Christian,
True. I also very much like the Wavelet Spectrgram. It does reveal resonances pretty clearly, and even more clearly with a close mic measurement than with the mic at greater distances.
I don't disagree. But one thing I want to emphasize is that my main purpose in sharing these methods is not to propose methods that replace, say, a standard frequency response curve, or even necessarily to tell you what a panel will actually sound like. Rather, my intention is to provide a tool that helps us to better assess the extent of damping that is introduced by any change in a design. Say you try a new coating on the panel, or a new type or arrangement of foam around the perimeter, or add a strip of CLD tape to the surface. You hear something different that you suspect is caused by an increase or decrease in damping. But did you really change the damping? How do you know? How can you verify it? How can you demonstrate it for the rest of us to see and understand?
So at that level, I don't really care if a particular peak in the impedance curve is a "productive" mode or not. I care only that the change in the sharpness of the peak in the impedance curve gives me a strong indication of the actual change in damping that the modified coating, or new foam, or whatever actually produced.
Eric
In the past two days I shared two methods of assessing the level of damping in a panel speaker build: an Impedance Test and then a Tap Test. Today I will describe the third test method, which I am call the Close Mic method.
In truth, the close mic method is really just a variation of the Tap Test, except that it uses an exciter as the "hammer". And as the name of the test implies, the mic is placed close to the face of the panel, and acts as the accelerometer. This test also uses REW , but this time in regular SPL frequency sweep mode.
One minor disadvantage compared to the tap test is that an exciter is required. The inclusion of the exciter also means that this test is much more difficult to use than a Tap Test to determine a panel's elastic properties, but that is not our purpose now anyway. But a nice advantage of this Close Mic test is all the views that REW provides for SPL sweep tests (spectrogram wavelet in particular), that are not available in an REW RTA measurement.
I think it's self evident by now, but the steps in the Close Mic Method are:
The five mountings are as follows:
1.Hanging Free: Panels were hung from wires attached with small alligator clips, about 5" from the two ends of the panel.
2. Four Points 3M Indoor: Panel attached at four points to a frame using 3M Indoor double sided foam mounting tape. The four points were along the two long edges, about 5" in from the corners. The tape pieces were each approximately 3/4" square. This foam tape is quite soft and rebounds quickly when compressed.
3. Full Surround 3M Indoor: Panel attached to frame around the entire perimeter, excluding about 2" from each corner. The width of tape in contact with the frame and panel was about 0.5" wide.
4. Full Surround 3M Extreme: Similar to directly above, except using 3M Extreme double sided mounting tape. This foam tape is much firmer, and has much slower rebound than the 3M Indoor used in mountings 2 and 3 above.
5. Full Surround Poron 92: Similar to 3 and 4 above except the panel was attached to the frame using Poron 92 foam (about 0.25" thick). This foam is very soft, with very slow rebound. The width of foam in contact with the panel and frame was in this case was about 1.25" wide.
The Close Mic SPL results (no smoothing) for those mountings are shown below.
The results show almost exactly the same thing as the Impedance Test and Tap Test. Each curve has peaks at each of the resonance frequencies of the panel assembly. The first two curves (from the top down) exhibit very sharp peaks, with a very slight increase in damping for the third curve (Full Surround, 3m Indoor). The lower two show much broader, rounder peaks, indicative of significantly increased damping.
But (as Christian alluded in a recent post) the Wavelet Spectrograms (below) show a similar, but perhaps even clearer story.
Free Hanging:
Four Points 3M Indoor
Full Surround 3M Indoor
Full Surround 3M Extreme
Full Surround Poron 92
In this representation, vertical "lines" appear at each of the resonance frequencies. The widening of the resonances due to damping is evident in these plots, particularly near the time=0 (near the bottom of each spectrogram). Note how SPL near t=0 is a much more uniform intensity (red color) for the final two mountings compared to the others. But the even more obvious indicator of damping (or rather lack thereof) is how far up the plots the vertical lines go. The length of those lines indicates how long after the driver "activates" a particular frequency that the panel continues to resonate (i.e. ring) at that same frequency. This characteristic (resonating time) is perhaps an even better indicator of damping than is the sharpness of the peaks in the frequency response curves. Being able to see this Wavelet Spectrogram representation is the biggest advantage of this method over either the Impedance Test or tap Test methods.
Eric
In truth, the close mic method is really just a variation of the Tap Test, except that it uses an exciter as the "hammer". And as the name of the test implies, the mic is placed close to the face of the panel, and acts as the accelerometer. This test also uses REW , but this time in regular SPL frequency sweep mode.
