. The callback in matlab is on "value changed" (there is no "value entered" callback, as far as I know), so it doesn't ever run a new simulation unless the value is literally changed. I'm trying to figure out a workaround, I promise!
So I think I worked out how to put the results I shared yesterday in dimensionless form. Seemingly it requires a new dimensionless parameter.
As a reminder, the question I'm trying to answer is how much error does thin plate theory introduce to the determination of the natural frequencies. My hypothesis is that the error could be large for high order modes, since their wavelengths are shorter and at sufficiently high frequencies could approach the thickness of the panel. So again, I expect that the ratio of the wavelength to the panel thickness should be the main factor determining how much error is introduced.
So the idea was simply to come up with an expression for the ratio of wavelength to thickness. It' basically the same thing I did for the "coincidence frequency" analysis, except this time looking at the solution for an arbitrary frequency, rather than particularly at coincidence.
So here it is: Please check my math!
:
I want to highlight here that while this index also represents lambda/t, it is a different index than the one that applied to the coincidence frequency, and unlike that one, this one is a function of not just the material constants, but also frequency and panel thickness. (Since index #2 applies generally, and index #1 only at coincidence, maybe #1 can be derived from #2, but I have not tried yet).
I calculated the value of "thin plate index #2" for the results I shared yesterday, which compared the natural frequencies determined by Pettals and by FEA. Then I plotted that frequency ratio as a function of the new index, below.
This plot actually includes more data than the earlier plots, and includes data for two different sizes of EPS panels (orange dots) and four sizes of Acrylic panels (blue dots). As I'd hoped, all of them seem to collapse onto a single curve. The results suggest that as long as the value for "index #2" is greater than about 10, the error is relatively small. However, below 10, the error rises rapidly with decreasing index.
Assuming this analysis is not flawed in some way, there are a couple of ways this result could be used in Pettals. Perhaps the simplest would be a warning that would pop up when the index gets too low saying something like: "the results above x Hz may be inaccurate due to limitation of thin plate theory". But it might also be possible to use this approach to apply a correction to the modal frequency values which is based on the value of the index.
Eric
As a reminder, the question I'm trying to answer is how much error does thin plate theory introduce to the determination of the natural frequencies. My hypothesis is that the error could be large for high order modes, since their wavelengths are shorter and at sufficiently high frequencies could approach the thickness of the panel. So again, I expect that the ratio of the wavelength to the panel thickness should be the main factor determining how much error is introduced.
So the idea was simply to come up with an expression for the ratio of wavelength to thickness. It' basically the same thing I did for the "coincidence frequency" analysis, except this time looking at the solution for an arbitrary frequency, rather than particularly at coincidence.
So here it is: Please check my math!
:
I want to highlight here that while this index also represents lambda/t, it is a different index than the one that applied to the coincidence frequency, and unlike that one, this one is a function of not just the material constants, but also frequency and panel thickness. (Since index #2 applies generally, and index #1 only at coincidence, maybe #1 can be derived from #2, but I have not tried yet).
I calculated the value of "thin plate index #2" for the results I shared yesterday, which compared the natural frequencies determined by Pettals and by FEA. Then I plotted that frequency ratio as a function of the new index, below.
This plot actually includes more data than the earlier plots, and includes data for two different sizes of EPS panels (orange dots) and four sizes of Acrylic panels (blue dots). As I'd hoped, all of them seem to collapse onto a single curve. The results suggest that as long as the value for "index #2" is greater than about 10, the error is relatively small. However, below 10, the error rises rapidly with decreasing index.
Assuming this analysis is not flawed in some way, there are a couple of ways this result could be used in Pettals. Perhaps the simplest would be a warning that would pop up when the index gets too low saying something like: "the results above x Hz may be inaccurate due to limitation of thin plate theory". But it might also be possible to use this approach to apply a correction to the modal frequency values which is based on the value of the index.
Eric
Dave,
Another minor point, but I'm thinking of it now: When there is an alert in the message field, it should be bolder, or red, or something. It's easy to miss, and not realize why it won't run. Also, the alert should disappear when the issue is corrected.
Eric
Another minor point, but I'm thinking of it now: When there is an alert in the message field, it should be bolder, or red, or something. It's easy to miss, and not realize why it won't run. Also, the alert should disappear when the issue is corrected.
