A Study of DMLs as a Full Range Speaker

I did see a difference between the front and back at 2-3kHz, but not more than about 5 dB in any case. Are you seeing more difference than that?
Eric
Hello Eric,
Having a new look to measurements (after a good night!), the difference in SPL for the cavity resonance might come from REW, how the IR is windowed or not.
In the FR I posted yesterday, the IRs are windowed (gated) at 5ms in order to collect almost only the direct sound and rejecting the room reflections.
Today, I have opened the same files of the clear polystyrene panel (see below) this time with default window settings of REW (left 125ms, right 500ms). the difference in SPL is around 5dB

1738569939105.png


In the plot below, I have duplicated the 2 FR above to show them with a long and a short window.

1738569900498.png

Front : 500ms window in dark red, 5ms window in orange
Rear : 500ms window in dark green, 5ms window in light green
In the 2/3k range, it is the front which has the most important influence of the window duration which makes sense as the main sound source is from the rear. Part of the sound from the rear is "added" to the front one thanks to the reflections (long observation time).
Christian
 
I am also trying to understand this better. Is cavity noise or resonance a problem? Can it be heard or only shows up in measurements?

André, it looks like a hole in the panel does not solve the issue, but instead creates a bigger one. See your post here:

And it seems damping the inside cavity did not help either. See this post:
Thanks Jerry! Memory like an elephant compared to mine like Swiss Cheese eh?
 
it looks like a hole in the panel does not solve the issue, but instead creates a bigger one.
Humm... Just thinking loud after the post to try to answer to Eric about the influence of the IR window on the visibility of the cavity resonance peak...

It might be that making a hole solves the cavity resonance problem (= a strong wild source from the rear side) but then makes appear another problem. It makes sense seeing the level of the cavity resonance and the fact the power response are almost flat despite this additional source...
Using a common expression : "plague or cholera?" or FR : "du moindre mal, il faut choisir" translated by Deepl in EN "the lesser of two evils"
Arghh...
 
  • Like
Reactions: twocents
I have suspected the same effect at times also.
Impedance looks much better.
Two thoughts:
  • It would be interesting to see what it looks like with a heavier exciter, like the 25FHE or 25VT. Those would separate the magnet resonance from the panel resonance.
  • I'm assuming that your test case is with "fixed" edges, as in your recent paper, but I wonder if it wouldn't be better to evaluate the impedance model performance using free edges, rather than fixed. IME it's really hard to get truly fixed edges. First, there is really no perfect clamp, so the panel dimensions are not really well defined. Second, the clamp itself always adds a little damping, but how much? But hanging a panel with a pair of long masking tape strips comes really close to "free" edges, and you don't have to worry about having introducing damping at the perimeter.
Eric
Hi all -

Two new pictures of impedance simulated in my PETTaLS model with an updated, more advanced electromechanical model for the exciters. First one is for the same acrylic panel as before, Second one is for an aluminum panel. Results match much better than previously.

I'm working to incorporate more exciters into the model now that this is sorted out. I was actually developing this with the primary intent of giving designers a tool to choice between exciter models, hence why it's called "exciter and tactile transducer loudspeaker simulator." So my first big goal is to incorporate as many exciter models as possible.

I'll also do some measurements with free panels as soon as I'm able!

Some people had questions about free magnet vs. magnet attached to a spine - I'm planning to make another video soon, and I'll include that as a discussion point.

Still some things to clean up in the UI before it can be available as a download, but hopefully that won't take long.

Acrylic_impedanceNEW.jpg


Aluminum_impedanceNEW.jpg
 
To dampen/attenuate the backwave or not?
Below are 2 polar plots of the same panel I use for test (30x40cm clear polystyrene, I think close to acrylyc). There is a peripheral suspension and a frame.
It is plotted with and equalization of the +/-20° listening window (on axis).
In the left, it is in the open back conditions. The null at 90° is clearly visible (also some 30 to 50° lobes and the cavity noise at the rear)
Right is the same with a thick absorber on the rear and then a plate. It is not a closed box, the sound can probably escape from the sides of the absorber. Nevertheless, it seems the sound is able to go more than 90° when the frequency decreases. The cavity noise is absorbed, the front lobes are reduced also (is it their DML part remaining?).
Nothing say it is better or not for music! It was just a measurement to get an idea to what change when the rear wave is attenuated.

