I remember now it is a topic I detected "something" long time ago without conclusion except "the mono exciter conservative way". The picture now is done. Good. On the multi exciter configuration, Steve pointed long time ago also the low frequency coupling. I suspect also some possible complication because of the impedance of each exciter when there are in series.
Christian
You describe them as if the there is a sequence in time in which the PIP happens first, then DML. And that is correct in a way, but I'm not sure if it's meaningful way. That is, if you imagine that there is no signal to the exciter, and then instantaneously a signal to the exciter, then yes there is a short period of time when there is PIP response of the panel, and which lasts approximately until the first reflections from the panel perimeter make it back to the exciter, after which it arguably becomes "DML action". That time period will vary somewhat depending on the panel material, its size, and the frequency of the signal, but generally it's likely to be well under 10 ms. As an example, I estimated that for one of my cf/balsa panels at 500 Hz, the time period would be about 2 ms. Is that period long enough for our auditory system to perceive it? I don't know, but I'm skeptical that it is.
Also, the signal to the exciter doesn't actually start suddenly from silence. Except at the start of any track, there is always some earlier signal to the exciter, and hence, residual DML action, in the panel. So while it may take 10ms (or 2ms) for the DML action to catch up to the signal at the exciter, at any arbitrary instant the panel already has residual DML action which is the result of the previous exciter signal. And if the damping is low, those reverberations can easily last 100 or even 200ms or more after the original signal. So I would say that there is always "DML action", it's just a question of how far back in time it comes from.
My view is that the transition from DML action to PIP action is not so much one that happens as a function of time, but rather happens as a function of frequency. That is, at low enough frequencies, where the panel velocity profile looks like the top two images in the figure below (from one of Dave's publications), the panel behavior is "DML", while at higher frequencies, where the panel displacement becomes localized around the exciter, (see lower right image at 10k), that panel's behavior is PIP. Of course, at intermediate frequencies (like 2k below) it is a combination of both.
I don't really think that is possible. My understanding is basically that increasing damping mainly acts to shift the transition from DML behavior to PIP behavior to lower frequencies. This is shown in the image from Zenker, which compares the behavior of a panel with low vs. high damping values. With low damping, the behavior is "standing wave" (i.e. DML), while with high damping the behavior at the same frequency is "traveling wave" (i.e. PIP).
So I don't really think you can get "large quantities" of PIP without also having high damping.
Eric
Thanks a lot for your valuable feedback and suggestions. I will ask a friend for help, testing the panel / taking measurements. Initially the plan was to use 3.2mm thick plywood for this panel since I can get larger sheets (24"x36"). Thinner plywood (2mm or 1.58mm) I can get only in smaller sheets sizes (12"x36"max.) If you feel that 2mm might be already a bit to thick, I could go for 1.58mm instead, but in that case I would need to downscale my panel size to 490x300mm if I want to keep the same proportions and take the max out of that plywood sheet. What do you think? Should I stay with my initial plan or would you downscale and go for the thinner ply? Since I haven't got any experience in building DML's, I will gratefully follow your advise. shido
Thanks Steve. I will share my process and results here soon.
Thank you for your valuable feedback, to which I will respond in detail later...need to go to work....have a great day !Shido,
This sounds like a neat idea. I like your honeycomb composite panel idea. I considered doing something similar at one time but never actually did it. I found this drawing I made (in January 2020), which looks similar to yours! I can't recall for sure but I think my idea at the time was to use wood veneer for the face, rather than fiberglass, but the concept is the same.
View attachment 1426336
I do agree with Christian and Steve with respect to the basic idea of testing some prototypes before going all-in on a particular design. Start with a single panel in a crude frame and see how that sounds, and then proceed from there.
I also agree with Christian concerning his "coincidence frequency" comments, but to be honest I would not concern myself too much about that if I were you. It is what I consider a second or third order effect, to be concerned about after you have sorted out more basic things like frequency response and efficiency, Maybe worry about that on your tenth design!
Concerning the composite panel itself: Have you made fiberglass composites before? In your description, you refer to "adhering" the fiberglass fabric with PU. What is not clear is if you understand that in order to work properly the fiberglass fabric itself has to be completely saturated with resin (PU or epoxy). I have only experience with epoxy resins, but I understand that there are PU resins that are also made for this. If you are not familiar with composites, there are great videos here:
https://www.easycomposites.co.uk/learning
Also, I would suggest making the the honeycomb openings as small as possible. It looks to me like yours are about 15 mm wide. I fear that is too long a distance to span with a thin fiberglass skin. When the fiberglass is on the compression side of a bending wave, it may buckle, and not provide the stiffening that you want. The openings in aramid honeycombs for composites are typically on the order of 5 mm. You may want to get as close to that as is practical.
Concerning the mounting of your panel to the frame, I would suggest that you plan to try many different configurations, rather than simply planning to mount at the two corners you have chosen. I think more mounting points will work better, and likely the best is attaching the panel to the frame around the entire perimeter using a soft foam or double sided mounting tape. The main benefit of attaching the panel around the entire perimeter is an increase in SPL at lower frequencies. I have done many test which show this, including these two:
But as I said, you should plan to test different configurations and decide for yourself what you like best. Good luck and keep us posted on your progress.
Eric
I have no experience to help you predicting the properties (bending stiffness, areal mass, coincidence frequency) of this composite. I wanted just to warn you about a possible too stiff (for its weight) material. Maybe start by making one sample to test your process and the panel. Measurements are of great help to focus on one aspect, one property of the panel; listening to it even suspended by a simple tape ribbon, in mono, (beware of the parasitic noise) help to get the overall picture.
Christian
The pebble in the pound...
I had under the dust of my hard drive a Python script written beginning of last year to simulate the transient of a plate under an impulse... The exciter is at 2/5 of a rectangular plate simply supported. The test case is not very realistic as there is no damping, the exciter is a point, the panel is force driven (it is not an exciter driven by a voltage)... but it is encouraging. See it as a curiosity.
It is an animated gif file. It seems accepted as format. Maybe need to be downloaded? The exciter is at the red dot.
Christian
PS : just click on it and you should see the (animated) pound!... A bit long, the animation goes up to 8ms...
I had under the dust of my hard drive a Python script written beginning of last year to simulate the transient of a plate under an impulse... The exciter is at 2/5 of a rectangular plate simply supported. The test case is not very realistic as there is no damping, the exciter is a point, the panel is force driven (it is not an exciter driven by a voltage)... but it is encouraging. See it as a curiosity.
It is an animated gif file. It seems accepted as format. Maybe need to be downloaded? The exciter is at the red dot.
Christian
PS : just click on it and you should see the (animated) pound!... A bit long, the animation goes up to 8ms...
Attachments
Last edited:
Thanks for your feedback! I will order panels in different sizes and thickness and do some experiments. I changed my CAD file and reduced the size of the hexagons to 5mm (side to side) and hope that a wall thickness of 1.5mm is save to cut without creating a fire....I haven't worked with fiberglass or carbon skins yet but the videos of the the easy composite website you were sharing are showing pretty much what I had in mind as well. I do not have a vacuum pump with a bag at hand but I should manage to press a panel in that size manually with evenly distributed weights on a glass plate. I do have a pneumatic heat press (20"x15") which I can use for the smaller panels. The idea is apply an even coat of PU onto the glass plate (which has been treated with a release agent) then to place the fabric onto the plate and adding more PU. With a squeegee I will remove the excess of the PU and assure that they are no air pockets by looking at the glass from beneath. If that is all good I will place the honey comb ply panel on top and press it with thicker, heavier plywood sheets and weights and for the smaller panel doing the same process but just using with my heat press. After curing I will repeat the steps for the other side as well.Shido,
This sounds like a neat idea. I like your honeycomb composite panel idea. I considered doing something similar at one time but never actually did it. I found this drawing I made (in January 2020), which looks similar to yours! I can't recall for sure but I think my idea at the time was to use wood veneer for the face, rather than fiberglass, but the concept is the same.
View attachment 1426336
I do agree with Christian and Steve with respect to the basic idea of testing some prototypes before going all-in on a particular design. Start with a single panel in a crude frame and see how that sounds, and then proceed from there.
I also agree with Christian concerning his "coincidence frequency" comments, but to be honest I would not concern myself too much about that if I were you. It is what I consider a second or third order effect, to be concerned about after you have sorted out more basic things like frequency response and efficiency, Maybe worry about that on your tenth design!
Concerning the composite panel itself: Have you made fiberglass composites before? In your description, you refer to "adhering" the fiberglass fabric with PU. What is not clear is if you understand that in order to work properly the fiberglass fabric itself has to be completely saturated with resin (PU or epoxy). I have only experience with epoxy resins, but I understand that there are PU resins that are also made for this. If you are not familiar with composites, there are great videos here:
https://www.easycomposites.co.uk/learning
Also, I would suggest making the the honeycomb openings as small as possible. It looks to me like yours are about 15 mm wide. I fear that is too long a distance to span with a thin fiberglass skin. When the fiberglass is on the compression side of a bending wave, it may buckle, and not provide the stiffening that you want. The openings in aramid honeycombs for composites are typically on the order of 5 mm. You may want to get as close to that as is practical.
Concerning the mounting of your panel to the frame, I would suggest that you plan to try many different configurations, rather than simply planning to mount at the two corners you have chosen. I think more mounting points will work better, and likely the best is attaching the panel to the frame around the entire perimeter using a soft foam or double sided mounting tape. The main benefit of attaching the panel around the entire perimeter is an increase in SPL at lower frequencies. I have done many test which show this, including these two:
But as I said, you should plan to test different configurations and decide for yourself what you like best. Good luck and keep us posted on your progress.
Eric
You are absolutely right about the mounting of the panel and thanks for sharing your findings. I will do different tests and will share my results here. shido
Christian,
That is very cool! I did some modeling of the transient earlier this year using the "dynamic response" mode of LISA. Also without damping. But I was only able to do a beam model. A plate required too many elements and the model couldn't handle it. Also, I never figured out a way to share the ouput.
What was the nature of your impulse? Could you drive it with a sinusoidal force?
What I did with my beam model was to drive it with a sinusoidal point force at various frequencies and driving points. It was neat to see how the behavior differed when the driving force was at a natural frequency and when it was not. It would be a good test of your model to drive it at some of the natural frequencies calculated by theory (or model) and see if the theroetical modal shapes start to develop. Also, it would be interesting to see how long it really takes to go from PIP to modal.
Eric
It is a very short half sinus (50µs). My idea is to input something as close as possible of a very short pulse (Dirac?) in order to get an impulse response and then use an FFT.
Seems possible as the driven force is described over the time. It is an iterative simulation based on time. A variant of the finite difference method we shared some months (more than one year!) ago with Paul.
Christian
Thank you for posting these two pictures.
They clearly show dml action on the left picture, but on the right the dml action has been damped or killed by the damping.
You can clearly see the pistonic pulse in the centre of the exciter area, then the waves moving out from this.
But even the waves are being damped by the damping.
By doing this, You have created a bending wave panel with pistonic action in its central excitation area.
Not a dml panel, which has to have all three , to operate properly without constraints( if using an exciter).
How much of pistonic, bending wave, and dml is produced by the panel can vary depending on the panel material .
A light bending panel will produce a larger pulse in the cone area than a very heavy rigid wood panel
So when I say I like to have large quantities of all three, I mean ideally a light flexible panel without damping.
Steve.
That's what I was assuming. Great work.
Christian, I seriously think this could be a really valuable tool for visualization of the panel motions, including transients. Of course it's easy for me to say (since you have to do the work!) but if you can solve the equations without damping, you can add damping without too much trouble, right? You are already solving the equation below using the first two terms, if you can do that, adding the "b" term is pretty trivial, isn't it? (Not for me, but for you!).
I think it would be very interesting and informative to see what several cycles (say 10?) at various frequencies across the audible range would look like in your simulation, especially if you could include damping.
I did that with my LISA beam dynamic model, and it was pretty interesting, but very limited because I could only do a beam, not a plate, and could not include damping. I'll share some output from that model in a second post, if I am able to figure out how.
I'd love to see you do more with that transient model. Of course, modeling how the plate moves is only half of the full simulation, but half is still a lot.
Eric
The first action on a panel using an exciter is the pistonic action not the PIP.
Without the pistonic action and the PIP, there can not be any dml action.