Eric
That would seem to be the preferred approach - but the ultimate output of PETTaLS (Dave - can we 🙏 workshop a better name for this before you fully release it ? -😊) for speaker design is the FR graph so my question is if the natural frequencies are 30% higher at high frequencies than that indicated in P....LS, does that imply that the FR graph is shifted to the right in that frequency range??
Eucy
I have to agree with Eucy that I don't love the name Pettals. But I haven't got a better suggestion either.
Eric
Yes, I think so, as long as what you mean is that the Pettals result is shifted a little to the right, and the correction would shift it back to the left. At least it would for thick panels at the highest frequencies. I think for most panels under 3 mm thick any correction would be pretty small. 25 mm thick EPS would be another story.
Eric
Thinking about the name. The acronym thing isn't really working for me. How about "Waves" (Wave Simulator? if you must), or "Bender"
I'll have to ask my wife. She has a communications background. She thinks the name is the most important thing. "Who cares if it works, what's it called!"
Eric
I'll have to ask my wife. She has a communications background. She thinks the name is the most important thing. "Who cares if it works, what's it called!"
Eric
When I was but a wee lad, our long disgraced media personality Rolf Harris marketed a popular item..
it was a piece of Masonite painted in Aboriginal motifs. You gripped it at both ends and shook it back and forth to make a somewhat pleasing rhythmic sound... It was used in his famous song "Tie me kangaroo down, sport". Well.. famous here...
It was called a 'Wobble Board'.
Not that I'm suggesting "Wobbler" of course😂
Eucy
Ridiculous acronyms are all the rage in academia. I also do research at work in deep learning, and the acronyms are absolutely out of control in that space.
I agree that PETTaLS is a silly and convoluted name. I was originally thinking of this software as something that would let people try out all of the different exciters offered by PE/Dayton Audio, and since they call them "exciters and tactile transducers," that's the wording I used in the title. It's still that, but it's becoming much more than just that, so a broader title is probably appropriate.
I moved the current frequency display to the title of the graph! I'm not going to keep uploading new releases every few days anymore, though, because that's getting to be a bit confusing. I'll collect a handful of updates at a time and then produce a new release.
I apparently can't change the text color, but I added an error symbol...
That goes away when the error is fixed...
I eventually downloaded the software (Version 1.3, now a bit outdated already) and managed to get it running on my ancient Windows laptop. I had great fun last night playing with the basic parameters, but I’m afraid the DML thread and now the new PETTaLS thread is developing so fast I can’t keep up and don’t understand a lot of the discussions and technicalities.
Anyway, I am curious to see how PETTaLS can help me with my next dml build and it would be great if Dave can maybe help to simulate my panel with parameters more true to my build. Maybe use the advanced version? Here is what I am planning - adding atmosphere dmls to compliment my existing system. To date I have not been able to build dml speakers that will dethrone my current MLTL full-range speakers. So, I’m hoping to add small (very small) panel speakers that can bring some dml magic to the table, and also act as centre channels like member @aagas has done.
Here is what I have - an A4 framed canvas with a 2mm acrylic panel (230mm x 140mm) glued to the front of the canvas with contact rubber based glue. Not the whole panel is glued, but only a 10mm border around the outer edges of the acrylic panel. I then cut the unglued portion of the canvas away with a Stanley knife at the back of the framed canvas. So the exciter can be glued directly onto acrylic panel from the back and you have the whole acrylic panel exposed (not covered by the canvas) at the front. So the acrylic panel is actually just suspended by the canvas. I can share more details (and pictures) later if someone is interested. Let’s see how it performs first.
So, I have XT32-4 exciters available for this build, but they are not fixed unto the panel yet. The 19mm exciters might be betterfor this small panel. That is where PETTaLS can perhaps help - where is the best exciter position? This is in fact a question that many future builders might want to answer, using PETTaLS. Is there a way to predict/calculate/discover the best exciter position (except just trail and error) - given that the panel material, dimensions and boundary conditions are already decided upon?
As a start to simulate my particular build I played around with PETTaLS using the following parameters:
Acrylic, 2mm, 230mm x 140mm, using SSSS boundary conditions, and the closest exciter - Dayton DAEX25VT-4
I played around with the exciter X and Y positions until I found a reasonable FR plot. But is it the best position? And what will it look like using the XT32-4 exciter? Time will tell.