One minor disadvantage compared to the tap test is that an exciter is required. The inclusion of the exciter also means that this test is much more difficult to use than a Tap Test to determine a panel's elastic properties, but that is not our purpose now anyway. But a nice advantage of this Close Mic test is all the views that REW provides for SPL sweep tests (spectrogram wavelet in particular), that are not available in an REW RTA measurement.
I think it's self evident by now, but the steps in the Close Mic Method are:
- Mount an exciter to the panel. As in the other tests, the location you choose does effect which natural frequencies are excited and how effectively, but it really doesn't matter that much for the assessment of damping. Any reasonable place within the region of the center should do fine. In the tests shown below I used the 0.4/0.4 position.
- Place the mic very close to the face of the panel (within about 1/8" or 3 mm work well). In the tests below I placed it in the same 0.4/0.4 position as the exciter (but on the opposite side, obviously). Other places would likely work fine too.
- Run an SPL frequency sweep with REW. As you can imagine, due to the close placement of the mic to the panel you will need to turn the power input (volume setting) way down to avoid overloading the mic.
- Examine the SPL curve (no smoothing) and the Spectrogram (wavelet mode). I've gotten accustomed to the 1/6 Frequency resolution for the wavelet spectrogram, but others may also be useful. The impulse response may also be interesting.
The five mountings are as follows:
1.Hanging Free: Panels were hung from wires attached with small alligator clips, about 5" from the two ends of the panel.
2. Four Points 3M Indoor: Panel attached at four points to a frame using 3M Indoor double sided foam mounting tape. The four points were along the two long edges, about 5" in from the corners. The tape pieces were each approximately 3/4" square. This foam tape is quite soft and rebounds quickly when compressed.
3. Full Surround 3M Indoor: Panel attached to frame around the entire perimeter, excluding about 2" from each corner. The width of tape in contact with the frame and panel was about 0.5" wide.
4. Full Surround 3M Extreme: Similar to directly above, except using 3M Extreme double sided mounting tape. This foam tape is much firmer, and has much slower rebound than the 3M Indoor used in mountings 2 and 3 above.
5. Full Surround Poron 92: Similar to 3 and 4 above except the panel was attached to the frame using Poron 92 foam (about 0.25" thick). This foam is very soft, with very slow rebound. The width of foam in contact with the panel and frame was in this case was about 1.25" wide.
The Close Mic SPL results (no smoothing) for those mountings are shown below.
The results show almost exactly the same thing as the Impedance Test and Tap Test. Each curve has peaks at each of the resonance frequencies of the panel assembly. The first two curves (from the top down) exhibit very sharp peaks, with a very slight increase in damping for the third curve (Full Surround, 3m Indoor). The lower two show much broader, rounder peaks, indicative of significantly increased damping.
But (as Christian alluded in a recent post) the Wavelet Spectrograms (below) show a similar, but perhaps even clearer story.
Free Hanging:
Four Points 3M Indoor
Full Surround 3M Indoor
Full Surround 3M Extreme
Full Surround Poron 92
In this representation, vertical "lines" appear at each of the resonance frequencies. The widening of the resonances due to damping is evident in these plots, particularly near the time=0 (near the bottom of each spectrogram). Note how SPL near t=0 is a much more uniform intensity (red color) for the final two mountings compared to the others. But the even more obvious indicator of damping (or rather lack thereof) is how far up the plots the vertical lines go. The length of those lines indicates how long after the driver "activates" a particular frequency that the panel continues to resonate (i.e. ring) at that same frequency. This characteristic (resonating time) is perhaps an even better indicator of damping than is the sharpness of the peaks in the frequency response curves. Being able to see this Wavelet Spectrogram representation is the biggest advantage of this method over either the Impedance Test or tap Test methods.
Eric
Christian, I'm not sure I understand this question. Can you clarify?
Eric
Gentlemen, thanks for your insightful comments and observations. I'm trying to align your findings with my empirical tests.
Please comment.
Here are the curves, (taken at around 500mm where necessary) for SPL, impedance and spectrogram wavelet all for the same panel:
I have not done a tap test yet, but the fundamental resonance, (Fs, 68.6Hz) is obvious in the impedance curve.
How do the various curves relate to each other? I can see no relation between Fs, SPL, spectrogram or distortion figures.
Edit: Note that my spectrogram is limited to +500ms/-100ms and 1/24 resolution.
Last edited:
Yes sure. It is in fact a question aside the main object of your posts (evidence of damping). My questions started from my last series of tests about the suspension. For the configurations I tested, it seems the distortion was higher when the suspension is not all around the perimeter. So I was wondering if you have also this in your measurements.
