DIY Flat Panel Speaker Love

[rmeinke]

Re: DIY Flat Panel Speaker Love

Rich,
I see on the AudioCircle thread you have tried a 1/8" plywood panel. I've very curious to know what happens to both the sound and the measured results when going from flat and then to stressed by slightly bending it. I suppose you would have to pull the sides toward on another and hold them in that position with some sort of cable or something from side to side. You might need several of them along the length. The places where these cables connected would seemingly constrain the movement of the panel edge to some degree, which could be good or bad.

Hey Dan!
I'll have to try that. I did it with just my hands the other day and it seemed the tighten the bass as expected. This was done at the same session that I clamped on the balsa to tight the panel horizontally. This was just a quick, very unscientific test but was what I would have expected given the flexible, even somewhat "floppy" panel. But to your point, by constraining any of the edges its hard to test the impact of stressing the panel +- the edge clamped suspension.

After testing with the balsa, stressing the panel with a curves piece of balsa might be the best bet. The balsa is incredibly light and may not impact the panel as much as I thought. Might simply act as a single panel material however I have to imagine that the waves will terminate differently and may cause some waves to travel back into the panel. The smaller the piece of balsa the better off we would be.

I'm going to try measurements later tonight. Get back to you.

[rmeinke]

Re: DIY Flat Panel Speaker Love

It is in the section where they are discussing radiation models (around pg 13-16 or so). They discuss the first model, which call the rectangular array of elementary sources and say this is the NXT model, concluding the following:

This model has been proposed by NXT, and included in their DML design software ([1]). Although we have decided to present it here, we do not think this model is a good model. There are several reasons to doubt the correctness of the model:
• The choice to observe the normal velocity Vn on a control surface S instead of on the vibrating surface itself can be justified (and is not mathematically wrong). However, the validity of the assumption that the velocity distribution can be represented by an array of elementary sources has not been proven satisfactorily.
• The validity of the choice of the spacing dw = dh = π/k is not given by [1]. It would be more appropriate to use π/kb as a measure of the spacing dw and dh.
• Setting Vn constant over the entire panel is quite dubious.
• The calculated number of sources has to be set to an integer value in order to be able to actually perform the summation of (2.34). Hence, the predicted acoustic response is equal over a frequency range.
• The model breaks down for small panels: the number of elementary sources is dependent on the physical size of the panel. For small panels, the number of predicted sources can be smaller than one, which leads to erroneous predictions. For instance: the AM1-3 panel has a size of 12 by 20 cm, hence for frequencies underneath 1450 Hz a single sources is predicted for the small side.
• Given the previous points, the model is most appropriate for large panels observed from a great distance, although these conditions are seldom fulfilled in practice. The model of elementary sources seems to have very limited applicability. It is included here for reasons of completeness.[/COLOR]

Then they talk about the second model, which the call the Velocity model and conclude:

This radiation model seems to be more applicable than the model given in the previous section. This model does not involve dubious assumptions and simplifications. Remember that the pressure calculated by this model can only be a far-field pressure. Otherwise, this model is universally applicable to calculate the acoustic pressure P(r,φ,θ, k) induced by a vibrating panel.

Thanks for the reference Dan.

Yes, seems that the model of spacing/number of nodes on a panel is not very tight... assumes the model breaks at lower frequencies on smaller panels but I wonder what the real impact is exactly? Are there are other factors in the model that, when the model is summed, compensated for this??? Surely the model was tested against real measurements and I have seen other papers that show modeled vs. measured freq. resp. even for small panels that are extremely close so it would be more interesting if they demonstrated the impact rather than point to very dubious math.

Not supporting or denying this paper or NXT's model. I have seen more model vs. measured responses so I'll side with the model until the weakness of the model is demonstrated.

[klane]

Re: DIY Flat Panel Speaker Love

I have some concerns about testing of DML. Should I expect any difference when I test the DML compared to a conventional piston-type speaker. If there are differences, what should I expect. I will be using a calibrated mic with either LMS, TruRTA or a SPL meter. I was reading the link I have posted below about the podium DML. At the bottom of the page it states conventional methods of measurement are completely inappropriate for vibrating panels.

http://www.podium-sound.com/index-A.html?menu=70836&id=70824

How should I measure DML's if I can not used conventional methods?

I am designing a DML that can be used outdoors. Is a DML recommended for outdoors? I realize the SPL does not obey the inverse square law (-6dB as distance doubles). Since there will be less drop in SPL, it seems like a DML would be good for outdoor use. Is there anything I should be aware of about using a DML outdoors.

[rmeinke]

Re: DIY Flat Panel Speaker Love

I have some concerns about testing of DML. Should I expect any difference when I test the DML compared to a conventional piston-type speaker. If there are differences, what should I expect. I will be using a calibrated mic with either LMS, TruRTA or a SPL meter. I was reading the link I have posted below about the podium DML. At the bottom of the page it states conventional methods of measurement are completely inappropriate for vibrating panels.

http://www.podium-sound.com/index-A.html?menu=70836&id=70824

How should I measure DML's if I can not used conventional methods?

I am designing a DML that can be used outdoors. Is a DML recommended for outdoors? I realize the SPL does not obey the inverse square law (-6dB as distance doubles). Since there will be less drop in SPL, it seems like a DML would be good for outdoor use. Is there anything I should be aware of about using a DML outdoors.

I know of 2 things related to measuring DML panels.

1.) Because of the random nodes that are generated on the panel, a single mic location is not accurate. You really need to test across multiple locations. I have experienced this in my own testing. It was stated that measurements should be taken from multiple locations in the room.

2.) If a panel is EQued flat, the power response/tonal balance is tilted too far to the high frequencies... a gradual increase to hearing limits. I have EQed flat and had to back off the HF by at least 6dB. I'll look for the paper that illustrates the power response on a DML panel. I have an open question about using OmniMic to measure the "power response". If anyone reading this thread know has insight into this, a little guidance would be appreciated.

okay... maybe 3.)
One additional area that I personally question is the CSD waterfall plots. The bass appears clean but the high frequencies certainly are NOT. This, you have to wonder, could be an artifact of the random generation of nodes across the panel. They are not coming from a single point source as is the case with a typical done tweeter. Instead, the high frequencies are generated over the entire panel. So if you took very small tweeters and made an array 2 feet wide and 30 inches tall would you get a waterfall similar to what I am measuring with OmniMic? I would have to think that it would measure very close to what a DML panel measures. This is not a problem with the lowest frequencies as the number of cycles over the test window are small. Unlike the high frequencies where the wave form is very short with a high number of cycles over the test window and generated across the entire panel. Busy tonight but want to test this theory by increasing the number of measured cycles to see at what time in the waterfall the decay stops. Want to tie the length of the wave forms to the height/width of the panel.


Aside from these 2 and my own question about waterfalls plots, from all the papers that I have read, nothing else has been mentioned. So with all the modeling and measurements how many more measurement differences are there... especially in regard to freq. resp.????


Finally, I wonder how old that particular comment from Podium is in regard to measuring DML's? Might be worth a quick email to Podium to find that status of that research!?!?

[rmeinke]

Re: DIY Flat Panel Speaker Love

Wanted to say also that DMLs do really follow the inverse square law (-6dB as distance doubles). If you built a very small panel you would still see a -6dB reduction with each doubling of distance. HOWEVER, if you build a tall panel, it begins to operate just as any line array and will only see a -3dB reduction. Have not read that specifically, but experience this in my own testing.

[rmeinke]

Re: DIY Flat Panel Speaker Love

Quick update on the Dayton Audio "Ultra" exciters.

First, the leads from the input tabs to the VC are fragile and because they seem to be attached to the end of plastic spider, they are not the most robust. I blame myself somewhat as I'm using solderless connectors and because the tabs are very short and the panels are vibrating, they can become disconnected... when I reconnected with music playing it caused the exciter failures. Assuming that there was a large excursion caused by reconnecting which breaks the fragile lead wire on the positive tab (attached to the end of the spider that moves larger distance big excursions).

Despite this... I still ordered 4 more.

Got them remounted recently on XPS panels and they sound fantastic as I suggested, even only after a brief listen, in earlier posts. If anyone wants to build a fantastic sounding DML panel I can recommend these exciters based on sound quality... despite their apparent fragile construction.

All the qualities of a high efficiency panel are here in spades with this exciter and highly recommended on 1in XPS, sanded and treated with the 1:1 treatment of white/wood glue and water as first proposed by Sedge over at Audio Circles.

You most certainly won't be disappointed!

[klane]

Re: DIY Flat Panel Speaker Love

What does the sanding and white/wood glue do to the performance of the speaker? Are there other treatments to XPS that would help the performance? If there are other treatments, then how to they change the untreated XPS?

[UnfitNitwit]

Re: DIY Flat Panel Speaker Love

Would scoring or carving shapes into the XPS foam have an effect on distortion at certain frequencies? Something like carving the shapes made in these type of videos: https://www.youtube.com/watch?v=1yaqUI4b974

[rmeinke]

Re: DIY Flat Panel Speaker Love

What does the sanding and white/wood glue do to the performance of the speaker? Are there other treatments to XPS that would help the performance? If there are other treatments, then how to they change the untreated XPS?

Let me preface this by stating that the panel treatment was discovered over at Audio Circles. It is said that when you listen to a DML panel, you are essentially listening to the "skin". With EPS/XPS there is a "skin" left from the cutting process. Listening to these materials untreated is described as sounding "plasticy" and was, in fact, dismissed for DML usage for years as it's performance was, at best, mid-fi.

The treatment is said to aid wave propagation across the panel surface. I'm not sure if that is true. What I do know is that an untreated XPS panel does not sound good. Treating a panel damps and possibly even stiffens it. If you run your fingers across a PVA:water treated panel the sound is amplified and is really quite loud.

I know that some have tried shellac but it over damps the panel and reduced high-frequency output. I think the PVA/water treatment works well because it is really thin as it goes on like water. I bet shellac could be used to better effect if it was thinned. I'm sure that there are many treatments that could work as long is the treatment is very thin.

Takes a lot of time to test different panel treatment... if only we had more time to test stuff!

[klane]

Re: DIY Flat Panel Speaker Love

Am I correct in the fact that you treat the side of the panel with the PV:water and do you sand before treating? Do you treat the side that is visible to the listener? Is it most important to treat the side where the exciter is located?

[rmeinke]

Re: DIY Flat Panel Speaker Love

The panel preparation goes in this order :

1. Round the corners. I use a 6 in diameter bowl to mark the line and cut the panel with a steak knife. I will go smaller in my next build; 3 inch diameter should be fine.

2. Break the edges with a quarter-round router bit. This is just to rough out the edge... rough is the key word.

3. With a palm sander + 100 grit, sand front and back of the panel removing the XPS dust before it builds up as it can leave heavy swirl marks. It will have swirl marks either way but I try to keep the panel as clean and smooth as possible. I usually finish with 220 grit but the last time I didn't and the panels looked about the same.

4. Hand sand the rough quarter-round edges with 100 grit to straighten the lines and round the edges uniformilly. I finish the edges with 220.

5. Treatment : After mixing the treatment, I roll it on with a foam roller to front and back but try not to get it on the edges. Some folks don't treat all the way to the edge, leaving about an inch but I go right up to where the panel begins to round over. I use a hair dryer to dry the panels so I can get through the treatment quickly.


Just a note, if you don't get off all of the skin, the glue:water treatment doesn't stick to the panel. I always make sure the entire "skin" is sanded off.

I hope that you will report back on your findings. I'll be very curious what you think of the sound quality of the untreated vs. treated. While my test wasn't completely controlled, I felt I heard improvement in the sound quality for 4-5 days until the panel treatment was completely cured.

Good luck!

[klane]

Re: DIY Flat Panel Speaker Love

Does the treatment improve efficiency? I would think the added weight would reduce the SPL.

[rmeinke]

Re: DIY Flat Panel Speaker Love

Would scoring or carving shapes into the XPS foam have an effect on distortion at certain frequencies? Something like carving the shapes made in these type of videos: https://www.youtube.com/watch?v=1yaqUI4b974

That an interesting video and thanks for sharing. My mind was totally blown last night... I'm still wrapping my head around what I saw.

Soooo.... out came the salt and a spare panel last night.

While I was able to get patterns to form on the panel they took much longer to form and where not as complete (had I added more salt, I think I would be able to). My material is softer and more damped than the small and stiff metal plate that was used in the video so the pattern formation is not as crisp.

I would have to think more about this, but first pass... I can see how you might be able to carve out or add a material such as thin felt to the center of each formed pattern to damp at that specific frequency. But is it useful for a Hi-Fi panel speaker??? I doubt it. From what I see, the patterns formed are quite specific to a frequency played and the formed shape of the pattern changes with only a few Hz difference. So the damping would would be in a VERY NARROW freq. band. EQ is much more effective than some physical attempt to alter the freq. response of the panel.

Others have added mass to the panels at certain frequencies but as they point out, it just pushes problems to different parts of the freq. resp. These types of attempts to alter the panel has been described as Roundabouts and Switchbacks... no real solutions have been found.

Will say that there *may* be more experimentation needed in the lowest frequencies where the number of patterns on the panel *should* be greatly reduced, say, in the lowest octave. 20Hz-40Hz might have only 4-5 patterns and can only imagine that the patterns are closer in shape?!? This is really "out there" stuff but a few minutes with a salt shaker might at least provide more learning's.

Thanks for sharing UnfitNitWit (nice handle btw)

[UnfitNitwit]

Re: DIY Flat Panel Speaker Love

I was thinking if there was a frequency that needed cut or attenuated(?), that tone could be played and the resulting pattern in the salt could be traced out. Then something applied or material removed in that pattern to make the panel sound better.

[rmeinke]

Re: DIY Flat Panel Speaker Love

I was thinking if there was a frequency that needed cut or attenuated(?), that tone could be played and the resulting pattern in the salt could be traced out. Then something applied or material removed in that pattern to make the panel sound better.

I'm with you here and understand what you are suggesting. The issue that I see is that each pattern is specific to a very small freq. range (a few Hz). Our ears can not detect narrow freq. peeks so attempting to attenuate such a very narrow range is not useful for practical. We are only interested in broader peeks. For example, a freq. hump from 2.2K to 2.8KHz. How do we reduce such a range where there are many patterns that are very likely overlapping and conflict with one another. Even with the small amount of testing I did last night I can see the conflict in the patterns over a short freq. range.

A point of clarification... the lines that form the pattern are nulls, the center of each open area is where the highest output of energy would be on the panel at that specific frequency.