It's supposed to represent the pressure at 1m, facing the center of the speaker. Rotating the speaker by 90 degrees (switching Lx and Ly) is essentially the same as just rotating the microphone by 90 degrees at this position, hence why the on-axis measurements stay the same. It's a 'slice' through the center of that polar plot.
Is this what you were thinking it was, or something different?
Sorry, yeah, there was a bug in that part temporarily. I've fixed it and a new version is up, if you don't mind the hassle of getting the new(est) version! I also rewrote the part that plots the surface modal patterns so that it's easier to rotate and so that the colors are always scaled to the current pattern. I was previously scaling the colors to the maximum velocity over the whole bandwidth, but this usually just made everything blue other than the highest velocity mode. I also fixed the impulse response interpolation so that those high frequency artifacts aren't there anymore (mostly, although it's impossible to eliminate everything).
Also sorry if I can't keep up with all of the comments, I'll try my best to get to them as soon as I can - I'm most concerned with fixing UI bugs in pettals right now, and some of the technical questions require me to go look things up in my thesis or a textbook... which takes time!
No worries. Personally, I'd rather you focus on Pettals! But that won't stop me from asking.
Eric
I decided to look at the effect of Young's Modulus (E) and density (rho) on DML sensitivity, using the Pettals model.
Christian and I have many times mentioned that several sources noted that sensitivity should depend strongly on the factor E/(rho^3).
The Pettals simulation is consistant with that, based on the results below.
I modeled three panels with hypothetical properties. For the first I used E=10 GPa and rho=1000 kg/m^3. For the second I halved both E and rho, and for the third, I halved them again. That way, the ratio E/rho was constant, so the shape of the frequency response should stay the same, and clearly it did.
But, also, by halving both properties, the ratio E/(rho^3) increased by a factor of four. And for each step, the predicted sensitivity increased by 6 dB. I'm pretty sure that is consistent with theory, though I must admit I'm not sure.
Christian, was the exact (theoretical) relationship in any of the papers?
Eric
Christian and I have many times mentioned that several sources noted that sensitivity should depend strongly on the factor E/(rho^3).
The Pettals simulation is consistant with that, based on the results below.
I modeled three panels with hypothetical properties. For the first I used E=10 GPa and rho=1000 kg/m^3. For the second I halved both E and rho, and for the third, I halved them again. That way, the ratio E/rho was constant, so the shape of the frequency response should stay the same, and clearly it did.
But, also, by halving both properties, the ratio E/(rho^3) increased by a factor of four. And for each step, the predicted sensitivity increased by 6 dB. I'm pretty sure that is consistent with theory, though I must admit I'm not sure.
Christian, was the exact (theoretical) relationship in any of the papers?
Eric
Hi Todd --I guess it is only a hijack if no-one or very few have anything of note to add on different topics - the thread has been very quiet of late (apart from out now departed member) - I for one would appreciate more involvement and topics of discussion - we're just trying to get to the end result in the best way possible - PeTTaLS is a tool which was hitherto unavailable, so it deserves the attention it is receiving IMO, and many thanks to Dave for making it so.
Regards
Eucy
@EarthTonesElectronics
Hi Dave - V1.2.1 - A basic run with CCCC, 700x300x3 ply again with Q of 1 yields the following FR's :
Left to Right: On-axis, +90,-90-, Av, Angular
Given the even angular spread and the on-axis/+90/-90 graphs, how do you reconcile the dips in the Av graph? - I suspect there may be a bug lurking in there somewhere, which may also explain the dips I queried in the FR graph I posted previously??
Eucy
Hi Dave - V1.2.1 - A basic run with CCCC, 700x300x3 ply again with Q of 1 yields the following FR's :
Left to Right: On-axis, +90,-90-, Av, Angular
Given the even angular spread and the on-axis/+90/-90 graphs, how do you reconcile the dips in the Av graph? - I suspect there may be a bug lurking in there somewhere, which may also explain the dips I queried in the FR graph I posted previously??
Eucy
Last edited:
Accepting PeTTaLS is accurate, low frequency sensitivity is driven by E, and mid to high frequency sensitivity is driven by rho - see below:
Left to right: CCCC case - 700x300x3 panel - Q1 used to eliminate 'noise'
(1) Reducing E increases low freq range, no significant effect above 100Hz (for the chosen panel).
(2) Increasing density reduces mid to high level output in a non-linear manner
(3) As for (2) above - pale blue line overlays a SSSS case on the orange CCCC - indicates constraints principally affect low frequencies only.