Looking at Christian’s post (#13361) and plots it looks like there is a strong case to be made to attenuate the back wave if I interpret the polar plots correctly.

Am I correct to think the measurements on the right (panel with a rear absorber) shows a much better dispersion and even spread of sound? Can we say that it is no longer a dipole in nature? I don't know of any open baffle speakers with backs or absorbent material at the back.
What are your views or listening impressions - should the back wave be partially or completely dampened or not? What noticeable differences does this make to the sound? Sorry, I know this is a subjective question or the answers will be subjective, but I would like to know other’s impressions.

Also, the polar plot on the left clearly shows the nuls at 90 degrees or at the edges. Yet we have had members here reporting that loose hanging dml panels actually sounds very good at the edges. Member Burntcoil’s Snow White panels (see post #6782) were made to listen on-edge. Could room reflections cancel the nuls then? How big influence does room nodes and reflections have on the sound of a dml? The same as for a normal speaker?
 
Hello Christian.
It is nice to see someone taking the problems in the exciter area seriously.
I have been mentioning these problem for too many years now 😕
BMRs use large holes in their very long coil formers .
I have removed the centre area altogether and even filled this gap with a phase plug.
The only problem with this is that you loose the hf output up to 20k from this area.
That is assuming your panel and exciter combination can reach 20k.
We never did finish building a panel between us, as you shot to the end of the experiment before I even started 🫣😃
We didn't even get to sorting out the 10k peak and dip .
But you were also using a very high density eps which has its own problems.
It is very difficult to generalise about DML panels as the different materials react differently.
I use different methods to try and tame different panel materials.
I never even got around to measuring the responses of using matchsticks to vent the exciter mounting.
I was expecting it to reduce and even lower the cavity frequency resonance.
My exciters already have a vent hole which probably acts like a helmholtz resonator.
Andre used a CD with a smaller hole than the exciter ring, this would be like reducing the port size on a reflex speaker.
This is why I used a light fabric dome on my very thin panels, so that the exciter cavity area could breathe a little but also retain a lot of the hf up to 20k, similar to a fabric dome tweeter.
But it's primary function was to reduce the cavity resonance.
Which it did.
The cavity resonance and so called oil can resonances sound pretty awful to my golden ears, I'm glad that your frequency plots agree 👍
Steve.
 
To dampen/attenuate the backwave or not?


Looking at Christian’s post (#13361) and plots it looks like there is a strong case to be made to attenuate the back wave if I interpret the polar plots correctly.
At least, it is interesting to try to find a solution to reduce the cavity resonance so that the rear sound has less difference compare to the front one. The cavity resonance is not specific to the DML and its exciter. Below is the same kind of plots for a standard full range Visaton FRS8 on an open baffle of similar dimensions as the DML test panel. It would be interesting to have opinion from OB (Open Baffle) builders. A good reason to have a look to back wave attenuation is to test a loudspeaker less sensitive to its placement or more tolerant to a close front wall.
Am I correct to think the measurements on the right (panel with a rear absorber) shows a much better dispersion and even spread of sound? Can we say that it is no longer a dipole in nature?
You are right, it is no more a dipole. Maybe something more in the category of the cardioid?
I don't know of any open baffle speakers with backs or absorbent material at the back.
Neither I do... In a certain way it is against the concept. Is it a taboo or is there no necessity? In an other hand, there is probably not so many directivity measurements of OB ?
What are your views or listening impressions - should the back wave be partially or completely dampened or not? What noticeable differences does this make to the sound? Sorry, I know this is a subjective question or the answers will be subjective, but I would like to know other’s impressions.
Excellent question...
Also, the polar plot on the left clearly shows the nuls at 90 degrees or at the edges. Yet we have had members here reporting that loose hanging dml panels actually sounds very good at the edges. Member Burntcoil’s Snow White panels (see post #6782) were made to listen on-edge.
I remember that also and I wonder how it can be. The IR of a DML is in my opinion quite good over a large angle but not a 90°. Does that orientation compensate some other problem?
Could room reflections cancel the nuls then? How big influence does room nodes and reflections have on the sound of a dml? The same as for a normal speaker?
If you make measurements with a long observation time (50 or 100ms compare to the 5ms of the plot I shared), this null is not visible. It is the case of most of the FR shared here. 5ms is a kind of anechoic measurement. The null at 90° of a DML is in my opinion not different from an OB. Like for the coincidence frequency, it was expected from the theory but not seen in practice, a question of measurement. The advantage of such directvity shown for example by S Linkwithz is the side walls are in the direction of the 90° so a low level of 1st lateral reflection. I have in mind that for how we perceive sound the 5ms and the 50ms horizon of time are important. The 5ms to get the temporal information (the precision of the image?), the 50ms for the loudness evaluation. Our DMLs with a low frequency in the range 100 to 300Hz work mainly above the Schroeder frequency of our rooms so not mainly in the range where the room is driven by its modes but more by its reflections.

You are right to open those questions. A balance measurements/listening experiences is needed.

Christian

Visaton FRS8 in OB
1738610678618.png
 
  • Like
Reactions: twocents
It is nice to see someone taking the problems in the exciter area seriously.
Hello Steve,
I was about to contact you to see with you what you tested against the cavity resonance. I of course remember your tests, the central hole, the tea bag... but I was not sure of the consequence on the HF. Thank you for the summary in your post.
Sorry if I went to fast in the steps of the EPS DML The good side is it leads to measurements of things we were waiting for like the coincidence frequency, the cavity resonance, the dipole behavior.
In addition, compare to the step we stopped, I just understood this week-end that my EPS test panel doesn't suffer from the effect of its coincidence frequency which is in fact hidden by the low pass filter in relation to the diameter of the ring of the exciter but most probably to the important lobes out off axis that are reflected in my rooms which are a bit reverberant. The effect of the lobes is an assumption while I have evidences for the coincidence frequency hidden in the low pass filter effect of the ring of the exciter.
I am also satisfied that the measurement approach is in agreement with what you experienced, for the cavity resonance but also about the materials regarding the presence or not of the coincidence measurement.
Christian
 
Looking at Christian’s post (#13361) and plots it looks like there is a strong case to be made to attenuate the back wave if I interpret the polar plots correctly.
I'm not saying I agree or disagree, but I did put a back on one of my panels a little more than a year ago. Results are below for on-axis SPL and Impedance. Solid is open, and dashed is with the back in place. Adding the back panel seemed to suppress the output at the fundamental (around 90 Hz), and boost it at 160 Hz, making the on-axis response less flat in that region. You could see the change in the impedance near 90 Hz also. In this case the "cabinet" was about 5" and not completely closed. There was foam support around most of the panel perimeter, but not within a few inches of each corner or the panel, so the back wasn't sealed off.
Eric

1738621697845.png
 
  • Like
Reactions: twocents
It is tricky because indoors you need a good acoustic environment to test, especially in the lower range of the spectrum. The exact setup and proximity to walls will be an important factor when determining the effect of the back panel at measuring position. But one of the benefits with a closed back is reducing reflections from a back wall, so you cannot really just go on neutral measurements at close range or made outdoors either.

Been thinking a bit about it recently since I'm considering if I want to close the back some way, at least for my studio plates.
Just jotting down some of my thoughts here....most of the considerations are the same as for regular speaker design, and being somewhat of a novice in that as well I might have missed some important aspects or got some things wrong. The alternatives I can think of are:

Open back
➕ Dipole pattern which can be useful in rare situations.
➕ Open "non-boxy" sound.
➖ At lower freqs the speaker will act more like a regular cone, and without a baffle rear and front waveforms will cancel out.

Open Baffle
➕ Should also give a dipole pattern and open sound.
➕ Reduces front and back wave cancellation.
➖ Plates are already large, and adding a large baffle becomes unwieldy.

Infinite baffle
➕ Would think this gives the cleanest sound, eliminating back waves while not impeding the plate, allowing it to move freely as when fully open.
➖ Only for installations or home/studio use.

Semi-open (back plate for absorption)
➕ Will reduce sound emitted to eventual back walls or allow for better directing sound outdoors when needed.
➕ Reduces front and back wave cancellation.
➖ Can start to introduce some slight boxyness.
➖ Cancellation of back and front waves will still happen, but more controlled and at a shorter wavelength.
➖ Slightly more things to think about when designing. Not a design commonly used for regular speakers, so have to experiment with material used, distance of back plate and perhaps directing the openings.

Sealed
➕ Should avoid back waves.
➖ Inefficient.
➖ Might sound more boxy.
➖ Will increase size and weight.
➖ A bit more complicated to design if wanting to make it small as possible and not just a big infinite baffle.

Ported
➕ Should both avoid front and rear cancellation.
➕ Efficient.
➖ Ported also is summation of two sources, port and element instead of front and back. But they should be in phase, so better, but still a compromise in sound.
➖ Probably most boxy sounding.
➖ Hard to design correctly. Need to find the parameters of the plate assembly and might be difficult to use tools designed for regular cones to tune the design.
➖ Will increase size.

The main differences with DML compared to regular speakers when it comes to enclosing are:
1) Diffuse waves means that summation of out of phase back and front waves might not be as bad
2) Plates need to be a bit bigger than your typical mid-top cones, making some designs less practical. Horns are pretty much out of the question, and making big boxes around the plates makes it very bulky since you will need quite many plates for high power applications.
3) DML has a very open spacious sound that many love, which might be compromised by enclosing them

My conclusion is that the enclosure option I think it is worth trying out for PA applications, apart from fully open, is semi-open.
For studio/home, semi-open is probably better than fully open if plates needs to be close to a wall, but mounting the plates in the wall for infinite baffle would be ideal. Also a sealed enclosure could work since size and efficiency is not as important.
 
Last edited:
In all the responses I have seen and listening tests, the high frequency lobes appear to come from the central area of the voice coil.
Do you concur?
Steve.
Hello Steve,
I have no recent investigation about the areas sources of HF lobes. What it appears for different panels I tested in directivity is a peak in the 10k on axis which is already known here but also higher SPL in the angles from 60° as the frequency increases. See below the EPS test panel we discussed. The yellow/orange or even red spots show higher SPL. They are mainly on axis in the 1k range to be more and more off axis while the on axis SPL is lower (dark green on the upper left plot) to come to the on axis HF peak. it is what I called "lobes" in the previous posts. It is visible in the polar plot (lower right) which is a simulation of on axis equalization (front is 0° to the left, 90° the sides, 180° the rear, low frequency out to HF to the center) with the 'orange slice" from 2k and up from 40° to maybe 70°.
Working on the damping is maybe a way to reduce (smooth) this effect?... but is it needed. This is measurements not listening tests.
Maybe a next step in the connection of what you experimented and measurements
Christian

1738670562645.png
 
A remark that I hope won't create a new entropy... I just linked what I shared in the previous post to answer to Steve (@spedge ) and the remark from @twocents before with @BurntCoil listening the DML at high angles.
I have extract below the FR at 0° and 60° of this EPS test panel

The 2 upper FR (red and green) are the FR at 0° and 60° windowed at 100ms so including the room reflections. Kind of energy response. They are very similar the panel was turned by 60° (the mic has a fixed position in the room) showing the angles effect can be seen, hides by the room reflections.
The 2 lower FR are in blue 0° and orange 60° with a 3ms window so a rejection of the room reflections to get the direct sound (before reflections)
The FR at 60° is almost flat compare to the 0°...
A flat anechoic (reflection free) FR with a smooth decreasing power response is a target in my understanding.

Might it be this panel more pleasant at 60°?
If we imagine a equilateral stereo triangle so 60° between the listener to the left source direction and the listener to the right source, then to listen according to the 60° direction of the panels, they are parallel to the side walls, perpendicular to the front wall...

Who is ready to test after this strange (in first approach, silly after?) remark?

Note that but I have almost no experience in the domain, EQing this panel on axis to get it flat leads to too high SPL at 60° (what the polar plot before shows) and probably an almost flat power response which from comments here or from what we can read about equalization is not the target (see down sloping house curves used with EQ).

Christian

1738673123516.png