So the PIP is very important, crucial even, no matter how miniscule the time window.
Are you saying that the next note on a musical instrument will not be hear because the panel is already vibrating.
We already know this is not the case.
Steve.
Christian,
I couldn't upload a movie of my LISA beam transient simulation, but here are a few still pics from it from start to end.
The "panel" has properties similar to plywood, 25 cm long and 5 mm thick on simple supports. In this case it's driven at the center by a point force at 11.55kHz, which is natural frequency of the 11th mode. It still has PIP-like response in the second image below at about 0.14 microsec, but by 0.26 microsec it already has the mode shape established, which just grows larger in amplitude after that (as you would expect at a natural frequency).
It would be so neat to see you plate version, in motion, in a similar simulation.
Eric
I couldn't upload a movie of my LISA beam transient simulation, but here are a few still pics from it from start to end.
The "panel" has properties similar to plywood, 25 cm long and 5 mm thick on simple supports. In this case it's driven at the center by a point force at 11.55kHz, which is natural frequency of the 11th mode. It still has PIP-like response in the second image below at about 0.14 microsec, but by 0.26 microsec it already has the mode shape established, which just grows larger in amplitude after that (as you would expect at a natural frequency).
It would be so neat to see you plate version, in motion, in a similar simulation.
Eric
I think you have forgotten the video I posted on this forum some time ago.
Showing two exciters as close together as they possibly can be.
One is stationery and the other is jumping up and down like a yo-yo.
They are not acting as one exciter, they are impinge on each other, in opposition.
In your first link the two exciters mounted close to each other , clearl shows the two exciters interfering with each other.
The picture of the two stones in a large pond is basically the picture I was going to draw for you and Christian.
That will save me a drawing.
The interference between the two point sources caused by the comb filtering ,clouds the sound compared to a single driver, which does not suffer from this problem.
I sometimes use weights to change the frequencies on a panel, which also alters the sound of the panels .
You must remember that an added exciter is also another weight, which is also randomly placed.
Not only that , but the exciter is a very large weight that also vibrates!
There is a cloudiness and lacking of detail using two exciters on the same side of a panel.
I got around this by having them on opposite sides of the panel and mounted in push pull mode.
But then there is the problem of the exciter body in the way of the sound 🙄
But should be fine for lower mid to low frequency use.
Steve.
Killed is a harsh word. Is it killed? Or mitigated? Or controlled? Or even remedied?
That would indeed be the point.
I don't really care what it's called, only that it reproduces recorded music accurately.
If you are saying that the pistonic motion of the exciter initiates the bending motion of the panel, I agree. But in the region of the exciter, the panel itself is bending as much or more than in any other region. So I think it's wrong to say that the panel's action is pistonic there. Pistonic means "moving as a rigid body" (i.e. without bending). The exciter itself is arguably pistonic, but the panel itself is certainly bending, and clearly not pistonic.
Not that it won't be heard, but rather that it can be muddied by the overlap of the old sounds with the new. That is, unless the resonances are adequately controlled.
Eric
@EarthTonesElectronics
Dave,
Both your model and laser vibrometer results show that the panel velocity profile becomes localized at high frequencies. It seems to me that there could be a couple of different mechanisms which could all result in that localization:
One is the effect shown in Ben Zenker's paper, when the damping is high enough so that a traveling wave pattern is set up, instead of a standing wave pattern.
As second possibility, however, could arise from the fact that (even without high damping), at high frequencies multiple modes are excited. And all those excited modes have antinodes at or near the exciter location. So near the exciter, all those modes will be in phase with the exciter's motion, and have relatively large velocity there. But once you get far enough away from the exciter, it seems likely that the velocity will tend toward zero, simply because some of the excited modes will be "in-phase" with the exciter, and other will be "out-of-phase" at any particular location far enough from the exciter. And hence, the average velocity in this case would also diminish as you move farther away from the exciter. I suspect this is what you consider to be the "statistical" region, If I am not mistaken.
And finally the third reason is the "drum" effect caused by the exciter's diameter.
But considering the first two cases. Is there a difference in the sound radiation characteristics of the two? Do the two cases even exist in reality? And if so, does your model "show" that difference in any way?
