Hello Dave,
My understanding is PETTaLS is based on GB Warburton's paper for the modes and mode shapes calculation. Is that correct?
Only the simply supported case has an analytical solution which includes in the paper. For the other boundary conditions, it is approximations.
In the paper square plates and some other proportions seem to need specific solutions.
Is all of that part of PETTaLs?
Christian
Eucy,
That's one area that I am very interested in, but have mostly have neglected so far. I think it is an untapped area for exploration. The challenge I see is to get a significant amount of damping without adding so much weight that your kill the sensitivity. I'd be looking at materials that provide constrained layer damping.
But at the same time I'm also wondering if there is really much value to putting the damping in (or on) the panel, if I know I can get very effective damping with good foams at the perimeter. I guess the question is this, is there anything to gain from free or nearly free panels that would eliminate the possibility of getting much edge damping, so then the only option would be to have it built into the panel? I wonder.
For fun I once made a panel with two layers of very thin plywood and with a layer of 3M Extreme double side tape between the plywood layers, on the entire interior surface. I think it took a whole roll of the tape. The impedance curve was flat as a pancake, and the FR was pretty flat too, but the sensitivity was abysmal, and the LF was gone.
Eric
Another deep topic to unravel in this seemingly endless quest.
I do remember saying 'no more' some time ago....
I was also considering a small gap between the panel and something like this stuff:
https://www.nanolayr.com/product/so...0IleRFI70y2njw3uUuzKu1YdUteoaPn4aAnW9EALw_wcB
Very esoteric and expensive no doubt... An alternative to try may be lightweight needle punched geotextiles, which have a furry structure and come in various thicknesses. Attach to the frame maybe 10mm away from the panel.
If it's light enough, it could also be directly attached either broadly or selectively.
Eucy
I do remember saying 'no more' some time ago....
I was also considering a small gap between the panel and something like this stuff:
https://www.nanolayr.com/product/so...0IleRFI70y2njw3uUuzKu1YdUteoaPn4aAnW9EALw_wcB
Very esoteric and expensive no doubt... An alternative to try may be lightweight needle punched geotextiles, which have a furry structure and come in various thicknesses. Attach to the frame maybe 10mm away from the panel.
If it's light enough, it could also be directly attached either broadly or selectively.
Eucy
Dave,
I would like to get back to the question of "pebble in pond" or PIP waves, because I still am not sure that I understand how, or even if, they are considered in Pettals. I don't love the PIP analogy, but at least I think you know the conversation I am referring to. To be clear, I'm not looking for a quick response. I just want to keep it "on the table" for discussion. I don't mind at all if you focus on Pettals first. Reply at your leisure!
I want to try to rephrase the question, because I'm not sure I have done a good job of that so far. Let me try this way, sorry if I am repeating myself:
Considering generally high frequencies, we know that the surface velocity profile in Pettals becomes concentrated around the exciter. If I understand correctly, this happens mainly (or entirely?) because at high frequencies all the driven modes have an antinode at or very near the exciter, and those antinodes move in phase with the exciter. Hence the net velocity there is high. Further, at any significant distance from the exciter, those same modes may be in-phase, or out-of-phase, with the exciter. Hence, as you move away from the exciter the net average velocity tends to fall. Is it correct so far?
But now consider separately the situation where the damping in the panel or the perimeter is strong enough that the reflections from the perimeters are so weak that modes don't really develop. The only bending waves in the panel propagate radially from the exciter and decay as they move away from the exciter. (This is what I mean by PIP waves). The waves in this case would have a crest that starts at the exciter and travels out radially and loses amplitude as it does so. These are clearly waves of a different sort than the former waves. However, even these latter waves would have a similar average velocity profile as the former ones. That is, both types would be centered on the exciter and decay in magnitude as they get farther away from the exciter.
All that said, I guess my questions are:
Eric
I would like to get back to the question of "pebble in pond" or PIP waves, because I still am not sure that I understand how, or even if, they are considered in Pettals. I don't love the PIP analogy, but at least I think you know the conversation I am referring to. To be clear, I'm not looking for a quick response. I just want to keep it "on the table" for discussion. I don't mind at all if you focus on Pettals first. Reply at your leisure!
I want to try to rephrase the question, because I'm not sure I have done a good job of that so far. Let me try this way, sorry if I am repeating myself:
Considering generally high frequencies, we know that the surface velocity profile in Pettals becomes concentrated around the exciter. If I understand correctly, this happens mainly (or entirely?) because at high frequencies all the driven modes have an antinode at or very near the exciter, and those antinodes move in phase with the exciter. Hence the net velocity there is high. Further, at any significant distance from the exciter, those same modes may be in-phase, or out-of-phase, with the exciter. Hence, as you move away from the exciter the net average velocity tends to fall. Is it correct so far?
But now consider separately the situation where the damping in the panel or the perimeter is strong enough that the reflections from the perimeters are so weak that modes don't really develop. The only bending waves in the panel propagate radially from the exciter and decay as they move away from the exciter. (This is what I mean by PIP waves). The waves in this case would have a crest that starts at the exciter and travels out radially and loses amplitude as it does so. These are clearly waves of a different sort than the former waves. However, even these latter waves would have a similar average velocity profile as the former ones. That is, both types would be centered on the exciter and decay in magnitude as they get farther away from the exciter.
All that said, I guess my questions are:
- Is this a reasonable description of the possibilities for bending waves of significance?
- Does Pettals consider both those types of waves, or only the first?
- Would the radiation pattern (i.e. beaming/diffusion pattern) of the two wave types be different?
- If there are waves only of the latter type, would the sensitivity be so low as to be impractical?
Eric
Hello Eric
Sorry if this interfere in the question, I hope not being wrong. I have already mentioned it. There are 2 possible representations of what happens. Thinking in steady state in the frequency domain and thinking in transient in the time domain.
My understanding is the first part of your description referring to modes is in the frequency domain while the second part with the PIP is in the time domain.
A physical system can be described in one domain or the other. There are tools to switch like the Fourier transform.
PETTaLS is a tool working in the frequency domain.
In his thesis, Dave explored the time representation from the frequency representation using the Fourier transform to get the transient (time representation). See §3.4 p34 and below from p42
In the time domain, you have a precise date of what happens but no knowledge about the frequency as all the frequencies (or at least a wide range) are used. This is for the impulse response (the stone). A single sinus of a specific frequency could be used. It is what you asked me... not done, the question of the Q calculation and behind the damping representation came on top.
In the frequency domain, you have a precise information of the frequency but not when something happens as the time is suppose to be infinite or long compare to the transient.
Mixing both has always messed my brain.
So the question might be rephrased in "what happens when the level of reflection of the edge changes (from fully reflective, to fully absorbent)?, from the frequency or the time observation.
Christian
@EarthTonesElectronics
Hello Dave
I just downloaded 1.3.
One remark about your web page : the link to PETTaLs at the beginning of the page goes directly to the .exe (so Windows) when the link below goes to Github with the 2 versions Linux and Windows. Both on the page are "Download PETTaLS free".
This is not my main point.
I am trying to make impedance measurements to get some material parameters like the quality factor. I tested an exciter DAEX25FHE and a piece of Depron (XPS), 6mm thick in A3 format.
In order to try to understand which mode is behind the impedance peaks, I ran a simulation.
The exciter is at 0,5 / 0,5. The strange point is PETTaLS shows asymmetric mode plots. Not sure it is important but I don't see physical reasons to that.
The simulation :
Some frequencies I picked from the peaks in the impedance plot : see the asymmetry 23 (at 0,0), 34, 44, 100Hz (at 0,2 horizontal) for example
Hello Dave
I just downloaded 1.3.
One remark about your web page : the link to PETTaLs at the beginning of the page goes directly to the .exe (so Windows) when the link below goes to Github with the 2 versions Linux and Windows. Both on the page are "Download PETTaLS free".
This is not my main point.
I am trying to make impedance measurements to get some material parameters like the quality factor. I tested an exciter DAEX25FHE and a piece of Depron (XPS), 6mm thick in A3 format.
In order to try to understand which mode is behind the impedance peaks, I ran a simulation.
The exciter is at 0,5 / 0,5. The strange point is PETTaLS shows asymmetric mode plots. Not sure it is important but I don't see physical reasons to that.
The simulation :
Some frequencies I picked from the peaks in the impedance plot : see the asymmetry 23 (at 0,0), 34, 44, 100Hz (at 0,2 horizontal) for example
Yes, I switched over to all equations from this paper for consistency. They seem to work very well. My reading of the 'extra modes' section for square and nearly square plates is that this is descriptive rather than prescriptive, e.g. the square plates have interesting modal patterns that emerge from the overlap of multiple actual modes which have node lines that are not parallel to the panel edges.
For example, here are some of those FFFF special modes in pettals with different aspect ratios and driving points:
Eric - everything you have discussed is all the same type of wave (propagating waves). All of these waves can be described either in the time domain or the mode domain. As you point out, though, there are really 3 regimes, and I've plotted an example of each along with its k-space transform (showing the magnitude of each mode that makes up the pattern):
- Modal: where the excitation point is not apparent, vibrational pattern is dominated by a single mode or very few modes.
- Localized (maybe): Vibrational pattern is distributed across the panel but is highest in amplitude around the excitation point.
- Pebble in Pond: Vibrational pattern is not distributed across the panel, is only localized around the exciter and decays quickly.
This textbook is a great reference on relating the time domain waveforms, k-space representations, and acoustic radiation. Unfortunately to answer all of your questions fully would essentially take up an entire course that covers this textbook!
The one difference where the two representations may not be equivalent is when there's some kind of nonlinearity or spatial dependence to the wave propagation.
Last edited:
Oh, I didn't even realize that there was a link at the top of the page. I took that out. Thanks!
This effect comes from the discretization of the exciter shape (even the point force), which is almost always coming out to be a little bit lopsided based on whether or not certain points in the grid are counted as being part of the exciter shape or not. I'm still going to play around with finding a solution to this, but it shouldn't impact the results much.
And, last thing for today... it's better to think about the radiation pattern in terms of what's happening in the k-space (spatial frequency domain) rather than the actual waveform on the panel. And that totally depends on how each k (mode) radiates acoustically. The main difference is probably that the localized type of pattern has fewer k's, and the PIP waveform always has many more k's that smooth out its radiation pattern. This is, of course, ignoring things like coincidence frequency that also impact the radiation!
Thank you Dave for the confirmation of the use of the Warburton's formula.
How to explain the asymmetry I got? Note it happened on rather low frequencies.
Christian
PS : Ignore the question above... Sorry I have seen after your post 51.
What is the grid step?
Last edited:
Dave,
Thanks, this is a really great explanation. It clarifies a lot of things for me. It also reminds me that I have to spend some time undestanding K-space better.
Eric
No worries Christian, you know I always appreciate your perspective.
Eric
I'm doing 250x250 for discretizing the exciter shape. This is scaled to the size of the panel, so larger panels have a more coarse spacing (only for exciter shape, not for mode shape). Now that you mention it, I should play around with this grid spacing and see how impactful it is...
Just uploaded v1.4 to github with these changes:
- Added a smoothing option (unsmoothed, 1/6 octave, 1/3 octave, 1 octave). Smoothing is only applied to the SPL graph, not to the polar plot or the impulse response. The "compare results" checkbox allows you to keep previous results when plotting new smoothed results.
- The "compare results" checkbox now affects the new SPL graph, so you can compare on-axis to off-axis, etc.
- Fixed a bug in the code to implement the equations for FF mode shapes with odd indices.
I loaded v1.4, but it won't run at all for me. It loaded fine (or so it seemed), but nothing happens when I press "run". I tried rebooting the PC but no change. I can make entries to the fields but that's it. I can't run a simulation.
Can I get back 1.3 temporarily?
Eric
Yeah, weird, it worked fine on my development desktop when I originally compiled the binary. After you said this I tried it on a different computer and it didn't work, like you said. I guess I need to check it on multiple computers before I upload it in the future!
I recompiled the binary and it's working for me on both computers now. I've reuploaded v1.4 to github, but I also put v1.3 up there in case there are still issues.
Last edited:
Should there be a different (i.e. github) link on the first post? What is it? Does the link there now not work?
Eric