Hello Lekha,
There is almost no realization or feedback of panels with box. Please keep us inform of the results.
In the video you linked, if we suppose there is no change in the recording set-up (the room being the same), the sound of the rectangular panel is better (at least to me) than the one of the circular driver even if the rectangular has something strange in the bass.
In this case, FR at the listening distance would help to understand the difference between both while some proximity measurements (few centimeters from the panel at the exciter position, a second one at the vent) would help to understand what happens in the bass, if the bass reflex is correctly tuned.
Be ready to make different dedicated measurements.
Christian
Hello Mark,
There are far in the past of this thread posts about this kind of material. I found this one #8594. There are probably others before. I don't remember results shared about it.
If I am not wrong in the calculation, the one you linked is 5.7kg/m² (2000kg/m³). Twice the density of a polystyrene glass (which is close to acrylic), between three to four time the density of plywood.
Not good for the efficiency.
The core being a foam so not that heavy, the alu layers might be thick so the panel really stiff.
Like that I won't put it in my short list.
Christian
A flat panel is a different kind of beast, so the software designed for testing cone speakers has no understanding of how a flat panel should sound.
I'm not at all interested in any kind of measurements, especially at a 1-meter distance. After all, everyone has different ears. Attaching an exciter or a few of them to a flat panel is an easy task, but finding something aesthetically pleasing is the challenging part, especially when one doesn't have (m)any rooms available for listening. One can cut all kinds of panels to various sizes, hang them on strings, frame them, etc., and this has been happening for over a decade, yet we are still at that early stage.
Anyway, here's an image of handmade speaker drivers in a handmade speaker box, created by a certain specialist in another country. The mid/bass speaker membrane is actually convex, while the tweeter cum mid cone is concave. This tells that a membrane doesn't need to be terribly massive to produce good sound, and it doesn't have to be concave, let alone flat. (No, I have not heard them, but I can believe those who have. No measurements are available.) All I know is that they are made of paper and painted.
I agree with Christian here. I built quite a few prototypes back in the day using different types of aluminum sandwich board material, and there are definitely pros and cons of the material, with cons probably outnumbering pros for most situations.
It is low-Q (e.g. internal damping is very high), which makes it sound good with vocals, but it's going to sound "dead" with most music. It's also heavy, which reduces the sensitivity/efficiency but also makes it easier to mount close to a wall without disturbing the resonances too much compared with a lighter sandwich board material. Another plus is that printing on aluminum looks very nice, and I've seen several companies try to make speakers out of wall art printed on aluminum.
Dave
Hello Dave,
With this introduction of the quality factor of a material, you make a bridge between a metric (something subject to a measure) and how a material can be subjectively perceived. This is fully in relation with what for example Steve (@spedge ) reported from his different tests of materials (I have in mind his test of 2 alu sheets glued together which gave a smooth FR but not a good panel, his proplex panel seems to be in the good range, I don't have in mind example of too ringing material... uncoated XPS?)
After the possibility of detecting the coincidence frequency with the directivity tests, it is for me among the next important steps.
To help all to understand (if needed) what is behind : the quality factor, as for any resonant system so here for the modes, is a measure of how sharp the peaks of resonance are (basically it is the ratio of the frequency of resonance to the -3dB bandwidth around it). A material highly resonating has a high Q, a material highly damped has a low Q. For the DML, to get the right Q allows to reproduce the frequency in between modes. Side comment : For now if I understand how a not too high Q helps to get a smooth enough FR, I don't understand the physical effect of a "too low Q"...
To come back to the topic :
Would you help us in that?
I see it in several sub-topics :
- Is there a range of suitable Q?
- How is the Q of some usual materials (EPS, XPS, plywood, polypropylene, glass polystyrene, acrylic...)
- How a DIYer could evaluate the Q of the material he has sourced or he is working on (ie while applying successive layers of damping) ? I see 2 tests as candidates : the tap test and the impedance measurement (probably because being more familiar with it, I am more in the favor of the second which is I think more robust in its set up - pure opinion!).
The material I'm referring to is reported to have a solid plastic core. So denser than foam. It probably would exhibit less damping but it sounds like the extra mass would be problematic.
If your panel material is something like this, they should work quite well, even if the panel is not as large as yours. Or, something like this. You'd be surprised at the sound you can get from an A4-sized panel, or less than that. 🙂
Mark'51
I have tried that type of material, and as both Christian and Dave had said, it's efficiency is very poor due to its weight. There is another similar material called Alumilite Economy but made with a corrugated polypropylene core rather than a solid core, and hence has lower weight and better efficiency. The 5 mm had the best efficiency of the two, but is overall too stiff for my taste, as it requires a fairly large panel size to reach lower frequencies.