(4) As for (3) above, red line overlays Q5 SSSS case on the pale blue SSSS Q1
Eucy
Dave did say at the beginning of this discussion ,should he start a new forum.
I see no problem with showing measurements of panels, but the continuing updates could go on for some time.
It is a highly technical discussion of pettals design and should probably have its own forum.
Steve.
You are right that most probably we are going to several (not say many) iterations either in the tool itself or in exchanges more or less technical (probably more than less with some math...) to use correctly the tool.
A possibility would be to move this discussion to a new thread or to reuse the one we had @Veleric about simulation.
The risk being a disconnection or the need to follow 2 threads. For sure, I will be in both.
Christian
Hello Eric,
When searching something, you can have a look to my Github page where I store some papers about DML.
In this page there is what I wrote about efficiency.
The sources we had at this time were a thesis about piano soundboards and Heron's patent
It is from that, after some experiments, I made an heuristic
Using the heuristic, the efficiency with your figures are respectively 82.6, 85.6, 88.6dB so by 3dB steps.
One difference is PETTaLS FR is given in a half space (infinite baffle DML) while the heuristic comes from open back panels in a room.
I don't know if it explains the difference.
Christian
Hello Eric,
I can't answer for now to your question... It needs more investigations. I just started yesterday digging this topic...
I don't think there is a "simple" extraction of the Q possible directly from the impedance curve. I had a tentative to extract something from the equations but it needs too many simplifications to lead to something.
This is the translation of the fact the impedance curve shows the panel at the exciter terminal so adds on the panel characteristics the exciter ones. In addition, the modes are seen differently depending on the exciter location compare to the anti nodes location of the mode.
Nevertheless, there is maybe a way based on the fact that for each mode with its anti node located at the exciter axis, the exciter "sees" fully the mode. In addition, around the resonance frequency of each mode, the panel can be modeled by a mass, a stiffness and a damping (the components at the origin of the second order response in PETTaLS for which we are looking for the Q). So locally around one mode fully driven, a DML behaves from the impedance curve view like a standard loudspeaker so the classical method to extract the parameters of a loudspeaker should work. Compared to a classical loudspeaker, there is probably an additional effort to have to separate the panel characteristics from the exciter one... at least in a first time as I have no idea of their respective values.
So my proposal would be to use the technique of the added mass. 2 sets of 2 measures each should be needed :
- exciter alone, exciter with an additional mass to get the moving mass, the stiffness and the damping of the exciter
- exciter on the panel and exciter on the panel plus an additional mass to get mass, stiffness and damping of exciter + panel.
I think a simple step up is to drive a sample panel in free edge condition with the exciter at its center. To do it then with an additional mass at the exciter position in order to create a frequency shift. Then for the modes which are directly driven, extract the information. Easy! Maybe?
Where you could help Eric, it is by showing on such an impedance curve which are the peaks of impedance related to directly driven modes (=exciter at the anti node).
Christian
I can't be the only one that finds it ironic. After 10 years and nearly 700 pages of the most wide ranging discussion I have ever seen in any single DIY Audio thread, along comes a thing that is clearly the most relevant development in the DML area, that thing should have it's own thread? You can't make this **** up!
Eric
Eric
Apparently I was averaging both magnitude and phase, allowing for phase cancellation when you average over all the points. That doesn't really make sense, so I'll fix that in the next update so only magnitude is being averaged. Thanks for pointing this out!
Just thinking about this as a basic mass-spring system from physics, the displacement or velocity of the mass (analogous to rho for a plate) is inversely proportional to the mass, so it makes sense that halving the mass (rho) would increase sensitivity by 6dB. I wouldn't expect to see a 6dB increase in the acoustic FR, though, but rather in the average panel velocity. Not sure if there are any rules to predict the exact increase in SPL... I'll have to think about it.
I think you should be able to measure mode Qs by this method using impedance, and Christian is right that the mode peak amplitudes are highly dependent on exciter placement - but I'm pretty sure that the Q will stay the same even if the peak is lower in amplitude. 50% points are used to measure Q in a measurement proportional to power, and 70.7% points for a measurement proportional to the square root of power - impedance is proportional to velocity, and velocity is a square root term, so I think we should be using 70.7% points.
Haha, I've been advocating for a new thread for quite a while! It's up to everybody else, though. I certainly don't want to bog everybody down with complaints about how best I should be formatting graphs in my application - that's not relevant to DML design at all.
Christian and I never found a "rule" for SPL either. In Christian's white paper he referenced the Heron patent, which was the first place I ever saw the the "stiffness-to-mass cubed" ratio for radiation efficiency. (Heron refers to panel stiffness as B instead of D). I think that patent may have been the origin of the NXT work that was to follow.
Incidentally, I think Heron's speaker may have been the worst DML ever (intended entirely above the coincidence frequency!) But in his defense, it was intended for loudspeaker applications, not hifi. That said, I think it's a great patent because he describes various other "ratios" for use in panel speaker design. Dave you may find it interesting if you have not read it already.
https://patentimages.storage.googleapis.com/9a/43/32/b09245ff903cac/WO1992003024A1.pdf
Heron wrote:
Eric
From somewhere I had also been thinking like Dave, that the peak amplitudes would change as a function of exciter position but that the peak sharpness (Q) woould be the same. But I don't know where I got that idea from (maybe I made it up!) and I never really tested it to see if it was true.
Okay, but 70.7% from where? From 0 ohms, or from the baseline (i.e. nominal impedance) of the exciter? It seems like it should be from the baseline rather than zero, otherwise a peak that peaks at say 5 ohms would never fall to 70.7%, if the baseline impedance is 4 ohms.
All that said, I'm wondering now if it wouldn't be better to use the "tap test with close mic" method. In that test the y value is SPL, so I think I would be looking at the -3dB points, which should be less ambiguous. Also, that can be done or without an exciter mounted, and with or without the panel mounted in a frame. So it should be easy to sort out the the contributions of each. What do you guys think about that?
Tap Test Description and Examples
I have done tap tests before on panels with and without exciters mounted, and there is definitely a difference.
Eric
Having an interest in producing a DML Speaker to be used in a few different listening environments.
Firstly to fit into a very small listening Space and then the model produced to potentially become a Computer Speaker at another date.
At another date to have Panels produced to work in a room dimension of 6mtrs x 4mtrs and be aesthetically integrated into the Space, preferably a Large Panel or a Hybrid as recently witnessed in a few designs.
In relation to the latest info being supplied and observed by myself, where Graphs and Simulations are seen to be used, and suspected to be supplying data for working towards producing a betterment for the Exciter / Panel Interface, is quite advanced to many posts already read by myself.
Is there any discoveries made with these latest investigations being made, that is now considered when implemented an improvement on DML Panels already produced in previous build reports?
My own interpretation being a Bespoke Exciter might be the only way forward to control the energies that are being produced, with the outcome being a particular transfer of produced energy is the cause to raise a concern.
Firstly to fit into a very small listening Space and then the model produced to potentially become a Computer Speaker at another date.
At another date to have Panels produced to work in a room dimension of 6mtrs x 4mtrs and be aesthetically integrated into the Space, preferably a Large Panel or a Hybrid as recently witnessed in a few designs.
In relation to the latest info being supplied and observed by myself, where Graphs and Simulations are seen to be used, and suspected to be supplying data for working towards producing a betterment for the Exciter / Panel Interface, is quite advanced to many posts already read by myself.
Is there any discoveries made with these latest investigations being made, that is now considered when implemented an improvement on DML Panels already produced in previous build reports?
My own interpretation being a Bespoke Exciter might be the only way forward to control the energies that are being produced, with the outcome being a particular transfer of produced energy is the cause to raise a concern.
Hello JohnnoG,
I am going to speak from my point of view as we haven't tried to summarize the last findings in a form having a global agreement. Your question is the opportunity to start.
- Before building any panel, the test of the material simply suspended in directivity is a a way to discard materials with a too low coincidence frequency (below 10kHz? there is probably not yet an agreement on this threshold) or with too high SPL in angles above 40°. In the remaining candidates in what I have tested are the plywood below 3mm thickness maybe better is 2mm, the plain transparent polystyrene (probably too heavy) and surprisingly the Depron (XPS) 9mm.
- The spine can modify a lot the rear wave.It should obstruct the passage of the rear wave as little as possible (low thickess)
- The exciter itself is a mask for the rear wave . Any part that increase its surface will increase the problem.
- The spine and the exciter masks are important contributors to what could be called "rear noise". A third possible source is the cavity noise, the sound from the air trap in the voice coil area. More investigations are needed for it.
- The central area within the voice coil ring may be subjected to high frequency resonances. A disc of aluminum epoxy glued was a solution for plain polystyrene. I will test it for other material.
- The HF performance is linked to the diameter of the voice coil. Tests with a diameter reduction (concentrator) were done but they didn't show real improvement. It might be a certain balance between HF performance and a certain level of filtering is reach with a 20mm interface (the point is still open).
The possibility of a custom exciter is mentioned time to time but what are its specifications? Or in other words, in which way should it be different from the current ones?
Christian