Thanks,
Eric
Dave,
Both your model and laser vibrometer results show that the panel velocity profile becomes localized at high frequencies. It seems to me that there could be a couple of different mechanisms which could all result in that localization:
One is the effect shown in Ben Zenker's paper, when the damping is high enough so that a traveling wave pattern is set up, instead of a standing wave pattern.
As second possibility, however, could arise from the fact that (even without high damping), at high frequencies multiple modes are excited. And all those excited modes have antinodes at or near the exciter location. So near the exciter, all those modes will be in phase with the exciter's motion, and have relatively large velocity there. But once you get far enough away from the exciter, it seems likely that the velocity will tend toward zero, simply because some of the excited modes will be "in-phase" with the exciter, and other will be "out-of-phase" at any particular location far enough from the exciter. And hence, the average velocity in this case would also diminish as you move farther away from the exciter. I suspect this is what you consider to be the "statistical" region, If I am not mistaken.
And finally the third reason is the "drum" effect caused by the exciter's diameter.
But considering the first two cases. Is there a difference in the sound radiation characteristics of the two? Do the two cases even exist in reality? And if so, does your model "show" that difference in any way?
Thanks,
Eric
Well, because my simulation is a modal one, this is in fact exactly what's happening. I don't ever simulate traveling/evanescent waves on the surface - the localized pattern arises from the constructive overlap of many modes being actuated in that particular way. For those familiar with DSP theory, the wider bandwidth something gets (in this case, bandwidth refers to k-space, or wavenumber space), the narrower it gets in the transform domain (e.g. the actual vibrational pattern). So the more modes that are actuated at once, the narrower the vibrational shape becomes, and at high frequencies many modes are being actuated at once.
I'm happy to say that I've made the free version of PETTaLS available here: https://sites.google.com/d.umn.edu/profdaveanderson/pettals
You'll have to request access for the download - this is mostly so that we can be made aware if any companies are interested in the software, and the University can follow up with them. You'll also have to download the matlab runtime environment as part of the installation process. This installer also only works on Windows right now.
The free version only has three exciter types (point source, DAEX19CT-4, and DAEX25VT-4). Someone asked me about Xcite models, and I'm planning to get to those (and others) eventually - right now we have a giant pile of Dayton exciters that we're working through measuring and verifying.
On that note, I'm hoping to also release the expanded version soon that'll include a ton of different exciter models, the ability to simulate multiple exciters independently on the same panel, the model for sandwich boards (though it's probably still too simple for Eric), additional boundary conditions, stands, movable weights, etc.
I'll make some videos about this version at some point - one thing to be aware of is that I have the program default to setting the bandwidth at only 2kHz after most changes. It's possible to set the simulation to a very large, very thin panel and at a 20kHz bandwidth this can take hours to run. It's best to step up the bandwidth from 2kHz and make sure that the full 20kHz bandwidth isn't going to take forever.
Steve,
No I have not forgotten it! That video is indeed a curious puzzle!
There is no doubt that multiple exciters will act in opposition at some frequencies and in concert at others. And it is possible that those interactions can result in a net negative, or or net positive, effect on the speaker, depending on the execution and goals.
But the video you refer to is indeed very interesting and curious. As I recall, it shows two apparently identical exciters, placed very close together on the same panel. The speaker plays a familiar song (Wicked Game?). At certain instants, one exciter bounces like crazy, while the other remains relatively motionless. As I recall, only one ever bounces, and never the other, but please correct me if I'm wrong.
Exciters only bounce like that when they are driven at really low frequencies. And they bounce at nearly the same frequency regardless of whether the panel itself is moving much or not. My best guess is that either (a) one of those drivers was getting a filtered signal (with no low frequency), or (b) no signal at all.
Eric
Yes, exactly this equation with the "b" term. It seems even possible to include a viscous damping (to take into account a variation with the frequency).You are already solving the equation below using the first two terms, if you can do that, adding the "b" term is pretty trivial, isn't it? (Not for me, but for you!).
Progress might be possible in the 2 next weeks.
It is necessary to include it otherwise the simulation is divergent (infinite displacement at the modes and nothing out of the modes).
What would be great is to find the relation between b, Q used by Dave and the "nu" in % used by B Zenker to link all those simulations.