Christian,
I finally got around to looking at your 9 mm (oops, 9.4 mm) PS foam tap test results.
Your conclusion about the flexure mode for hanging by the short side was correct. That is, for the 20x30 version of the panel, the 219 Hz peak was the resonant frequency of the lowest flexural mode for that direction. The 100 Hz was the flexure mode for the other direction, which you identified in the other test. Normally, when I look at the tap results for hanging from the short side, I ignore any peaks that I saw in the previous two taps (torsion and long side flex) and look for the lowest "new" resonance that hasn't already been found in the torsion and long side flex tests.
Using the FEM model and fitting your tap results, I got the best fit for:
E1=28 MPa
E2=26 MPa
G=12 MPa
Note, I assumed the density was 32 instead of 33, assuming that you used 9 mm instead of 9.4 mm when you originally calculated the density. I'm not sure if that assumption was correct or not. In any event it would not matter a lot.
When I looked at the tap results for the larger panel (20x60), I had to use a lower value for E1 (about 24 MPa). I don't know what the reason for that is.
Eric
I finally got around to looking at your 9 mm (oops, 9.4 mm) PS foam tap test results.
Your conclusion about the flexure mode for hanging by the short side was correct. That is, for the 20x30 version of the panel, the 219 Hz peak was the resonant frequency of the lowest flexural mode for that direction. The 100 Hz was the flexure mode for the other direction, which you identified in the other test. Normally, when I look at the tap results for hanging from the short side, I ignore any peaks that I saw in the previous two taps (torsion and long side flex) and look for the lowest "new" resonance that hasn't already been found in the torsion and long side flex tests.
Using the FEM model and fitting your tap results, I got the best fit for:
E1=28 MPa
E2=26 MPa
G=12 MPa
Note, I assumed the density was 32 instead of 33, assuming that you used 9 mm instead of 9.4 mm when you originally calculated the density. I'm not sure if that assumption was correct or not. In any event it would not matter a lot.
When I looked at the tap results for the larger panel (20x60), I had to use a lower value for E1 (about 24 MPa). I don't know what the reason for that is.
Eric
Last edited:
Sonnar,
Cool stuff about the violin. I like that they call it tap testing too. I thought I made that name up myself!
Eric
Hello Eric,
Thank you for your feedback.
From my point of view, this tapping test is validated as an additional valuable method in the tool set around DML. It allows to approach the characteristics of the material we have in hands. A higher precision might be better but as it is, it is much better than values from the net from material which are perhaps not the same.
About the FEM software, I started to try using freeCAD. It is a 3D tool I used some months ago to design enclosure for a bass unit. It has FEM possibility; different mesh tools and different solvers are possible. It runs under all the main OS (Linux, Windows, Mac); I am at the very first steps tryng to solve some errors to get the examples working. The capacity for modal analysis seems present.
To make 3D drawing (I was used to make 2D drawing from the time the drawing board was the tool!) it is not really difficult. There are many tuto.The function to export from the 3D a 2D drawing is even nice. FEM is an other (new) story.
As a Linux user, the good point is this software is now available from the repository of the distribution I use (Manjaro/Arch)
Christian
I'm glad you found something that might work for FEM. If it was LISA I could help more.
FEM won't answer all the questions, but it's a huge help. Together with tap testing, it's kind of like actually seeing for the first time. Designing a DML becomes a bit more like engineering and a little less like guessing.
Today I was playing with the idea of using PS foam for a subwoofer. I don't care for the sound quality of XPS foam in the voice range, but the high efficiency of XPS makes it a good candidate for a subwoofer to fill in the octave from 50 to 100 Hz. My XPS is about 14 mm thick. I found that, unlike for a plywood panel, the addition of the exciter (25fhe) increases the frequency of the fundamental. In my case it was about 10 Hz, (say 40 Hz to 50 Hz). And when I tried a DAEX32U the shift was even greater.
