Hello all,
During time I've collected all kind of information about esls.
What strikes me is that all cumulative spectrum decay plots (CSD) from ESLs I've seen sofar, all show resonances at mid and high frequencies. Not very bad, but very good (dynamic) tweeters are able to exceed the ESL on this excercise and show CSD plots with a more rapid decay.
This is somewhat against the popular belief that the ultralight esl-film is effectively damped by the surrounding air and immediately stops with no ringing.
In a dutch book (Fikier) it is stated that the delays shown on a CSD plot can not be heard as they are 'masked' in case of an ESL. But I couldn't find any theory supporting or explainig this. The only thing I can think of is that the several resonances will cancel at a longer distance as they travel across the film while the CSD plot is closed miked and shows a small area only.
Does anyone have some opinions about this masking effect?
( it may help to explain the subjectively experience of micro-resolution in ESLs)
Martin-Jan
During time I've collected all kind of information about esls.
What strikes me is that all cumulative spectrum decay plots (CSD) from ESLs I've seen sofar, all show resonances at mid and high frequencies. Not very bad, but very good (dynamic) tweeters are able to exceed the ESL on this excercise and show CSD plots with a more rapid decay.
This is somewhat against the popular belief that the ultralight esl-film is effectively damped by the surrounding air and immediately stops with no ringing.
In a dutch book (Fikier) it is stated that the delays shown on a CSD plot can not be heard as they are 'masked' in case of an ESL. But I couldn't find any theory supporting or explainig this. The only thing I can think of is that the several resonances will cancel at a longer distance as they travel across the film while the CSD plot is closed miked and shows a small area only.
Does anyone have some opinions about this masking effect?
( it may help to explain the subjectively experience of micro-resolution in ESLs)
Martin-Jan
Have you seen this thread? I inserted a discussion of the CSD measurements of panel speakers in post #31 and followed up with a few others. Perhaps it'll be useful.
Few
Few
Hi Few,
Thanks for the link!
It seems from the comments of Atkinson about a decay-plot, that there might indeed be a possibility of canceling out of the resonances in case of large flexible membrane.
All the decay plots shown in the thread are worse than top-tweeters like scan speak....
Thanks for the link!
It seems from the comments of Atkinson about a decay-plot, that there might indeed be a possibility of canceling out of the resonances in case of large flexible membrane.
All the decay plots shown in the thread are worse than top-tweeters like scan speak....
Hi,
the quality of CSD of ESL is depending on the design of the ESL.
Curved designs with highly tensioned membrane show worse CSD. This is obvious since multiple resonances caused by bended shape and high tension of small areas wont enable fast decay.
Best CSD is possible using nonsegmented flat panels with optimized width-height ratio and very important optimized membrane tensioning.
Segmented ESL panels can have a very good CSD, but it is a long way of trial and error to find out appropriate segment areas and segment filtering. Segmented areas and depending filters will store energy.
Capaciti
the quality of CSD of ESL is depending on the design of the ESL.
Curved designs with highly tensioned membrane show worse CSD. This is obvious since multiple resonances caused by bended shape and high tension of small areas wont enable fast decay.
Best CSD is possible using nonsegmented flat panels with optimized width-height ratio and very important optimized membrane tensioning.
Segmented ESL panels can have a very good CSD, but it is a long way of trial and error to find out appropriate segment areas and segment filtering. Segmented areas and depending filters will store energy.
Capaciti
Hi a.wayne,
yes that CSD is from my segmented fullrange panel. Membrane width is 20 cm, having 17 segments.
Capaciti
yes that CSD is from my segmented fullrange panel. Membrane width is 20 cm, having 17 segments.
Capaciti
Capaciti,
Based on your experience would it be accurate to say that some of the "hash" in the CSD when using segmented panels comes from having neighboring regions of the diaphragm driven by different signals? Your 17 segment, 20 cm wide panel would presumably drive neighboring segments of the panel with quite similar signals (because there are so many segments), and a non-segmented panel will, of course, have the entire diaphragm driven by the same signal.
Also, can you say a little more about what you've found regarding diaphragm tension? As you progress from low tension through the best tension to too much tension, is there a trend in the CSD that you've observed? This is the first I've heard of a connection between tension and energy storage, apart from the fundamental resonance, so I'm very interested to know what you've observed.
Finally (sorry to have so many questions!) have you seen any clear trends associated with the number or pattern of diaphragm clamping points, such as silicone dots between the diaphragm and stators?
Thanks for your helpful posts.
Few
Based on your experience would it be accurate to say that some of the "hash" in the CSD when using segmented panels comes from having neighboring regions of the diaphragm driven by different signals? Your 17 segment, 20 cm wide panel would presumably drive neighboring segments of the panel with quite similar signals (because there are so many segments), and a non-segmented panel will, of course, have the entire diaphragm driven by the same signal.
Also, can you say a little more about what you've found regarding diaphragm tension? As you progress from low tension through the best tension to too much tension, is there a trend in the CSD that you've observed? This is the first I've heard of a connection between tension and energy storage, apart from the fundamental resonance, so I'm very interested to know what you've observed.
Finally (sorry to have so many questions!) have you seen any clear trends associated with the number or pattern of diaphragm clamping points, such as silicone dots between the diaphragm and stators?
Thanks for your helpful posts.
Few
Hi Few,
1. you are right, if segmentation is not optimized adjacent areas will interact with each other in a negative way.
2. My experience shows that a just mechanical strechted membrane even can take years until remaining resonances due to stretching are gone. Be aware that any resonance of the membrane will disturb CSD. E.g. a resonance at 200 Hz will show up as a multitude at e.g. 1000 and 2000 Hz where its critical to the ears
3. Thats why i perform a mixture of mechanical stretching and thermal treatment.
4. Silicone dots can have negative impact to CSD when distributed wrong. This cannot be calculated, just try different positions and measure. symmetrical is worse than nonsymmetric distribution of dots.
5. Most critical to CSD is the surrounding frame where the membrane is glued to. Often this is just some rectangular spacer material with a sharp edge. i am adding a damping layer at the outer edges of the membran area to minimize longitudinal resonance modes.
1. you are right, if segmentation is not optimized adjacent areas will interact with each other in a negative way.
2. My experience shows that a just mechanical strechted membrane even can take years until remaining resonances due to stretching are gone. Be aware that any resonance of the membrane will disturb CSD. E.g. a resonance at 200 Hz will show up as a multitude at e.g. 1000 and 2000 Hz where its critical to the ears
3. Thats why i perform a mixture of mechanical stretching and thermal treatment.
4. Silicone dots can have negative impact to CSD when distributed wrong. This cannot be calculated, just try different positions and measure. symmetrical is worse than nonsymmetric distribution of dots.
5. Most critical to CSD is the surrounding frame where the membrane is glued to. Often this is just some rectangular spacer material with a sharp edge. i am adding a damping layer at the outer edges of the membran area to minimize longitudinal resonance modes.
Hi Few,
The resonance modes on a rectangular membrane are defined by:
F = sqrt(T/M) * sqrt( (m/a)^2 + (n/b)^2)
where:
T = tension per unit length(uniform in all directions)
M = mass per unit area(uniform)
a = membrane length
b = membrane width
m,n = integer mode identifiers (m=n=1 give fundamental resonance)
The pic shows the first few modes, but you can imagine what the higher order ones look like.
As Capaciti mentioned, with fundamental mode(1,1) at 200Hz there are many higher order vibration modes dancing on the diaphragm that you can see with a flashlight while sweeping slowly with a frequency generator. You can also see the modes in near field measurements as you move the microphone slowly up and down or side to side on the panel.
As with other mechanical resonating systems, with mass held constant and tension increased, the damping ration goes down, so the CSD would show a slower decay.
Of course once you start adding in silicone dots things get much more complicated as tension in no longer uniform.
@Capaciti,
Have you experimented with different tensions for length vs width for your flat panel as is done for curved panels? I wonder if having tension different for the two directions in the diaphragm adds to the poor CSD of the curved panels, or if the higher tension alone is the main reason.
Attachments
Last edited:
Hi Martin-Jan,
Looking at the attached pic of some of the diaphragm modes, notice that even modes have equal areas radiating in and out of phase. For example, (1,2), (2,2), (2,4),...etc. The acoustic output from these modes does cancel out at longer measuring distances.
The odd modes (1,1), (1,3), (2,3), (3,1), (3,2), (3,3),...etc have unequal areas vibrating in and out of phase. So, these do show up in measurements made at a distance.
Attachments
Thanks, all, for the posts. The theory dictating the pattern of drumhead modes is very familiar (two-dimensional quantum particle in a box...) but I haven't seen good measurements made on real ESLs showing how clearly each of the resonances show up. Can anyone point me to such a set of measurements? I guess the results would be very dependent on the construction of the ESL so maybe it just needs to be done on a case by case basis.
I'm building my stretched wire stators with about 40% open area and fine wires in the hope that the increased damping will help control some of the less desirable diaphragm behavior but I'm shooting from the hip: I haven't looked into the question of whether such an approach is more likely to affect the higher or lower frequency modes. I'm guessing the fundamental would be strongly affected, based on previous discussions of adding damping cloth behind the rear stator.
Capaciti: When trying to damp the resonances by adding damping material around the perimeter of the diaphragm I would assume that very soft material would have to be used in order for it to make any difference. Otherwise I would think the impedance mismatch between the diaphragm and the damping material would reduce the damping effect. Is that what you have found in practice? Any suggestions for damping materials that work well? (I'm not asking for trade secrets, just non-proprietary information; I realize you're looking to make a living off your designs.)
Few
I'm building my stretched wire stators with about 40% open area and fine wires in the hope that the increased damping will help control some of the less desirable diaphragm behavior but I'm shooting from the hip: I haven't looked into the question of whether such an approach is more likely to affect the higher or lower frequency modes. I'm guessing the fundamental would be strongly affected, based on previous discussions of adding damping cloth behind the rear stator.
Capaciti: When trying to damp the resonances by adding damping material around the perimeter of the diaphragm I would assume that very soft material would have to be used in order for it to make any difference. Otherwise I would think the impedance mismatch between the diaphragm and the damping material would reduce the damping effect. Is that what you have found in practice? Any suggestions for damping materials that work well? (I'm not asking for trade secrets, just non-proprietary information; I realize you're looking to make a living off your designs.)
Few
I've done near field measurements of the resonances many times. Scanning my files I found an example that might help give you and idea what to expect. The diaphragm was 60" long by 3 5/8" wide. I measured the top half of the panel at 10 equally spaced locations: Mark 0 thru 9. On each plot, the red curve is the response at Mark 0(the middle of the panel). The microphone was placed in the middle of the width dimensions. You get a whole different family of resonances with the microphone displaced left or right of the middle. I was mainly interested in the LF behavior so the dB scale was pretty large, but you can still see the resonance behavior up at 500 hz. CSD plots done near field highlight the higher frequency modes better than these impulse response measurements.
Oh, forgot to mention that open area was 48% with crossbars every 3"...which made for the conveniently equal spaced measurements.
It's just one data point, but I did not notice any significant increase in damping when comparing wire panels with 43% open area with 51% open area.
Not sure if you had seen this before, but here is my positive experience with using very thin mesh made for silk screen painting as damping material.
http://www.diyaudio.com/forums/plan...con-dots-resonance-control-3.html#post1958582
Attachments
Last edited:
Bolserst: Thanks very much for posting the measurements. Very interesting. I hadn't considered simply comparing frequency response measurements taken at different locations close to the diaphragm. I was envisioning mapping out the nodal patterns by looking at changes in the phase relationship between the drive signal and the acoustic signal as the microphone is moved--or perhaps using two mics at different nearfield locations and comparing the phases of their signals. Your method has the advantage of displaying all the frequencies, and therefore multiple resonances, at once.
I guess the take home message is that while ESL evangelists (and I guess I'm one of them) like to brag about the uniformly driven diaphragm, your measurements demonstrate that uniform drive sure doesn't necessarily imply uniform, pistonic diaphragm motion. I'm not surprised there are resonances, but they are more pronounced than I expected. Have you looked at the CSD plot for the driver you used for the sequence of measurements you posted?
Few
I guess the take home message is that while ESL evangelists (and I guess I'm one of them) like to brag about the uniformly driven diaphragm, your measurements demonstrate that uniform drive sure doesn't necessarily imply uniform, pistonic diaphragm motion. I'm not surprised there are resonances, but they are more pronounced than I expected. Have you looked at the CSD plot for the driver you used for the sequence of measurements you posted?
Few
Hi Few,
No, I didn't grab any screen captures of the CSD plots for that measurement series. At the time I was trying to verify the locations of the peak amplitudes for the first few(most prominent) odd modes and then apply a silicone damper to these locations and observe the results. But, I may still have the impulse data saved and could recreate the CSDs. I'll see if I can locate them later this week. You won't see anything unexpected in the near field. Basically ridges of decay at each of the modal frequencies.
Getting back to the comment made by MJ Dijkstra in the beginning of this thread, when you pull the microphone away from the diaphragm toward the listening position the acoustic outputs from the even resonance modes cancel. The odd modes are reduced in amplitude but remain because the amount of in-phase and out-of-phase panels areas are similar but not equal. The most prominent modes(besides the fundamental) are (1,3) and (3,1) where the in-phase and out-of-phase areas differ by 50%...something Capaciti warned me about last year when I started experimenting with the silicone dampers and had placed some at the nodal lines for one of these modes which tended to enhance rather than damp the mode.
A random but slightly related thought that popped in my head...
B&W uses a similar technique in their FST midrange driver where even order break-up modes are encouraged with increasing frequency. The result is that the acoustic output from the cone breakup cancels at the listening position. Also, the apparent size of the source shrinks with increasing frequency as the area of the cone involved in the break-up behavior grows from the outside inward toward the voice coil.
Last edited:
Thanks for sharing your data--very helpful. When asking about the CSD plots what I was wondering was whether the nice clean far field CSD plot posted by Capaciti would correlate with lower amplitude resonances in your series of near field frequency response measurements. If so, it would support the argument that the hash seen in far field CSD plots comes from the same diaphragm resonances that can be seen using your near field technique. I expect the two to be correlated, but I hate to make assumptions that come back to bite me. Also, you near field technique suggests (I think--see next paragraph) where damping ought to be placed to tame a ridge in the far field CSD plot.
Did you have success damping selected resonances by strategically locating silicone dots? I assume you placed them at the anti-nodes of the troublesome standing wave patterns.
Few
Hi,
in a typical room CSD will become worse by increasing distance to the speaker.
Thats expected since reflections from walls, bottom and ceiling will radiate delayed energy and thus disturb CSD seriously.
Thats why long membranes are preferable. Bottom and ceiling reflections are minimized even towards lower frequencies.
So a long ESL sounds less disturbed at wider distances.
btw1: My damping at the outer edges of the membrane is a surrounding layer which behaves like silicone or rubber.
btw2: It is very important that the esl need to be absolutely quit at any frequency you run it. It took my long time to get there. My current design is free of any additional noise. I tested other brands and a lot of them showed noises when sweeeping a sinus from e.g. 40-20000 Hz. Often it is a ringing or rattling. Obvious that this results in bad CSD
Capaciti
in a typical room CSD will become worse by increasing distance to the speaker.
Thats expected since reflections from walls, bottom and ceiling will radiate delayed energy and thus disturb CSD seriously.
Thats why long membranes are preferable. Bottom and ceiling reflections are minimized even towards lower frequencies.
So a long ESL sounds less disturbed at wider distances.
btw1: My damping at the outer edges of the membrane is a surrounding layer which behaves like silicone or rubber.
btw2: It is very important that the esl need to be absolutely quit at any frequency you run it. It took my long time to get there. My current design is free of any additional noise. I tested other brands and a lot of them showed noises when sweeeping a sinus from e.g. 40-20000 Hz. Often it is a ringing or rattling. Obvious that this results in bad CSD
Capaciti