You can see the Alumilite here:
https://www.laminatorsinc.com/sign-panel-products/
Eric
Hi Dave,
Is there a less subjective description of the kind of response would sound "dead"? That is, lacking of some particular frequency range, or other measureable characteristic? And can you explain why low-Q would result in that? Like Christian asked, I would be interested to know what you think of as the optimum range for Q.
Also, I'm eager to hear of any progress on your model. With such a tool, we might be able to answer questions like this ourselves!
Thanks,
Eric
@homeswinghome @Veleric OK, I finally got around to posting a new video, it should be available here. I tried to keep the discussion to material parameters and Q factors. I'll be posting more videos as time permits!
I don't have any precise answers about material Q factors. The idea that a material has one Q value is an oversimplification, as Q tends to be frequency dependent and is based on a large number of effects and factors. As I'm sure you all know, adding extra damping at the panel edges can affect the Q value of the peaks in the response as well. DMLs are unique because they use resonances to recreate sound as compared with a traditional loudspeaker, which is largely resonance-less (besides cone breakup!).
In my experience, low-Q devices materials well as small panels, and high-Q materials work well as large panels with free edges, when the audio range entirely avoids the region of low modal overlap (where there are sharp, isolated peaks). It's very subjective, though, and that's part of the fun of these things!
By the way, here are some features that I added to my software:
I don't have any precise answers about material Q factors. The idea that a material has one Q value is an oversimplification, as Q tends to be frequency dependent and is based on a large number of effects and factors. As I'm sure you all know, adding extra damping at the panel edges can affect the Q value of the peaks in the response as well. DMLs are unique because they use resonances to recreate sound as compared with a traditional loudspeaker, which is largely resonance-less (besides cone breakup!).
In my experience, low-Q devices materials well as small panels, and high-Q materials work well as large panels with free edges, when the audio range entirely avoids the region of low modal overlap (where there are sharp, isolated peaks). It's very subjective, though, and that's part of the fun of these things!
By the way, here are some features that I added to my software:
- sensitivity (you can specify the input wattage)
- bulk material parameters (density, young's modulus, Q, etc)
- Sandwich board model support
- You can now specify any frequency for a surface plot
- Impulse response as a waterfall plot
- Support for fixed exciter magnets
- Support for simple enclosures
- Impedance (I derived this as an LTSpice model years ago - gotta figure out how to translate that into the PETTaLS code)
- exciter magnet displacement vs. frequency
- More exciter models
- More plate materials
- Maybe support for simulating unbaffled acoustic response? This one is questionable.
Oh, and I forgot to respond to the question about how to measure Q. The impedance method is probably the most straightforward - it'll work best if you place the exciter right on the peak excursion point of a big, isolated mode (as I'm sure you know). The tap test would work too, but acoustic measurements are prone to all sorts of extra problems from room reflections, weird modal radiation patterns, etc, so it'll be more difficult to get good data.
Dave,
Again, just awesome. I'm so happy to see your progress and it sounds like you think it will be available for us to try out soon. I can't wait! The additions are great. I have so many comments I want to make, but for now I will try to confime myself to some suggestions for your consideration.
1. For the material properties inputs, and especially the sandwich panels, I renew my suggestion to have the option to use an orthotropic Mindlen *(thick plate) model, where users would input their own set of material properties. For sandwich panels, there is no way for you to anticipate all the material combinations someone might want to try. You can test a few and add them, but you can't really even specify "posterboard" as a particular set of elastic properties, because for every thickness the apparent elastic properties will be different. I suggest putting the burden on the user to know or measure the properties of any material they want to input (other than a simple set of isotropic materials like you already have). Everything else could be described by input of the elastic constants. For an orthotropic plate one would need the six highlighted independant properties here:
I have already shared a methodology using equipment most of us have (REW), that allows deterimination of all of these with the exception of G23 and G13. But those can be simply estimated as the values of the G23 and G13 of the core material, which can often be looked up or guessed at. Also, I believe this treatment would eliminate the need to add the artifice of a sharp transition between a frequency dependant wave speed and a constant one, which you described in the video. In my mind, that "transition point" is a pure guess. With a "thick plate" formulation of the plate equations, the two shear (G23 and G13) will automatically (I believe) flatten the wave speed function appropriately. I fear my explanation is poor, but I'm sure you have a contact in Engineering Mechanics that can confirm (or refute!) this idea.
2. I'm really glad to see the impulse response added, but personally I hate the waterfall presentation. The attempt to create a 3D appearance gets in the way of seeing the whole thing clearly, and it's really hard to trace back the ringing ridges to their actual frequencies. It's not a criticism of you waterfall plot, but rather of all of them. I really find the wavelet spectrogram presentation to be much easier to interpret. I think it's based on exactly the same information, but just provides a much easier visual to interpret (to me anyway.) The spectrogram I'm talking about looks like below, for a high Q panel. The frequencies which ring the longest are easy to pick out in the representation. Some prefer the two axes swapped, either is good for me.
Concerning your list of things you'd like to add eventually:
1. The impedance model is the one I'd like to see most. Since that is a thing that is pretty easily measured, it would be a great way to double check model inputs, especially with respect to the Q value chosen, but also elastic properties.
2. I'm also wondering if you can add a "reflection coefficient" or similar to the boundaries? The plate itself can have a Q value, but oftentimes the mounting foam or whatever can provide damping effect which is much greater than the internal damping of the material. I'm always wondering how different the response would be from a panel with very low internal damping, but high boundary damping, vs a similar panel but with high internal damping and low boundary damping. I know it's especially tricky because the boundaries would no longer be "simple" or "free" or "clamped", but I think you can have boundaries that are nearly "simple" but still provide very substantial damping.
3. Have you thought about separating the input and results on separate "pages"? Or even multiple results pages? The plots end up being pretty small when you pack it all onto one screen.
Thanks again for your great work.
Eric
For those who are still interested in "slapping" an exciter onto a panel, there are still individuals writing patents. Here is one of them. It's quite recent, having been granted in 2021.
"Typical electrodynamic force actuators are designed with a cylindrical voice coil, the top surface of which has an adhesive material applied to it so that it may be bonded to the object to which the vibrational force is to be applied. The contact area between the force exciter and the panel is annular in typical applications. A problem with this configuration is that at high frequencies, the area of the panel within the force actuator's annular contact area may vibrate out of phase with the panel area outside the exciter's annular contact area. This occurs at the resonant frequency of the first “drum head” mode of the panel area within the annular contact region. Under these conditions, and due to the phase reversal between the vibrating region within the annular contact area and the panel area outside the contact annulus, the net acceleration of the panel, when integrated over the entire panel area, decreases noticeably. ..."
"Typical electrodynamic force actuators are designed with a cylindrical voice coil, the top surface of which has an adhesive material applied to it so that it may be bonded to the object to which the vibrational force is to be applied. The contact area between the force exciter and the panel is annular in typical applications. A problem with this configuration is that at high frequencies, the area of the panel within the force actuator's annular contact area may vibrate out of phase with the panel area outside the exciter's annular contact area. This occurs at the resonant frequency of the first “drum head” mode of the panel area within the annular contact region. Under these conditions, and due to the phase reversal between the vibrating region within the annular contact area and the panel area outside the contact annulus, the net acceleration of the panel, when integrated over the entire panel area, decreases noticeably. ..."
It's fortunate that someone has previously considered this "idea" here. Independently of the people here, another group had been experimenting with this type of "idea," even without being aware of the existence of this "patent." Interestingly, they are concentrating on the need to develop their own voice coils, coil formers (or the lack of that), and magnetic systems, which is essentially what this patent is suggesting. The presence of holes is not the main focus here or there.
Hello Dave,
Thank you for this new video. It gives a first idea of what could be a Q range. Acrylic with a Q of 8 is among the dead material, aluminum at 40 is much too ringing. So maybe a range 5 to 20 is already too wide.
It is interesting to see how the speaker changes from DML (there are modes) to a bending wave driver (low Q, no modes) with the reduction of the emission area when the frequency increases.
About your tool :
- in the "material box" you introduce a frequency at which the core is limiting the increase of the wave speed. I think you named it Fc (as Corner frequency?) but fc is also used for coincidence frequency... Maybe my understanding is wrong...
- as Eric, I am not a fan of the 3D waterfall. I prefer the 2D spectrogram view. The time scale of 200ms is useful for the tool development and to learn about the influence of very high Q but for a loudspeaker design point of view, the horizon is probably more 20ms or even less.
- About the scales. In the FR, a 50dB scale is generally used with a frequency scale so that the distance from 100 to 1k (a decade) has the same length (mm, px...) as the 50dB range (they draw a square). For a spectrogram, 30 dB is maybe enough.
Christian
By the way, the owner of that patent is the University of Rochester. If a business were to attempt to utilise it, the university would likely earn some revenue; otherwise, it may fade into history. In any case, if you're interested, it's worth checking what patents such universities hold.
You don't really have to search all over the internet if you know what you are looking for. It is right here. All you need to do is utilise the idea to implement it in that flat panel DML ... to create the membrane, and the voice coil, coil former (or the lack of that), and magnetic system.