Carbon fiber + nomex honeycomb sandwich DML panel construction project

@majerjack I forgot to add that probably the easiest path to a rule of thumb for crossover point that I've found is just to look at the general design of Tectonic's panels relative to frequency response and add a fudge factor.

For example:

http://cdn2.hubspot.net/hubfs/449794/Product_Data_Sheets/QSG_3_Web.pdf?t=1482293437245

I can't find an exact measurement for their enclosures, but it looks like they're around half again as long along each edge as mine, and are recommended to cross over around 100Hz - 125Hz in the case of their single panel speaker, which is higher than I've been doing.
 
Seemingly contrary to the earlier quote about foam backing not necessarily helping mids and highs, the abstract of Prokofiev, et al. 2002 makes it sound like it should (seems intuitive, right?):

AES E-Library Intensity Measurements of the Acoustic Emission from a DML Panel

An experimental investigation is reported into the effects of a porous layer on the radiated sound intensity from distributed mode loudspeakers (DMLs). For an unbaffled panel, the results suggest that attaching an absorbent layer behind the panel leads to a smoothing of the spectra of sound intensity. When a specially developed enclosure was used, an improvement in the low frequency response of DML panel was observed. The inclusion of a porous layer in this enclosure further reduced the fluctuations of the emitted sound spectra, and smoothed the resonance peaks in the low frequency range.

Again, anyone with AES access who can help us out with a better idea of what kind of "special" enclosure is involved and other details from that paper?

OK, I'd better make this the last post for a while and get back to real life...
 
It is interesting to me that the Tectonic panels cross over to a ribbon at 6,500 Hz. It makes me wonder if their panel material rolls off the highs? It also makes me wonder if a stiff and dense hardwood such as sugar maple might reproduce high frequencies better.

I don't know if I understand the radiation pattern of a DML accurately. Some say it is like a dipole, some say it is not, but it does seem that wideband attenuation of the back radiation might be a good idea if the panel is against a wall.

I have years of experience servicing pianos and building, repairing, and playing guitars, and I believe some of the things I learned might come in handy in the construction of a DML. I think a traditional soundboard material such as spruce might make for a good panel, but I also remember that dulcimer builders prefer hardwoods for their soundboards because they believe that spruce is too lively for that instrument (too little damping?). It has been my experience that the harder and denser the wood used in guitar construction, the better the reproduction of high frequencies (the body of the guitar also comes into play here, but can be ignored, I think, for the present purpose). I think that a panel made of sugar maple (of which I have pieces large enough for panels) might make for a good DML, but perhaps even better might be Douglas fir, which is the hardest and stiffest softwood known to me (and of which I also have pieces large enough to make panels).

Also interesting to me is the possible arching of a DML panel to provide additional projection and damping. Years ago a piano manufacturer made a sales tool available to its dealers to illustrate to customers the advantages of the company's steel tension frame that served to keep the piano soundboard in a precise arch. The sales tool was a tuning fork attached to a spruce panel which could be locked into a completely flat or an arched position. When the fork was struck it produced very little volume with the panel flat, but when the panel was then locked into the arched position the fork suddenly sounded much louder. "Flat top" guitar builders are aware of this property of a soundboard and will commonly build their guitars with a soundboard arch radius of about 25 feet.

What about "time alignment" when wedding a DML panel to a low frequency driver? Is it wise to try to match the front of the panel to the voice coil of the bass driver in the vertical plane, or will it make little difference due to the wavelength of the low frequencies involved?

I think I feel a bout of Speaker Building fever coming on. Must resist.......
 
It is interesting to me that the Tectonic panels cross over to a ribbon at 6,500 Hz. It makes me wonder if their panel material rolls off the highs?

I think I feel a bout of Speaker Building fever coming on. Must resist.......

I think they added the ribbon to control directivity, they claim pretty wide dispersion of 120 degrees with the ribbon installed.

Don't resist the urge, how else will rest of us find something to complain about?

My thought was to laminate a couple of 2mm thick veneers with hide glue, to make a curved panel using something like a curved garbage can lid as a former.



Sent from my Nexus 6 using Tapatalk
 
Panel response

Signal chain to panel:
- HDMI output from GeForce graphics card
- To Magnavox LCD TV
- Out the TV's SPDIF output, to a Behringer Ultragraph DEQ1024 (everything flat, just used as a DAC)
- To a Yamaha P2700 power amp, with the volume set to about -19dB.

Measurement:
- Lexicon Omega SV USB, channel 1, volume all the way up in both hardware and software
- Dayton EMM-6 measurement mic, with calibration (can't remember where I got it, must have been from Parts Express I guess, if that matters in terms of who would've provided the calibration)

The graph below doesn't capture that the output sounded pretty smooth except for a nasty resonance somewhere in the midrange I would guess between 500-1000Hz. I couldn't tell if that was panel self-noise or, maybe more likely, the ragged edge of the aramid honeycomb scraping against the pillow that I've got the panel set up on.

Seems to be doing what it's supposed to though :eek:
 

Attachments

  • daytonultrapanel.png
    daytonultrapanel.png
    108.2 KB · Views: 697
EDIT: these measurements are with Room EQ Wizard, if that wasn't clear.

I might be measuring something incorrectly. The phase, impulse response, group delay and distortion are unintelligible. I get similar results measuring my sub, so I don't think it's the general difficulty of measuring DMLs with conventional methods that some papers point to.

There's a fair amount of fan noise from my amps that are sitting out in the middle of the room right now. Would that do it?

Here's my measured sub response for a PPSL, 12x16x38.5" sealed enclosure, 2x GRS 12-SW4 drivers:
 

Attachments

  • grs12sw4-ppsl.png
    grs12sw4-ppsl.png
    39.6 KB · Views: 653
Last edited:
Yeah, they were faulty. I didn't realize that REW distinguishes whether you're measuring *from* the right or the left "channel", a notion that doesn't even really make sense to me when we're talking about mono mic inputs. The previous graphs were showing whatever noise happened to show up in the other idle mic input, I think.

With that fixed, it looks like response is pretty good from around 60Hz - 5K. Maybe Tectonic is onto something with adding ribbons.
 

Attachments

  • daytonsmoothed.png
    daytonsmoothed.png
    43.1 KB · Views: 723
With that fixed...
Wonderful to see data. That's the way to go!

Ordinarily, you'd want 1/12 smoothing (as a good compromise value).

And you need to specify the distance (with your listening chair as a good choice since REW records the initial sound and less so the reverberation in the room).

Of great importance (in light of the poor results in the other thread where a home-brew poster-board DML is tested), it to display the distortion (total THD and maybe 2nd and 3rd harmonics). Just a matter of clicking "Distortion" in REW and you get it clear as day based on your last run. Failing to post the distortion - which is so easily posted - might possibly expose you to critical comments, eh.

http://www.diyaudio.com/forums/full-range/272576-study-dmls-full-range-speaker.html

It's hard to specify loudness. But we can take for granted you are testing at a level comparable to listening level?

Your mic doesn't need a correction table since it is plenty good enough raw. But for your own peace of mind, always good to run a loop (the system without the acoustic part of the loop) in order to be sure all the electronic and software elements are flat and clean.

Ben
 
Last edited:
Yeah, they were faulty. I didn't realize that REW distinguishes whether you're measuring *from* the right or the left "channel", a notion that doesn't even really make sense to me when we're talking about mono mic inputs. The previous graphs were showing whatever noise happened to show up in the other idle mic input, I think.

With that fixed, it looks like response is pretty good from around 60Hz - 5K. Maybe Tectonic is onto something with adding ribbons.

That result is shown a plate that damps to hard or the most Likly,not rigid enough from front to back, exciting the backplate in the higher frequencys will not get trough the core material to the front plate. Try using it the wrong way around exciter facing you, it will confirm or disconfirm the core being the problem. If the problem persists it could be it damps to hard.

What I did was using a stiff paper directly on the coil as a tweeter that excites the panel. It is a bit like a whizer. My tiny panel went to 17-18 kHz down to 300 , the whizer should be rather stiff since it still has to excite the panel for the lower end of the spectrum

Here is a video about me trying to get anything useful
https://youtu.be/ONOxc1p0XBI
 
Last edited:
That tiny panel went to 17-18 kHz down to 300 , the whizer should be rather stiff since it still has to excite the panel for the lower end of the spectrum

Here is a video about me trying to get anything useful
https://youtu.be/ONOxc1p0XBI
That describes a legitimate "full range" driver, and is the best sounding playback in your video series.

There's genuine midrange response that can't be discerned in the earlier Rubanoid or current Magnetostat experiments.

This is a promising design.


Sent from my Nexus 6 using Tapatalk
 
That describes a legitimate "full range" driver, and is the best sounding playback in your video series.

There's genuine midrange response that can't be discerned in the earlier Rubanoid or current Magnetostat experiments.

This is a promising design.


Sent from my Nexus 6 using Tapatalk

Well thank you for the kind words! I hope it can be of use for anyone wanting to try.


At the same time it might be the ugliest kobbled together design :) it's almost hard the replicate it even for me :) haha
 
Can you elaborate on what properties of a material would lead to it being overdamped in this sense?

Rigidity front to back does seem like a likely problem given the thickness. I considered sawing the honeycomb panel in half along the third axis to end up with a panel around 5mm thick instead of 10. As it is, the panel is clearly thicker than it needs to be. Only reason I didn't was that I wasn't sure how sawing honeycomb along that axis would work, and didn't want to damage materials without a backup when I could try getting the simplest solution working first. Maybe that's a good option to try for the next round, along with using carbon fiber in tape form as recommended earlier in the thread.

I think my near term next steps though will be to make a mockup frame out of cardboard or foam board to provide some edge and rear wave dampening.

Distortion measurements to come. They looked pretty bad as I recall, like maybe 5-10% THD, maybe higher at the far ends of the spectrum. It certainly sounds better than I would expect based on those numbers. I'm not sure if that's wishful thinking or down to difficulties applying conventional measurement techniques to DMLs.

Li's thesis shows a response of a simpler panel that's somewhat similar to what I've posted here:

http://publications.lib.chalmers.se/records/fulltext/154618.pdf

It is seen that DML lacks of low frequency while a dramatic drop occurs at 300 Hz and below 300 Hz the SPL is low except two peaks. On the other hand, even though the conventional loudspeaker is in the band of stiffness-control at low frequencies therefore the SPL decreases while the frequency decreases, it still holds relatively higher amplitude than DML until around 120 Hz, which is about the lower frequency limit this size of the multimedia loudspeaker can produce. The lack of low frequency of DML can be attributed to the fact that the modal density is relatively low at low frequencies.

As for high frequencies, due to the transformation of bending waves into transverse waves at high frequencies, the SPLs go down gradually. It might be notices there is a rapid descend at 3 kHz, which could be because of the presence of the critical frequency around that frequency. In contrast to that, the conventional speaker maintains certain magnitude of SPL at high frequencies as high as at middle frequencies, irrespective of the larger variations at high frequencies. For overall behaviours, it is clear that DML shows an obvious modal behaviour that there occur many peaks at corresponding nature frequencies. Since the modes are the responsible for radiating in the design of DML, therefore they are greeted to be seen as many as possible and as evenly distributed as possible, while as for conventional loudspeaker it has been asked as flat as possible.
 
Can you elaborate on what properties of a material would lead to it being overdamped in this sense?

Rigidity front to back does seem like a likely problem given the thickness. I considered sawing the honeycomb panel in half along the third axis to end up with a panel around 5mm thick instead of 10. As it is, the panel is clearly thicker than it needs to be. Only reason I didn't was that I wasn't sure how sawing honeycomb along that axis would work, and didn't want to damage materials without a backup when I could try getting the simplest solution working first. Maybe that's a good option to try for the next round, along with using carbon fiber in tape form as recommended earlier in the thread.

I think my near term next steps though will be to make a mockup frame out of cardboard or foam board to provide some edge and rear wave dampening.

Distortion measurements to come. They looked pretty bad as I recall, like maybe 5-10% THD, maybe higher at the far ends of the spectrum. It certainly sounds better than I would expect based on those numbers. I'm not sure if that's wishful thinking or down to difficulties applying conventional measurement techniques to DMLs.

Li's thesis shows a response of a simpler panel that's somewhat similar to what I've posted here:

http://publications.lib.chalmers.se/records/fulltext/154618.pdf

i dont think you can saw the honeycomb. but here in europe you can get 1 1.5 2 5 etc mm honeycomb the 1 mm is only 18 euro m2. but yes i think because it excites one plate in this case the back plate the low end will get the whole thing excited , but my guess is the high end wont excite all the way trough to the second plate. and is damped by the core. as usuall there is no free lunch i think. either bass is good or the highs suck. since they both need different properties and as usual they are far apart.


about the distortion, i noticed this as well , its impossible to get a good distortion like an esl of a planar. but at the same time it does sound rather ok not much noticable distortion. but if i remember correct it was mostly even order distortion witch is not easy to capture with your ears :) but it will screw up the total distortion graph, without hearing it