Thanks for the links. They give links to download pdfs in Japanese. The last link, JP2018121178A downloaded Japanese pdf gives some drawings to ponder about from page 7 onwards. JP2018121178A pdf. You've to read them side by side. I think the diaphragm can be anything. People here had been testing all kinds of material to get the best out of the transducer. Most probably, what we'd have to do is to cut out the legs of the Dayton "frog", or similar and stick a pencil or a nail in the middle of the transducer to check the point transfer of vibration (music) to the flat panel. There's a lot of ideas in the first few pages of this thread, by the way.
Last edited:
Christian.
I did take some measurements of my proplex panel and canvas panel from the side to show the sort of response expected in a room.
Similar to the burntcoil setup.
But have now forgotten which pictures are which🙄
So if you wish I can do them again ,but at night, with less traffic noise muddying up the LF.
Steve.
I did take some measurements of my proplex panel and canvas panel from the side to show the sort of response expected in a room.
Similar to the burntcoil setup.
But have now forgotten which pictures are which🙄
So if you wish I can do them again ,but at night, with less traffic noise muddying up the LF.
Steve.
Looks like there might be a way to get the far field response from a DML without doing the full Helmholtz FEM. There is a python interface to a library bempp doing advanced boundary element method calculations, including multiprocessor/GPU speedup. The technical documentation is completely impenetrable to me at this point, but there are some friendlier docs, and most importantly some acoustics examples as Jupyter notebooks, including one on analysing far field SPL from a loudspeaker.
So provided Elmer can give us the velocity distribution from the forced modal analysis, we may be able to use bempp.
https://bempp.com/handbook/intro.html
https://github.com/mscroggs/bempp-acoustic-tutorials/blob/main/tutorials/5_loudspeaker.ipynb
So provided Elmer can give us the velocity distribution from the forced modal analysis, we may be able to use bempp.
https://bempp.com/handbook/intro.html
https://github.com/mscroggs/bempp-acoustic-tutorials/blob/main/tutorials/5_loudspeaker.ipynb
Steve,
I'm not sure that we actually need a name for it, particularly since I'm not sure such a thing actually exists. But we do have names for lots of other abstractions (like "Sasquatch", or "infinite plate", for example), so maybe having a name wouldn't hurt.
A. Pistonic: can't be that, since that already means a speaker that creates primarily by rigid body motion.
B. Bendingwave: nor this, because that already covers the entire range of speakers that create sound primarily by transverse waves, including DMLS.
C Other: So it must be this.
If it must have a name, it would be good if the name hinted at the characteristic that distinguishes it from DML. So how about one of these:
amodal bending wave (ABW), or amodal panel speaker (APS)?
or even:
Quasi-infinite bending wave (QIB, or QBW) or Quasi-infinite panel speaker (QIP)?
Others?
Eric
A close reading of the Goebel patent shows both bending wave and distributed mode definitions are used.
https://patents.google.com/patent/DE10246792A1/en
It’s a patent designed to lock down the perimeter absorption method hence the stress on bending wave. Looks like a legal differentiation for the sake of establishing the patent.
I will from now on use Wobbly Flat Thing or WFT as a generic term to cover all non- pistonic speakers.
Burnt
https://patents.google.com/patent/DE10246792A1/en
It’s a patent designed to lock down the perimeter absorption method hence the stress on bending wave. Looks like a legal differentiation for the sake of establishing the patent.
I will from now on use Wobbly Flat Thing or WFT as a generic term to cover all non- pistonic speakers.
Burnt
Looking at Toru Maruyama's vibration speaker patent, he too uses the now standard pistonic motion from a moving coil, only difference being that his moving coil (voice coil) moves a spike with a point, most probably a ball, at a fixed diaphragm - a flat panel, which is fixed at its edges. As he says in his patent,
same as we try to experiment with the very same materials. Only, he might be using a concave panel, more like the speaker cone.
Hello Burnt,
Hopefully you could explain why pointing the speakers at each other improves the sonic experience by a passive edge I am at a lost to see why?
Paul & Eric,
With your FEM simulations The modal patterns is of a particular sine wave excitation can you input other excitation waveforms?
What is the excitation power used to calculate the mode?
Are the modes the same amplitude across the surface?
How fast can the modes change on the surface?
Again many thanks in advance for your responses.
Hopefully you could explain why pointing the speakers at each other improves the sonic experience by a passive edge I am at a lost to see why?
Paul & Eric,
With your FEM simulations The modal patterns is of a particular sine wave excitation can you input other excitation waveforms?
What is the excitation power used to calculate the mode?
Are the modes the same amplitude across the surface?
How fast can the modes change on the surface?
Again many thanks in advance for your responses.
Actually, reading the early entries is quite interesting. That's where the people really went out to experiment. For example, Toru Maruyama type accidental result, Post 140: https://www.diyaudio.com/community/...s-as-a-full-range-speaker.272576/post-4299791 and some ideas to it, Post 160: https://www.diyaudio.com/community/...s-as-a-full-range-speaker.272576/post-4299791
Later, experiments with a normal cone speaker with DML, Post 203: https://www.diyaudio.com/community/...s-as-a-full-range-speaker.272576/post-4331217, and creating a two-spider exciter, Post 213: Post 203: https://www.diyaudio.com/community/...s-as-a-full-range-speaker.272576/post-4331217 and few YT vids 7 years ago
Later, experiments with a normal cone speaker with DML, Post 203: https://www.diyaudio.com/community/...s-as-a-full-range-speaker.272576/post-4331217, and creating a two-spider exciter, Post 213: Post 203: https://www.diyaudio.com/community/...s-as-a-full-range-speaker.272576/post-4331217 and few YT vids 7 years ago
The last link should be, Post 213: https://www.diyaudio.com/community/...s-as-a-full-range-speaker.272576/post-4336674
Hi Tagis,
It neither improves or degrades the sound but it does reduce the visual impact.
I have several sets of DML’s at home and can do a direct comparison, plus a couple of B&W bookshelf speakers.
Burnt
Thank you for all the inputs about FEM. Gmsh is working now on my laptop. Next is to have a look to this Jupyter notebook.
I've been going from the beginning of this thread for the last few hours, trying to remember the best parts/posts. I'm already at the 40th page. It was nice reading your discussions with certain people, and some arguing too.
Reading again brings out new/forgotten ideas. Must go to sleep now. Will be reading all the way, when time allows, for the next couple of weeks. Thanks for such an input.Tagis,
There are certainly various levels of sophistication in FEM models. Mine is probably among the most simple for this type of problem.
What it does is calculate the mode shapes and associated frequencies for free (unforced) vibrations for a plate with defined elastic properties, dimensions, and boundary conditions. So the model doesn't actually use a particular type of excitation (sine or other) or power of excitation.
I'm not exactly sure I know what you mean about the amplitudes, but I'll try to answer anyway. Here is one mode for a particular plate: The amplitude varies across the plate from zero (in the blue areas which are so called nodal lines) to a maximum at the center of the four red or orange circles (antinodes). Often the displacement amplitudes of each of the antinodes is similar, but it need not be so. In this particular case, the antinodes at the ends of the plate exhibit greater amplitude than the two closer to the center.
My model is doesn't tell me anything about how quickly modes can change.
Hope that helps.
Eric
Hi Tagis
This is my understanding of the physics and our reasons for spending way too long on a computer:
The modal analysis gives you a set of mode shapes, and the set of frequencies at which each mode shape occurs. The mode shapes depend upon the shape of the panel, and how it is supported (its boundary conditions). They are all of the ways that the panel can vibrate "of its own accord", in the absence of damping, or any applied force.
A sinusoidal excitation (force) applied to the panel at one of these modal frequencies will give you a peak in the dynamic response, which corresponds to the resonant peaks observed in a measured spectrum using REW, or in a tap-test (essentially the impulse response).
The mode shapes just give the shape - the relative displacement at any point in relation to the displacement at other points on the object. I don't believe the displacement is in relation to anything absolute, and I don't think the relative displacements of different mode shapes can necessarily be directly compared at this stage.
The mode shapes are orthogonal in the sense that any particular shape that the panel can assume can be constructed from a sum of the mode shapes (in a continuous system, the sum is in general infinite). An arbitrary signal can be broken down into a sum of sinusoidal signals (Fourier). So if your signal consists of some sum of the resonant frequencies of the panel, it will excite each mode in proportion to the amount present.
For a sinusoidal excitations of arbitrary frequency (ie not one of the modal frequencies), and of an arbitrary location on the panel, the software can approximate the panel response by looking at modes in the viscinity of the excitation frequency, and estimating how strongly each mode would be excited, to give a estimate of the overall panel response. Exactly how it apportions the energy, I'm not sure, but for example the closer the mode to the excitation frequency, the more energy in receives, and the closer the excitation point to an antinode of an nearby mode, the more energy it receives. I know that in some situations - like how a panel responds to ambient sound - an assumption of equipartion of energy is made - ie all modes receive equal energy.
Some FEM software can proceed "from scratch" - that is, given properties of the panel, BCs and damping, can calculate the dynamic response or transient response. I don't know if Elmer can do this, but I assume that the simple elastic plate (2D) model I have been using just uses the modal analysis to give an estimate of the forced response. I would expect that an analysis 'from scratch' would take a long time to compute. In any case, I've not gotten that far yet.
So we proceed by baby steps. Currently we have a modal analysis, which tells us how modal frequencies are distributed. That is useful info already, particularly for low frequencies, which are the weakest aspect of panels. Eric has used it to get more accurate estimates of material properties.
Beyond that, we want to get an idea of the forced response, as described above - an actual estimate of how the panel behaves across the frequency spectrum, and how strongly each mode is excited in relation to others to a point-force excitation.
But that still doesn't give us an estimate of the actual frequency response of a panel, in terms of the sound pressure in air. How does the movement of the panel - the lumps and bumps of any given forced response of the panel - translate to what you hear? Some modes are 'non-productive' - they cancel in the near field - because they have antinodes of opposite phase too close together in relation to the wavelength of that frequency in air. Baffles make a big difference too because they help to reduce out of phase cancellation around the sides of the panel. And of course, the proximity of the floor, walls, and the entire room response affect the sound too, but that is normally dealt with separately.
So that's our 'holy grail' - to have tools that can give us a estimate of a panel's actual response without having to build it. So we can optimise the design by doing quick changes and seeing the difference. Even if only approximate, it will be very useful to see "better" or "worse" for design modifications or new ideas.
If you want to better understand some of the theory behind this, there are an excellent set of lectures in dynamics from MIT at https://www.youtube.com/playlist?list=PLUl4u3cNGP62esZEwffjMAsEMW_YArxYC
See in particular lectures 19 and later about modal analysis, vibration.
Eric may be able to add to or correct my understanding.
Last edited:
The youtube link posted by frank 40 on post 218 was the one I was unsuccessfully trying to re-find a few weeks ago.
Thanks.
Steve.
This raises the vexed issue of damping.
In general there are three types of forces at work in a dynamical system - inertial forces related to how mass responds to a force (F=ma Newtons law), elastic forces related to how a spring responds to a force (F=kx, Hookes law) and viscous damping forces which are proportional to velocity. Damping is different to the other two because it represents energy loss.
Without damping, the simplest model of a dynamical system involving lumped mass and elasticity is a mass on a spring, which forms a simple harmonic oscillator. Undamped SHOs will oscillate at one particular frequency, forever. This is the classical "second order system".
When you add damping, the oscillations decay in an exponential fashion, depending upon how much damping you add. When a system is underdamped, it continues to oscillate for possibly many cycles, when overdamped, it decays very fast, but may not give either the amount, or the speed of responsiveness you desire. So for example in a car suspension, you have a mass, you have a spring, and you have damping. You want the damping to be such that the car rides easily over the bumps, but does not then oscillate as a result. If the damping is very high, you might as well not have springs at all. If very low, you'll be bouncing down the road. Like Goldilocks' bed, it should be juuust right.
We can extend this analogy to higher-order distributed system like panels. For accurate sound reproduction, you presumably want the panel to faithfully follow the input excitation and then stop vibrating ASAP when the excitation changes or stops. There are some problems though. The system is now distributed and extended through space, so it will take time for a signal to propagate, even when damping is "ideal". Secondly, the damping affects different frequencies to different extent. Eg if all frequencies decay within one cycle, you are talking three orders of magnitude shorter time period at 20kHz compared to 20 Hz. Damping materials and their application also seem to be something of a black art - difficult to predict and apply.
Layered on top of all of this complexity you have valid discussions of what makes the character of a DML speaker 'special' (and most people who have listened critically do agree that it is, in some sense, special). What are the psychoacoustics of DML in listening spaces? Is the resonant character of DML and ostensibly poor impulse response of a DML part of what makes it pleasurable to listen to? How much does it matter that the DML does not reproduce sound as accurately as other technologies? And so on.
Finally, layered upon this, you sometimes have discussions where people cannot agree upon what it is they are even talking about (see discussion between Eric and Steve above). My take: always clearly distinguish marketing terminology or marketing claims from descriptions of physical dynamics. Otherwise, confusion reigns and madness ensues.
Last edited:
A few years ago, someone, who was a member here directed us to this company. Looks like they are still doing business. Other than them, I can't find anyone doing business, real business, producing DMLs. Actually, it is really educating to fly through the thread from the beginning. Once, you omit the unnecessary stuff, it is a fast going.
One person, who was here, continued to stress that what's important is what your own ears hear, rather than the microphone and what the programs interpret. Sound advice, as whatever you build, you are going to hear them.
Last edited: