A Study of DMLs as a Full Range Speaker

RMAM,
I'm wondering if the material you are talking about is PMMA, rather than polyethylene. PMMA is also known as acrylic or plexiglass. Among the materials I've tested, PMMA seems to have very good internal damping (as your post suggests), based on the impedance results in the link below.
Eric

https://www.diyaudio.com/community/...s-as-a-full-range-speaker.272576/post-7499940

Valeric, it turns out to be polystyrene, more precisely syndiotactic polystyrene (sPS or PS-st) so... in the family of EPS and XPS.
The chemical world is a multiverse, my daughter told me about sPS-graft-poly(methyl methacrylate) (sPS-g-PMMA).
Still waiting for the manufacturer specs for more details.
 
Nidaplast!
https://www.nidaplast.com/en/products/61-nidaplast-8
1707813218552.png

But this is polyprop.
I think Podium use a polycarb version in their panels.
 
  • Like
Reactions: 1 user
The plates are 500x330mm and the support points in the latest version are located at 140mm from the edges and 20mm long.
Leob
So is it 4 points of contact then, two on each of the long sides? Or 8 (on both long and short sides)? And what is the thickness of your panel? I was thinking I might do an FEA model of your plate/mounting, just to see what the mode shapes and frequencies are. It would only be a rough approximation, as I don't know the material properties of your plates, but I can probably make a decent guess.
Eric
 
  • Like
Reactions: 1 user
Thanks @Veleric that would be amazing!
It is these plates https://www.dekokopf.com/neopor-styroporplatten-3er-set-50x33x2-5cm.html
So 25mm thick.
I added hide glue and shellac and rounded the corners, mostly to make it easier to mount everything together.

I add on 2 points on each long side, from both directions, so a total of 8 pads.
Previous revision I had one point in the middle on each side, so same amount of pads, and then I got a lot better response below 150Hz...but with many other variables being different as well.
 
I'm experimenting with different honeycomb panels, and I get the impression that the material makes more difference to the FR than the aspect ratio/size. For instance I made a 4 mm thick panel from higher weight honeycomb nomex than my 3 mm panels, and it's bass light, and has a very pronounced and annoying peak at 3 kHz. Changing the dimensions didn't make a difference to the peak. The panels shouldn't be to stiff to get enough bass - or so it seems.

My first nomex paper panel that "sounded" really good has ridges on the surface - a random pattern of bumps because I used to much epoxy, It pooled under the paper with air bubbles in between. Could it be that that accidentaly is a good way to "randomly" dampen the panel? It still is the best effort so far. It gave an immidiate "This sounds really nice" reaction. Has anyone a clue what might be happening? I'll try to make another "proper" panel, and to mimic the ridges with a textured roller and thickened epoxy.

If I manage to get a similar sounding panel as my first one I might keep that as "the standard". This panel thing is so complex. It might be best to stick with what (found by accident) works OK.

BTW the panels I make are about 450 g/m2 now, so building lightweight is doable.
BTW II The response of the panels changes quite a lot in the first days after making the panel, because of the epoxy slowly hardening to it's end value. When cured for 24 hours there's not much HF, and the panel sounds cardboard-y. Two days later the HF by magic has appeared!
 
Last edited:
Nidaplast!
https://www.nidaplast.com/en/products/61-nidaplast-8
View attachment 1272671
But this is polyprop.
I think Podium use a polycarb version in their panels.
Andre,
I tried buying something very similar in the US some time ago, but the supplier never had the thickness I wanted in stock, so I finally gave up. I was planning to try adding a fiberglass/epoxy or carbon fiber epoxy layers on top and bottom. I think it would have been a very efficient panel, but maybe also have been a bit too stiff. It's worth a try I think.
Regarding the podiums, I think they used a nomex honeycomb with Mylar skins, though I can't find the source of that info just now...
Eric
 
  • Like
Reactions: 1 user
Valeric, it turns out to be polystyrene, more precisely syndiotactic polystyrene (sPS or PS-st) so... in the family of EPS and XPS.
The chemical world is a multiverse, my daughter told me about sPS-graft-poly(methyl methacrylate) (sPS-g-PMMA).
Still waiting for the manufacturer specs for more details.
RMAM,
Interesting. PS makes more sense than PE. I would have expected it to be garden variety PS (i.e atactic) rather than syndiotactic, however. I would have thought that sPS would be too expensive to use for anything but fairly specialized applications, but I could be wrong about that...
Eric
 
Hello!
I am new to the DML "sport" so I've been doing some homework.
I've been through much of the pages here as well.
Now here's something I found out, that might be interesting.
There's a paper from Chalmers University of Technology:
"A critical review of bending wave loudspeaker technology and implementation
Master’s Thesis in the Master’s Programme in Sound and Vibration
KUONAN LI Department of Civil and Environmental Engineering
Division of Applied Acoustics
Room Acoustics Group
CHALMERS UNIVERSITY OF TECHNOLOGY Göteborg, Sweden 2010 Master’s Thesis 2010:9"

You can look it up, it's free to download.
In there they analyze some example DML panels and they discuss Excitation point and Geometry of Panels.
They have an example figure in which Modal Analysis is illustrated.
DML Modes.PNG


They comment on this:
"The position of the exciter will decide which modes of the plate will be excited, affecting the radiation performance of DML. Thus it is essential that one makes sure that the excitation point couples as many modes as possible, and ensures de-correlation between panel velocity and drive point. One might use, again, FEA to determinate the position of the exciter. Usually observing the first 20 modes or so is sufficient to conclude the optimal point of exciter, which is often near the middle point of the panel but never exactly on it. This is because that while the exciter is on the middle, it only can excite even modes. Figure 3.5 illustrates the nodal maps of the first 20 modes of a panel quoted from [9], which can be done by using FEM. The region where fewer nodal lines pass is a better position for exciter; in this case, possible choices are near the middle and the corners"

As we can clearly see, the region at the middle section (blank space, not dead center on the cross point ) provides the best possible positions for exciter placement, the edges are not considered as good choices.
Now look at this picture below.
I overplayed in red lines a Phi Matrix (Golden ratio) over the original image, just out of curiosity.
Look at this! It's a perfect match!
DML Modes with Golden ratio matrix Matrix.PNG


Hmmm! I think there's some juice here.
The picture above has the two ratios of 0.618 and 0.382 lined out.
Look at the picture below where the 5 lines of phi matrix are overlaid.

DML Modes with Golden ratio 5Matrix.PNG



Nature works in patterns ;) and I think there's an obvious one here.
This could possibly give us hints for better placement.
I am definitely going to try this on.
I would appreciate your thoughts or experiments on such a possibility as well.

Manos.
 
  • Like
Reactions: 3 users
RMAM,
Interesting. PS makes more sense than PE. I would have expected it to be garden variety PS (i.e atactic) rather than syndiotactic, however. I would have thought that sPS would be too expensive to use for anything but fairly specialized applications, but I could be wrong about that...
Eric
Eric, you are not wrong, it becomes established that they have the sheets wrong placed and priced. Meanwhile the atactic PS. with same dimensions are now playing with some more eq (i.e, less dB`s) in the 2 to 8 kHz band compared to the previous. The price diferences don`t favor the s-PS.
 
The honeycomb isn't attached to the panel and is not used for creating a panel, it's used as a kind of waveguide for the sound emitted by the panel.

Obviously, although you can dampen the panel attaching it to the honeycomb wich is needed depending on material type, dimensions and panel load.

Nevertheless, i think is more cost effective to add exciters on the same panel, or chose different materials/exciters for separate mid/low freq panels to achieve the same goal.
 
Leob,
Here is what I have so far using the LISA FEA to model your plates. There are lots of assumptions, so take it all with a grain of salt, but hopefully it's is still interesting to you.
I used plate dimensions of 550 mm x 330 mm x 25 mm, with a corner radius of 60 mm. The supports are 20 mm x 10 mm centered 140 mm from the plate edges. I assumed an elastic modulus of 30 MPa, density of 40 kg/m^3, and poisson ratio of 0.3. Those came from some EPS I tested before.
Images below show the estimated frequencies and mode shapes for the first 15 modes.
Based on the mode shapes, the ones I would expect to radiate sound best would be: 1,3,8,9 and 12.
Eric

1708219344910.png


1708219414022.png 1708219496773.png 1708219550270.png 1708219616837.png 1708219669726.png 1708219734004.png 1708219792341.png 1708219835423.png 1708219873079.png 1708219955618.png 1708220000832.png 1708220054118.png 1708220104821.png 1708220148767.png 1708220201348.png
 
  • Thank You
  • Like
Reactions: 1 users
Thanks @Veleric! Very interesting and seems to correlate quite well with the FR where 1, 3 and 9 coincides with the prominent peaks.

Small correction to the material data....the GPS I use is 25kg/m³.

I'm trying to figure out the optimum positions to place the padding, and not sure if this data can help me in any way?
I guess I would need more simulations with different positions to get an idea if FR will improve when moving the pads closer to the edges, or placing them on the center of each edge. I did some tests listening to different clamping positions, but need to be more thorough and do some more tests and record the FR.
 
Eric .. Thanks... And wouldn't it also mean that any mode shown above with blue through the driven point/s would be not achievable.? In fact, one could surmise that it would only be in a low frequency resonance situation in which the maximum panel deflection may conceivably occur away from the driven point/s, and with the semi centrally located multi driver array used by Leob, this is unlikely to occur.
Research by Anderson and co indicates that above the modal crossover frequency of a particular panel, the distributed mode effect is replaced by displacements which are heavily concentrated around each driver. The close spacing of Leob's drivers are likely to result in a spreading of the driving "point", and perhaps unusual stresses on the individual voice coils, perhaps going someway to explaining the damage he's reporting. It is also worth pondering the likely effect of this close cluster on the high frequency performance of the panel.
Eucy
 
  • Like
Reactions: 1 users
The close spacing of Leob's drivers are likely to result in a spreading of the driving "point", and perhaps unusual stresses on the individual voice coils, perhaps going someway to explaining the damage he's reporting.
That is a very interesting point.
Could this mean that the voice-coils experience a lower mechanical impedance than what they're designed for, and thus bottom out too easily?
This might imply that drivers should be spaced out over the relevant nodes instead of being clustered into tight groups, contrary to Dayton et al recommendation.
 
Not sure, but possible.

I was thinking more in terms of asymmetrical loading due to the panel deflection being greater on the inside of the group than on the outside. Maybe a marginal effect but could cause issues over time

The problem in chasing the nodes is that the mere act of choosing a driver location changes the node patterns available, add another driver and the patterns change again, and so on
 
You understand the concept, but near the centre is claimed to be the best location, and importantly, it assumes only a single drive point.

So unless the location of each driver in an array is complementary in terms of modal patterns at each frequency (which may in practice be impossible), based on this natural frequency analysis, they will fight against one another, as claimed often by Spedge

In that case, a close cluster would logically perform with less interference than that of a spread pattern, or line array, but still with some disadvantage over that of a single driver, apart from the increased in output volume.

However, can a natural frequency analysis approach be adopted when multiple drivers are used?. I am unconvinced. It seems to me that a full analysis would require superposition of the effects of a number of single driver analyses to optimise the interactions..A massive task... Bring on AI...

Eucy
 
  • Like
Reactions: 1 users