Eucy,
One of the factors cited by Christian that I find most useful is the T factor. B/mu^3 or E/rho^3. I first found it in the work by Heron, which led to start of NXT, as I understand it. It really explains why PS foam is so loud! Can you find anything louder?! And why acrylic (PMMA) is so quiet.
But when you use it, you have to be really careful about the values you use for E and rho. For plywood especially! The elastic modulus of wood in the directions perpendicular to the grain are only 1/10 to 1/20 of the modulus parallel to the grain. So you can't take the textbook value for E (parallel) and assume your plywood has an effective modulus that even close to that in either direction.
Likewise, the density of any species varies widely. Plus, usually the core layers of plywood are not even the same species as the core. So the only way to have any idea of the moduli and density of plywood to use in the calculation is to measure them for the actual panel you have.
Eric
My guess is that the "hairy" response comes from the comb filtering that happens when the multiple modes combine. That is part of the nature of DML, and that kind of very fine comb filtering is something our hearing is very used to since most natural sound sources would behave the same. If you record for example a violin, I think you will see the same "hairiness".
Longer tails are also part of the nature of DML. If you want it to measure like a cone or planar, then why not just use a cone or a planar? If you do like the sound of DML, then I think you need to accept hairiness and a bit longer impulse response since that is part of the nature of DML sound.
I do try to ensure that response is reasonably short and even over frequency and time, but I move my frame of reference what even and short is for a DML speaker.
@Leob, understandable, but part of the fun is trying to make DML better 
I understand the nature of DML involves modes, but why do those modes need to keep ringing after they're done being excited? They don't. They are hard to stop though, and that's what I'm trying to achieve with the right kind of damping. Excite easily but stop quickly after excitement is done.
Here's some measurements, in case anyone else is interested, comparing off-axis FR for planar and DML from 3ft away at 0 degrees, 20, 30, 45. All done with same settings as my last post (EQ for different positions in the room, crossed over, other channels muted). Planar is 3" GRS, DML is 9"x19" exposed area of heavily PVA'd XPS with two coats of rubber.
I understand the nature of DML involves modes, but why do those modes need to keep ringing after they're done being excited? They don't. They are hard to stop though, and that's what I'm trying to achieve with the right kind of damping. Excite easily but stop quickly after excitement is done.Here's some measurements, in case anyone else is interested, comparing off-axis FR for planar and DML from 3ft away at 0 degrees, 20, 30, 45. All done with same settings as my last post (EQ for different positions in the room, crossed over, other channels muted). Planar is 3" GRS, DML is 9"x19" exposed area of heavily PVA'd XPS with two coats of rubber.
Yep - fully aware
Yep - except in the laserply I've been using, all plies are made with the same material .
Eric- I have access to Midas and it handles ortho, iso and aniso materials. I've run some sims on 3mm ortho cedar, but the results are weird so it needs more time on my part to sort out why. I guess you've noticed how this interest chews up hours
Eucy
Hello SapphireSloth,
I already gave you my opinion your DML is too ringing. I understand you design constraint about the width but with that your are not in good condition to stop it bouncing. The oscillations in the DML IR are in beginning of its FR, probably much below 2k.
About the close mic measurement technique, it is used for pistonic loudspeakers in order to get their low frequency response. This technique has a limit in its upper frequency. It is clearly defined in the case of a pistonic loudspeaker. When the frequency increases, the phase of the point of the emitting surface distant from the mic becomes too high leading to cancellation. In the case of the DML, it is a convenient measurement to have an idea of what happens locally but I don't know if their is a study showing the limit of this technique.
Below is the measure of a small 8cm wide range (Visaton FRS8 in a 1.2l closed box) at 1m without smoothing. Hairy too but not for the same reason. A DML is by nature "hairy" because working on modes, the pistonic speaker becomes hairy because of the reflections in the room. At the end, as Eucy wrote, our hearing system filters that and then has the illusion of something much more smooth.
+ @Veleric
Hello,
2 points,
About the exciter fundamental, I don't fully understand...
We can evaluate the modes of a panel. When the exciter is glued on it, it adds locally a mass (its voice coil) and a stiffness (its spider) which will shift up the panel modes. It is a new system which has 2 variants depending if the magnet is free or almost blocked by a spine.
In an other hand, for the exciter alone, it is possible to define 2 resonances. One of the voice coil, the magnet being blocked, one of the voice coil, the magnet being free and the voice coil blocked but this not the panel plus exciter system.
So is the proposal here to use the exciter magnet mode as to be lower than the first panel mode? What about the case of a spine which is mandatory for a long term design?
The second point is about the "low frequency" measurements. Those last days, I made measurements of almost all the speakers I have here, DML, pistonic. I did that as almost all my DML measurements, same room (my kitchen!) speaker and mic at their usual positions, 1m between them... and I rediscovered a fact that I completly fogets with DML : measuring low frequency is not easy, . Below are the FR of a pistonic 2 way speaker
- green : the FR at 1m
- blue : the FR at close distance of the woofer.
For a pistonic speaker, the way to merge the 2 measurements is a topic in itself covered by papers.
For DML, I don't remember reading something about that.
It opens to new questions :
- what the papers say about the frequency under which a measurement is at distance is not correct (related to the room, the Schroeder frequency?)
- is the close distance measurement of DML relevant as for a pistonic loudspeaker?
- the room transfert for the considered set can be estimated from the 2 curves below but will it be the same for a DML? A close box is a monopole at those freqneies, a DML is a dipole. Do we have to compare to a open baffle pistonic loudspeaker ?
All this second point to say to be cautious with the DML FR let say below 250/300 Hz?
Sorry Eucy, it doesn't answer to your question about DML design rule but if it can avoid others to be in this measurement trap, it is important.
Christian,
I'm not sure I understand exactly your question. But I agree that everything you say here is true. As you say, the panel modes will be shifted to higher frequencies when the exciter is added. And I will add for clarification that the fundamental panel frequency is shifted the most, by up to about 20 Hz, but higher modes are shifted by much less, and by, say, the fourth or fifth mode any shift is negligible.
Concerning the two resonances of the exciter, the one with the magnet free and voice coil fixed is the only one that appears in the impedance curves after the exciter is attached to a panel. I suspect that the resonance with the magnet fixed is useful for determining the voice coil mass, but otherwise I don't see direct evidence of it once it is attached to a panel.
Tentatively, I would say yes. And not just a little, but that the first panel mode should be 2x or 3x the magnet mode. But I'm not sure I can really state this as a good general rule. I can only say that when I have tried panels with a fundamental closer than that to the magnet mode frequency, the fundamental is not excited very well. But once the fundamental is over 2x or 3x the magnet mode, it can be excited very effectively. I once thought there might be an advantage to making larger panels, in order to increase the modal density, but when I tried it it was never better than if the first panel mode was very close to about 100 H, or about 3x the exciter magnet mode frequency.
That said, I must admit that most of my prototyping is done without a spine to support the exciter. And when I have compared results with and without a spine, they are almost always virtually identical. But I have almost certainly made those comparisons for panels with fundamental close to 3x the magnet mode. Would I get a different result for larger panels with a lower fundamental? I don't know! Maybe.
Eric
I did some tests recently with a 5mm thick XPS panel with epoxy on both sides. No fiberglass. I expected the HF to be better than thicker panels, but it was worse. I then drilled a circular grid of ~2.5mm holes over the driver mounting area and filled them with epoxy. This performed even worse in HF. Then I sanded off all epoxy from the driver mounting area on both sides, leaving the epoxy 'sticks' in the holes, and this performed the best in HF. There was probably only 3mm of XPS foam left in that area, so pretty floppy. This still was not as good as a bare 20mm thick XPS panel for HF.
Before this test I thought that the low compression strength of foam was reducing the HF, but it is not, directly. Or at least, reinforcing or replacing it with epoxy does not improve it. Now I'm thinking that the HF may be very low energy that cannot excite a very big area on the panel because it is lost to internal damping. The HF area needs to be flexible enough to oscillate tiny distances very quickly without getting damped out. That's my current theory but I don't yet know how to test it.
An exciter alone, without a panel, puts out a significant amount of dB for HF. I think maybe sometimes the panel reduces the HF below what the exciter alone would produce. Long ago I took a measurement of a lonely driver from 1ft away. I don't know what the volume level was but I'll recheck tomorrow with a new test. Here's the old one: