A Study of DML's as a Full Range Speaker

Edit oct 27, 2021: a nice Google drive library of DML publications by vdljev:

DML - Google Drive

Edit April 1, 2020: nice summary of DML patents by Burntcoil

A Study of DML's as a Full Range Speaker

Edit Aug 12, 2019: Nice summary here by BurntCoil on how to maximize performance.



I had some cheap exciters that I got from PE a while back and tested them out a while ago with a full 20x30in FC panel here:

Foam Core Board Speaker Enclosures? - Page 225 - diyAudio

I found that a large 20in x 30in panel can sound quite good with nice bass extension and a snappy transient response:

403240d1393817181-foam-core-board-speaker-enclosures-fc-exciter-ob-1.png


Impedance:

403242d1393817181-foam-core-board-speaker-enclosures-fc-dual-exciter-spl-impedance.jpg


Frequency Response & Harmonic Distortion:

403243d1393817181-foam-core-board-speaker-enclosures-fc-dual-exciter-hd.jpg


Impulse Response:

403244d1393817181-foam-core-board-speaker-enclosures-fc-dual-exciter-impulse.jpg


The results were interesting in that it sounded nice - with surprising bass and good midrange. I thought nothing more about it until recently prodded by the master of DML, CLS. He has worked on this extensively and is a treasure trove of good info. I since have discovered that you can really do some cool things with them by playing with driver placement, cutouts, adding mass, adding felt, adding ribs, making them huge, making them multi-way FAST, etc. the options seemed almost limitless. Take for example, a large center channel and a super 40Hz capable multiway that CLS built here: PIEZO NXT type panel - Page 60 - diyAudio and PIEZO NXT type panel - Page 61 - diyAudio

I feel like this technology just isn't getting enough attention. There are several very large threads on this on the web. It might be tough for a newbie to comb through. I am a newbie at this so thought I would document my journey for the Full Range forum to follow. I think it really could be a great full range speaker with some careful experiments. This will be really useful with some modeling using CAD and FEA modal analysis - such as available in many CAD packages like SolidWorks. One can play with shapes, cutouts, mass loading, variation in thickness and materials, boundary clamping conditions, etc. The idea is to spread the modes evenly with not any one mode dominating and causing a spike.

First thing is to play with it to get a feel of what we are dealing with. What is nice is that it is relatively cheap to play with. Exciters cost $3 to $20 ea and can be made by removing or cutting out large holes in the cone from an old driver and leaving the spider and VC to attach to the panel. Foam core or corrugated cardboard doped with shellac or PVA seems to be the materials of choice.

Some interesting facts:

1. Although it has no baffle, it is not an open baffle (OB) dipole in behavior - that is, there is no huge bass rolloff and it hits surprisingly low (40 to 50Hz is easy) for a zero baffle driver.

2. It is not a dipole but behaves more like a bipole or an omni.

3. It has very quick snappy transient response - nice drum sounds.

4. It is sensitive to how you mount it or frame it or hold it.

5. You are building a driver in reality - a driver and zero enclosure.

6. You want to avoid symmetric shapes and symmetric exciter placement in order to reduce the effects of the main symmetric transverse drum head eignmodes. Think reflection anti-symmetric shapes like uneven trapezoids, pentagons, blobs, etc.

7. It operates more by having high velocities and large areas for good efficiency vs large displacements - thus small drivers and large panels can be surprisingly loud.

8. The impedance is essentially "flat" relative to normal drivers in that there are modal peaks (many of them) but they range from nominal Re value (say 6 ohms to maybe 8 or 9 ohms throughout the 40Hz to 20kHz range - this presents a very flat load to an amplifier.

Here is a photo of a basic panel I was testing (right before I cut off some edges to form a trapezoid) - panels is about 1 square ft in size and made of standard dollar store foam core board. I suspended it with two pieces of twine from the top corners between a ladder to reduce the effects of edge boundary clamping:

476004d1428319574-study-dmls-full-range-speaker-trapezoid-dml-80mass.png


Then I started to play with mass loading by adding blobs of modeling clay (8, 5, 4, 2, 1, 0) so you can see the effect on the resonance modes and the harmonic distortion. What is neat with this test is you can do it live while playing music and immediately hear what sounds more pleasing to the ear. Note the 50Hz bass extension. It is also surprisingly efficient with 85dB and a puny little voice coil. These are 0.5m and 0.71v for equivalent SPL at 2.83v and 1m.

Frequency Response and Harmonic Distortion for...

No added mass:

476005d1428319574-study-dmls-full-range-speaker-trapezoid-no-mass.png


8 pieces of added mass:

476006d1428319574-study-dmls-full-range-speaker-trapezoid-8-mass.png


5 pieces of added mass:

476007d1428319574-study-dmls-full-range-speaker-trapezoid-5-mass.png


4 pieces of added mass:

476008d1428319574-study-dmls-full-range-speaker-trapezoid-4-mass.png


2 pieces of added mass:

476009d1428319574-study-dmls-full-range-speaker-trapezoid-2-mass.png


1 piece of added mass:

476010d1428319574-study-dmls-full-range-speaker-trapezoid-1-mass.png


Impulse Response of 1 piece of added mass, note the sharp transient capability with relatively low after pulse or ringing:

476011d1428319574-study-dmls-full-range-speaker-trapezoid-1-mass-ir.png


I liked the sound of the 5 and 1 mass the best and for a FAST with a 200Hz XO, the 5 mass might actually do fairly well. Of course, sound clips to follow once I have more time to work on this. Just to demonstrate how quick and easy this is, I did all these experiments in less than an hour including making the DML.
 

Attachments

  • Trapezoid-DML-80mass.png
    Trapezoid-DML-80mass.png
    396.1 KB · Views: 16,422
  • trapezoid-no-mass.png
    trapezoid-no-mass.png
    146.4 KB · Views: 14,716
  • trapezoid-8-mass.png
    trapezoid-8-mass.png
    146.4 KB · Views: 14,641
  • trapezoid-5-mass.png
    trapezoid-5-mass.png
    151.4 KB · Views: 14,593
  • trapezoid-4-mass.png
    trapezoid-4-mass.png
    151.7 KB · Views: 14,549
  • trapezoid-2-mass.png
    trapezoid-2-mass.png
    151.3 KB · Views: 14,527
  • trapezoid-1-mass.png
    trapezoid-1-mass.png
    143.7 KB · Views: 14,530
  • trapezoid-1-mass-ir.png
    trapezoid-1-mass-ir.png
    40.5 KB · Views: 14,432
Last edited:
XRK, if you'd like to try correlating modal analysis results with measured responses I can do a couple runs of your tests. I just need the weight, dimensions, and locations of the exciter and the masses. Any information needed to create an accurate model.

Unfortunately, I can't give you information about damping or relative amplitude. Only estimated frequencies and mode shapes. I suspect the mode shapes will be the most useful. Also, as we go we can try to refine material properties to get more accurate frequencies.
 
XRK, if you'd like to try correlating modal analysis results with measured responses I can do a couple runs of your tests. I just need the weight, dimensions, and locations of the exciter and the masses. Any information needed to create an accurate model.

Unfortunately, I can't give you information about damping or relative amplitude. Only estimated frequencies and mode shapes. I suspect the mode shapes will be the most useful. Also, as we go we can try to refine material properties to get more accurate frequencies.

MJ,
Thank you for the offer! I will get back to you when I take another shot at this and measure the masses and their placement carefully.
X
 

CLS

Member
2005-06-17 6:58 am
Taiwan
Several suggestions:

1. Bigger is better :D

2. Rounded corners (or irregular shapes) can help diffusing edge reflections. Think water tank - ripple experiment. Your added masses act like some stones along the banks, also diffuse the wave effectively. As you can see the effects vary a lot.

3. Foamcore board is light, so it's easily damped. Constrained layer damping can be applied locally around edges. I've try felt pads and they work well. See Manger's bending wave unit.

4. Foam is soft and can eat some vibration energy from the exciter. On such material, driving the front surface of the panel helps a lot in HF. Refer to the following sketch:

DML_driver_panel_zpscgrb2xi7.png~original


Cut a shallow hole and retain the front surface paper. Fill the hole with some other dense material - solid cardboard, plywood etc. So the exciter is now driving the front surface more directly, instead of riding on the spongy rear surface.

One step further, the dimension where exciter drive the board affects the highest frequency. My little exciter has 1" VC. The driving ring is about the same size. It can play up to 20kHz on proper panel.

If the rigid 'transmission' can be built in a smaller size, theoretically the HF can be even higher. Like the one on the right.

5. For the low end, in addition to bigger panel area, you'd need higher driving force. A powerful exciter, or multiple smaller ones.

6. Heavy 'panel' like framed glass can work. (or, every surface can) The main problem is the resonances could be too strong and hard to control.

I've also tried this. The glass is fix on a dry wall. So the exciter on glass is actually driving the glass and the whole wall. There's a huge peak at 32Hz, and other irregularities in frequency response here and there. Very amusing effect, but impossible to deal with those problem physically. Severe EQ is needed.

This is hugely fun, and has the potential of very good sounding by very cheap devices and materials, but a LOT of experiments are needed.
 
Last edited:

CLS

Member
2005-06-17 6:58 am
Taiwan
I could be wrong, but it doesn't feel right to load a DML with horn (or WG).

The driver of DML works in high velocity, while horn driver works in high pressure.

Waveforms in a horn need to be tidy, planar or sphere, as perfect as possible. But the radiation of a DML is, well, a total chaos.

About the radiating pattern, there have been many debates already. Wide and narrow dispersions have each own sets of merits and drawbacks.

For commercial products, Elac 4pi tweeter and MBL 101 sound very good to me. So, why omni? I'd say why not? :D
 
I'm surprised no one has tried to use the drivers of these to try to make Walsh type speakers.

What is the effect of driving two units on the same panel in reverse polarity?

Or sticking pins through the foam board in places (perhaps adding a dot of glue to both sides where it goes through paper)? Could that be used to 'tune' the coupling between front and back?
 

CLS

Member
2005-06-17 6:58 am
Taiwan
I'm surprised no one has tried to use the drivers of these to try to make Walsh type speakers.

There is.

http://www.diyaudio.com/forums/planars-exotics/207762-diy-walsh-driver-revisited.html

What is the effect of driving two units on the same panel in reverse polarity?

Since there're many (vibrating) modes across the whole area, so different polarity on the drivers is at least not harmful.

There're inevitably interferences among multiple drivers on a single panel. So it's advisable to spread them in a manner of different distances from each other.

This is the latest panel I built:
DSCF6859_zpse9fa5685.jpg~original


As can be seen, there're 5 exciters on it. Only the central one see the full voltage of the amp, and in normal polarity.

The two at top and bottom are wired in series, and both in - polarity.

The two at left and right are also in series and - polarity, and further attenuated by a resistor.

All these were driven by a stereo amp, with the central one on one channel, and other 4 on the other. Inputs of both channels see the same signal.

That's the way I wired them, and it was very good right at the beginning. So I didn't make any change.

Sticking pin(s) through different layers should be a great idea. I haven't tried such technique myself, but would expect great results. I think this would work very well on corrugated cardboard, too.


One thing I can't explain is the bipole character, polarity, and the impulse response.

If it is indeed bipole, the polarity on both sides should be the same. So, it should be no different driven by normal or inverse polarity. But how can we explain the impulse response? -- Very high positive peak and very low negative.
 
Last edited:
I was thinking along the same lines of reproducing instruments by making foam core violins and cellos, and perhaps a piano soundboard. Putting transducers on the "ensemble" and listening to my own chamber orchestra. CLS pointed out a neat web page on the digital piano thread with similar ideas:
The Hybrid Piano – Part 1 - NCF Music!

I have a cello and violin at home - perhaps put these guys on the sound board near the bridge and listen?