Here is what I got in PETTaLS. Any comments or suggestions are welcome.
Thank you, Dave, for a great tool!
Anyway, I am curious to see how PETTaLS can help me with my next dml build and it would be great if Dave can maybe help to simulate my panel with parameters more true to my build. Maybe use the advanced version? Here is what I am planning - adding atmosphere dmls to compliment my existing system. To date I have not been able to build dml speakers that will dethrone my current MLTL full-range speakers. So, I’m hoping to add small (very small) panel speakers that can bring some dml magic to the table, and also act as centre channels like member @aagas has done.
Here is what I have - an A4 framed canvas with a 2mm acrylic panel (230mm x 140mm) glued to the front of the canvas with contact rubber based glue. Not the whole panel is glued, but only a 10mm border around the outer edges of the acrylic panel. I then cut the unglued portion of the canvas away with a Stanley knife at the back of the framed canvas. So the exciter can be glued directly onto acrylic panel from the back and you have the whole acrylic panel exposed (not covered by the canvas) at the front. So the acrylic panel is actually just suspended by the canvas. I can share more details (and pictures) later if someone is interested. Let’s see how it performs first.
So, I have XT32-4 exciters available for this build, but they are not fixed unto the panel yet. The 19mm exciters might be betterfor this small panel. That is where PETTaLS can perhaps help - where is the best exciter position? This is in fact a question that many future builders might want to answer, using PETTaLS. Is there a way to predict/calculate/discover the best exciter position (except just trail and error) - given that the panel material, dimensions and boundary conditions are already decided upon?
As a start to simulate my particular build I played around with PETTaLS using the following parameters:
Acrylic, 2mm, 230mm x 140mm, using SSSS boundary conditions, and the closest exciter - Dayton DAEX25VT-4
I played around with the exciter X and Y positions until I found a reasonable FR plot. But is it the best position? And what will it look like using the XT32-4 exciter? Time will tell.
Here is what I got in PETTaLS. Any comments or suggestions are welcome.
Thank you, Dave, for a great tool!
@twocents
It is a great tool and surely thanks to Dave, but he's not here to run sims for us, so that's a bridge way too far.
He's responding freely to many requests and I'm concerned the project will become too much of a burden... Time maybe to be patient, including me.
Eucy
True. Sorry Dave. I seem to recall such an invitation at the start of all this. Could have been in one of the videos.
I also discovered (duh) that unclicking the ""Plot surface as simulation runs" causes the simulation to run so much faster. Now it's easy to do "trail and error" runs.
Any chance that you have the ability to measure impedance? The way you've described the setup, it sounds like it's probably going to be halfway between free edges and simply supported edges. Acrylic is going to be modeled very well in pettals, but it's hard to tell how to properly set up the simulation with the canvas boundary - sounds like a good idea, though!
The tough part to me about determining the best exciter location is that we have to define what best means - there's no achievable perfect acoustic response out there, so we just have to minimize some kind of cost function (this is the EE prof coming out). I haven't read the papers where people try to find the optimal exciter location in quite some time (I remember Ben Z gave a talk on this at an AES convention a number of years ago), but it seems like the objective was to minimize the peakiness in the on-axis acoustic response. This is not necessarily a bad target, but I suppose it's worth asking: do you what types of problems you want to try to minimize in your design? What bandwidth are you going for? Are lows or highs more important in the design?
Here's a set of simulations that I ran (I'm happy to do this if I happen to have time...) showing a 32mm exciter vs a 19mm exciter at a variety of locations.
32mm exciter: sharper roll-off in high frequencies. Much higher magnet displacement at low frequencies. Variability in FR with location stops around 2kHz, pretty spatially uniform between 3-20kHz.
19mm exciter: Less roll-off. Low magnet displacement. Variability in FR extends with location is from roughly 150 Hz up to 8 kHz.
In either case, the best exciter location is around the (1/3 x 1/3) placement, as we all know. I don't know that there's a ton of variability in response with slight variations in placement as long as you avoid the edges or direct center of the panel.
I can re-run this with a FFFF panel when I get some more time!
Hi Dave, thank you for your reply and taking the time to do some simulations with a 32mm exciter.
I am hoping to find the magic spot with REW’s RTA function. If I can get it working one day. If you have golden ears you can also determine the best sounding spot by moving the exciter around while music is playing. But I would like to measure and simulate it too.
For music I would be very happy if the dmls could cover Low mids (250 – 500 Hz), Center mids (500 – 2000 Hz), and Upper mids (2000-4000 Hz).
The best sounding panel speakers I have heard were canvas framed panels made with Nidaplast, and I was somehow inspired by other canvas builds posted on the DML thread, as well as the Sonance speakers mentioned in this post.
Maybe yes. I hope. I have a little Dayton Audio iMM-6 Microphone and I have REW installed. The rest is a steep learning curve mountain I still need to climb.
In my case I think it is rather something halfway between simply supported and clamped edges. The canvas surrounds are pulled really very tight. When you tap the panel lightly it sounds like a Bodhran drum.
Yes, I specifically chose acrylic because it is universally the same, available everywhere, it’s isotropic, and it has been used by others (even commercial speakers) with success. I actually wanted to use 2mm polycarbonate, which I think might even have better physical properties for dml, but I think I bought acrylic. Can’t remember really - it was bought long ago. I might even have a polycarbonate panel glued to my A4 canvas. How can I know/find out?
Yes, I agree. There is probably no universal magic formula for best. Too many other variables involved.
I am hoping to find the magic spot with REW’s RTA function. If I can get it working one day. If you have golden ears you can also determine the best sounding spot by moving the exciter around while music is playing. But I would like to measure and simulate it too.
Good questions and I don’t have clear answers. I was hoping to aim for a bandwidth of say 300Hz upwards to about 8kHz, so I was thinking mids and highs are more important for this design. The other design goal was to be used as centre channels to improve voice intelligibility when watching movies. My hearing is still good and well trained, but my wife’s hearing is deteriorating. Tough getting older. Wikipedia says - “In telephony, the usable voice frequency band ranges from approximately 300 to 3400 Hz.[2] It is for this reason that the ultra low frequency band of the electromagnetic spectrum between 300 and 3000 Hz is also referred to as voice frequency”. So my dml should be ok if it can cover this bandwidth I guess. The 32mm exciter should work fine here, I hope.
For music I would be very happy if the dmls could cover Low mids (250 – 500 Hz), Center mids (500 – 2000 Hz), and Upper mids (2000-4000 Hz).
The best sounding panel speakers I have heard were canvas framed panels made with Nidaplast, and I was somehow inspired by other canvas builds posted on the DML thread, as well as the Sonance speakers mentioned in this post.
No need for now, thank you Dave.
Last edited:
twocents,
I think Dave is right here. I would suspect your case is closer to "free" than "simple" edges. Clamped implies perhaps more constraint than you might imagine. It means that the edge of the panel can neither move, nor rotate, like the image below. It's frankly hard to achieve a truly clamped edge even if you are trying to. Envision two bulky frames with faces one inches wide or more between which you sandwich your 2-3m mm thick acrylic panel, using sturdy bolts every few inches. That would approach "clamped". No matter how tight you pull the canvas, the edges will be able to move a little bit, and rotate with virtually no restriction. So, probably not even close to clamped.
Eric
That is something I have been meaning to get to eventually. I'm really curious what others think. I have my own ideas about that. I'm pretty eager for that conversation, and I suspect there are different perspectives on it! I believe Christian is interested in all 360 degrees of radiation. I'm a little more interested in the front 180 degrees. I think the biggest challenge is the balance of sensitivity (for decent SPL without distortion) and damping (to avoid big peaks and dips and ringing). Others may see a different challenge. I'm most interested in about 100 Hz to 10kHz, others may have a different range in mind. Fun, fun fun!
Eric
Dave,
I want to emphasize how much I appreciate this post. I think it really helps me put some things in a new perspective.
In flat panel speaker "lore" there is a lot of idealization of the "pebble in pond" wave, as in these snips from the Goebel site:
I've always thought that the Goebel site was at least a little "hyperbolized" (like most audio lit), but now I'm more convinced that there is really no great advantage of the PIP pattern over the simply "localized" pattern. Both excite a sufficient number of modes to produce a very wide and even radiation pattern. So I'm a a lot less inclined to target the "PIP" response over a wide frequency range, if the "localized" response provides similar directivity without the need for extreme damping measures.
Does this make sense, or am I missing something?
Eric