Eric
About subwoofer, have you had a look around "infraflex"? It is a DML (without naming it like that in the forum about it) made from a large square EPS membrane (up to 1mx1m) moved by a woofer. If I remember, the membrane is clamped all around giving a circular opening. I think all is in French. Perhaps start here "Infraflex, visit at the builder" or "Infraflex pictures". There are other pages, I just pick 2 quickly. I can search later the most relevant pages if you have an interest for more detail; I think that about the membrane, the suspension you will learn few things... but who knows, perhaps a look around how the woofer is used as an exciter... there are pictures somewhere if I remember. It is a topic a bit like the main thread about DML... not so long but similar. If you search on that in the Melaudia forum, you will probably cross posts from a "Christian", not me. Just same first name.
Funny point : trying to find quickly some page relevant for you around the infralex, I found a piece of discussion about the suspension. One builder making test without suspension and having no bass, an other replying that the only solution is to work in flexion.
For FEM, I am solving some installation problems. As usual, the first step is to find the missing software module and which part of the tutorials are no more true due to the software improvements. I would have given a chance too Lisa if a Linux version have existed but I don't want to have to solve problem of compatibility in Linux environment in addition to the FEM learning.
Christian
One has to wonder what the phase is between say, the 4 spots about the end of the neck of the left one. Are two going out, while the other two are going in? Could be...
I've always been amazed by Gibson's ability to craft an instrument - either hollow or solid body - that has such resonances in tune with concert pitch. I've owned two, one of each type that did this. The guitars would "come alive" when you tuned up the strings.
I wonder what would show if one did the above to that other Joe's patented Horn guitar, from another thread?
I've always been amazed by Gibson's ability to craft an instrument - either hollow or solid body - that has such resonances in tune with concert pitch. I've owned two, one of each type that did this. The guitars would "come alive" when you tuned up the strings.
I wonder what would show if one did the above to that other Joe's patented Horn guitar, from another thread?
@Veleric + all
Just to share I started some additional tests in the idea of the Chladni patterns. No so nice pictures as with the guitar above but probably informative, at least that help to understand the DML behavior.
apart the possible link with the mode type, I was impressed seeing vibration in modes without sound. See the holes in the FR.
Below some slides to explain.
To be improved probably.
Advice welcome
Christian
Just to share I started some additional tests in the idea of the Chladni patterns. No so nice pictures as with the guitar above but probably informative, at least that help to understand the DML behavior.
apart the possible link with the mode type, I was impressed seeing vibration in modes without sound. See the holes in the FR.
Below some slides to explain.
To be improved probably.
Advice welcome
Christian
Christian,
The Infraflex thing looks interesting. But it's hard to know what exactly it is without understanding French.
Are they removing the cone from the woofer and physically attaching the coil to the panel with some sort or adapter? Or just driving the panel by driving the air into it? Zenker had an article about driving a panel with a woofer as I recall.
Eric
The Infraflex thing looks interesting. But it's hard to know what exactly it is without understanding French.
Are they removing the cone from the woofer and physically attaching the coil to the panel with some sort or adapter? Or just driving the panel by driving the air into it? Zenker had an article about driving a panel with a woofer as I recall.
Eric
Eric
Since I understood DML are not limited as the standard speaker in the bass or at least not by the same phenomena, yes it is interesting.
Yes I think they remove the cone and place a kind of cylinder just at the voice coil to come in contact with the membrane like an exciter.
I have stored some documents about this Infraflex. I will search for that and translate what you need or at least find some pictures. Will need probably some time.
Christian
Christian,
Nice work on the Chlandi plate! I tried a very similar thing a few weeks ago. I did see some shapes but not even as good as yours.
For my plate I used and aluminum sheet attached to a frame by foam, since I was trying to simulate something close to the construction method I use usually. I chose the aluminum plate only to minimize the damping and hopefully encourage the shapes to be much clearer. Like you, I used a strip of foam around the edges as a wall to hold in the particles. I tried salt and sugar, but not rice.
Overall, I was disappointed in my experiment turned out. Nothing as pretty as the Chlandi images you find online, or even as clear as yours.
This is a 20x30 plate, right?
Eric
Nice work on the Chlandi plate! I tried a very similar thing a few weeks ago. I did see some shapes but not even as good as yours.
For my plate I used and aluminum sheet attached to a frame by foam, since I was trying to simulate something close to the construction method I use usually. I chose the aluminum plate only to minimize the damping and hopefully encourage the shapes to be much clearer. Like you, I used a strip of foam around the edges as a wall to hold in the particles. I tried salt and sugar, but not rice.
Overall, I was disappointed in my experiment turned out. Nothing as pretty as the Chlandi images you find online, or even as clear as yours.
This is a 20x30 plate, right?
Eric
That would be very kind of you. I am interested but don't do too much work on my account! Take your time.
Eric
Eric,
Not easy to find some synthetic material on the infraflex... similar to the DML thread...
Basically, it is based on a large square membrane (70 to 100cm) from EPS 20mm. There is a circular opening (90cm for a 100cm square) The membrane is clamped between 2 pieces of some wood (ie plywood). Some mention to have glued it; I was not able to understand if ti works or not.
The edge of the circular opening are rounded to avoid membrane stamping.
For the woofer, it is a high BL one for the efficiency. It is fully dismounted. the membrane is removed, the basket (frame) also so only something like an exciter remains.
A part to link the motor to the membrane is used. It might be a simple nylon threaded rod to a more sophisticated part with the shape of the central dome of the woofer, or a cylinder to extend the voice coil. in the original design is was a recycled jar of cottage cheese (thank you deepl...).
The motor is mounted at the center thanks to radial brackets with a sufficient inertia
The exchanges show about no measurements neither tentative of theoretical approach. First resonance at 20Hz, second at 80Hz on an impedance curve.
Few mentions of DML.
The comment about the quality of such subwoofer are complimentary. with a nature of sound different from a standard woofer. similar to my perception of DML. No rear to front acoustic short circuit.
Some posts mentioned a sensitivity to the room. Not clear.
In all of that, I am wondering what is the most favorable design to create the acoustically efficient modes... free edge, simply supported, clamped, rectangular, round...
Christian
Eric,
Below a test with the XPS 9mm 20x30cm panel
I don't know what to think about hte humps between 50 and 60Hz. In the FR, the 54Hz hump seems linked to the room. I have another FR where the panel was at the door window, the rear side firing to the garden; this hump is not so present.
As there is no hump in regard the other peaks, I understand that for the next modes, if they don't appear in the FR, it is because they have a low or even no contribution to the sound.
A tap test at 1/3 1/3 might be instructive ?
Christian
Below a test with the XPS 9mm 20x30cm panel
- green is the FR
- dark green is the tap test without the exciter
- purple is the tap test with the exciter
I don't know what to think about hte humps between 50 and 60Hz. In the FR, the 54Hz hump seems linked to the room. I have another FR where the panel was at the door window, the rear side firing to the garden; this hump is not so present.
As there is no hump in regard the other peaks, I understand that for the next modes, if they don't appear in the FR, it is because they have a low or even no contribution to the sound.
A tap test at 1/3 1/3 might be instructive ?
Christian
Hello NaRenaud
IET Impulse Excitation Technique is based on mode excitation. The sample plate has to be held at a node of the desired mode. In flexion, this node line is at 0.22 x plate side (see attachement)
As you can see in the posts here, I experienced the method. The math behind are not really difficult with Excel or equivalent, just applied what is on the net. The more tricky aspect for me is to get the culture of modes, in one sample modes had the same frequencies so not easy to understand what is what.
In the videos about guitar top, you'll see it is the node line used for the tap test.
Christian
Attachments
Hey, I just stumled across your thread. Very nice read! Thanks for the tip with Freecad. I absolutely overlooked that one.
I am suprised to see that basic modes are in the audible spectrum. From some rough calculations I expected them to be much lower. This changes a lot for creating a good sounding bass. All the tricks with weights and asymetric panels shouldn't work in that frequency range, at least with the panel size and thickness that most people are using.
Have you tried suspending your plates with long rubber bands? That should make them free Mindlin-Rissner plates as there is no more momentum transfer to the suspending structure no matter the position of the suspension points.
You were also talking about the question whether some modes don't produce any sound at all bc all the positive pressure zones would cancel with negative ones. I stubled across this sentence: "When the bending wavelength is longer than the wavelength in air, real sound power is radiated to air" (A. Dumčius, L. Bernatavičius, The Research of the DML Loudspeakers Propertie, check the google drive collection for it). My understanding is that when this condition is not met high and low pressure zones can compensate directly in front of the plate. So you don't even get sound in the near field. In the far field I would expect the soundwaves to cancel out even if the condition is met but since we get sound out of these speakers it obviously doesn't. Mb someone here has an explanation why the sound doesn't cancel out when facing perpendicular to the plate.
I am suprised to see that basic modes are in the audible spectrum. From some rough calculations I expected them to be much lower. This changes a lot for creating a good sounding bass. All the tricks with weights and asymetric panels shouldn't work in that frequency range, at least with the panel size and thickness that most people are using.
Have you tried suspending your plates with long rubber bands? That should make them free Mindlin-Rissner plates as there is no more momentum transfer to the suspending structure no matter the position of the suspension points.
You were also talking about the question whether some modes don't produce any sound at all bc all the positive pressure zones would cancel with negative ones. I stubled across this sentence: "When the bending wavelength is longer than the wavelength in air, real sound power is radiated to air" (A. Dumčius, L. Bernatavičius, The Research of the DML Loudspeakers Propertie, check the google drive collection for it). My understanding is that when this condition is not met high and low pressure zones can compensate directly in front of the plate. So you don't even get sound in the near field. In the far field I would expect the soundwaves to cancel out even if the condition is met but since we get sound out of these speakers it obviously doesn't. Mb someone here has an explanation why the sound doesn't cancel out when facing perpendicular to the plate.
I modeled my current speaker which is a 1200mm*600mm*20mm xps board with rounded corners in Inventor and ran a mode analysis. I also recorded a frequency sweep with the exciter attatched at the common 3/5th position.
Now the great guessing game starts which one of the peaks resembles the base mode. Unfortunately I don't have the material constants for my specific xps plate, So I just used a rough estimate with values I could find online. Now comparing the simulated modes with the responsecurve I think the base mode is the 35.4 Hz one because the frequency ratio with the next one is very similar to the frequency ratio for the simulated modes (1.29 measurment 1.23 simulation). But when looking further up the frequency range there is a huge gap to the next simulated mode. The response curve does not feature such a gap at all. I also found a paper which simulated the the first couple of modes of a free mindlin plate: https://www.hindawi.com/journals/sv/2018/8562079/
They have a similar gap between the 2nd and 3rd mode, with the 3rd and 4th super close together again. Take note the paper has a rectangular shape, my board has a 2 by 1 aspect ratio.
So the question is what the other peaks are. I will definetly do the impulse excitation test you did as well. This should tell us if the other peaks are due to the exciter (impedance, god knows what...).
Now the great guessing game starts which one of the peaks resembles the base mode. Unfortunately I don't have the material constants for my specific xps plate, So I just used a rough estimate with values I could find online. Now comparing the simulated modes with the responsecurve I think the base mode is the 35.4 Hz one because the frequency ratio with the next one is very similar to the frequency ratio for the simulated modes (1.29 measurment 1.23 simulation). But when looking further up the frequency range there is a huge gap to the next simulated mode. The response curve does not feature such a gap at all. I also found a paper which simulated the the first couple of modes of a free mindlin plate: https://www.hindawi.com/journals/sv/2018/8562079/
They have a similar gap between the 2nd and 3rd mode, with the 3rd and 4th super close together again. Take note the paper has a rectangular shape, my board has a 2 by 1 aspect ratio.
So the question is what the other peaks are. I will definetly do the impulse excitation test you did as well. This should tell us if the other peaks are due to the exciter (impedance, god knows what...).
Attachments
Hello Harzer99
About Freecad, I haven't really pushed further for now... it is a bit aside my knowledge area but the main reason is 2 aspects are missing : for low mass membran, I think the exciter changes the behavior by its mass and stiffness and also a calculation of the SPL might be a plus. So good tool for some 3D drawing but for DML?
Different types of suspensions have been tried. The best low freq response seems given by some surrounding foam.
I have read also about the role of the wavelenght in the materail compare to the air but for now, I don't know what to do with it... You can make measurements at short disdances. If I remember, the FR is much flat so the hypothesis of areas cancelling at distance make sense.
Christian