It is just recently I have paid more attention to the distortion figures when I was not happy because of parasitic vibrations in a panel hang by 2 wires. Those parasitic vibrations are visible in the distortion view. They appear as a pic of distortion at the vibration frequency.
One of my conclusion at this step was the 2 wire suspension is not suitable. Too many vibrations go to the wires and even to the support. Maybe my suspension was too simple being a simple kitchen string.
So I adopted the suspension proposed by Steve which is a central ribbon of tape (one half on the front side, the second half on the rear side of the panel, a sponge between the 2 half ribbons of tape from the top edge of the panel). Here the distortion (or parasitic vibrations) was low enough.
Christian
Let me try to explain how I see those relations or more how I use those views (I am not a teacher or a doctor in physics or acoustics...)
First I would classify in 3 categories : the impedance measurement, the SPL (FR or spectrogram) measurement and the distortion.
For the impedance, the current create a force that moves the membrane. The reverse way exists also : a movement of the panel is visible at the terminals of the exciter in the electrical world. So you can observe in that curve some mechanical behaviors like the resonance of the magnet if the exciter is not supported by a spine, as Eric shown in the post before the damping or as B Zenker suggested but I haven't tested the coil former deformation in HF. The impedance curve helps to identify the problem in the FR and its cause.
The FR and the spectrogram give an information about the pressure level. For the FR if you don't apply more specific features, REW will show you the SPL over a constant amount of time which is called the window. All the presences of a frequency in this window are taken into account. You know the level over the window time but you have no information about when (and so why) the different contributions come. Typically, all the boundary reflections (floor, ceiling, walls) are included. Keep in mind (I can't explain why...) it is not possible to be precise in time and in frequency). The FR can be precise in frequency, not in time.
The spectrogram using wavelet has an other time/frequency balance. Less precise in frequency, a bit like when you choose 1/n octave smoothing of the FR. The observation window is frequency dependent, shorter when the frequency increases. You can see I think somewhere in REW they give the number of cycles (periods) of the wavelet. So with that, you have a view in frequency and in time. In the discussion about damping, the idea is to check the decay time at each frequency (how long to stop the vibration). In a spectrogram, you can also make a "slice" at a given time or frequency. I use it for example to have an idea of the spectral content of a reflection. For measurements at 1m, the floor reflection is around 4ms (peak in the IR). So make a slice at this time and you have an idea of the frequency content so the directivity of the panel. This shown me some panel material like EPS have an important level of HF out of axis.
The third category is the distortion. It is a specific graph in REW. I use it to detect "unwanted" contents in the response of the panel. It can be the distortion by itself (harmonic 2, 3 and above) so the non linearity of the panel which is generally in the low frequency due to the non symmetry of the motor. There is also a basic information in that view which is the noise floor (the noise coming from the background). A fan, a car in the street and, more difficult to detect some trouble in the electronic setup with the usb devices or laptop leading to harmonics of the main. The parasitic vibrations appear also is this distortion view : a loose nut, an undamped suspension string, an improper suspension termination.
I hope it will help a little.
Some aspects of your measurements are unfamiliar to me. Would you mind to share the REW file? The impedance curve is different from what Eric for example shared. The peaks in low frequency below 200Hz) are rather thin which is generally a synonym of low damping while there is almost no peaks in the mid or HF which is a synonym of high damping. Is this panel without spine to explain the 68Hz peak?
You can see here an example of the use of the different views : the 68Hz peak in the impedance curve is high, OK and then.... high peak means low damping so high displacement and so possibly distortion. I know there is probably an Xover that reduce the level at this frequency. It is just to give an idea how an information picked in a view may help in other views.
Christian
Andre,
Thanks for sharing your measurements. I have to say you give us a difficult task, for me at least!
One challenge is that the relationships between the Impedance, SPL, and Spectrogram Wavelet are most evident when the SPL and Spectrogram curves come from a close mic test. You may consider 500 mm to be "close", and it is compared to 1 meter, but if you really want to see the relationships, it is clearest when the mic is at 5 mm or closer, rather than 500 mm. I do realize that a measurement taken that close doesn't tell you much about what the far field frequency response will be. But it's also not the point. It does tell you a lot about what the response in the time domain will be, and it does so with minimal confounding effect from the room.
Second, you showed the Wavelet spectrogram at 1/24 resolution. At that resolution, every frequency looks like a resonance on just about any panel! Can you use 1/6 resolution, or is there some reason you cannot?
Also, I have not done any experimenting with the panel-on-canvas style construction, so this is outside my range of experience. For sure the impedance test is unambiguous to me, at least in the following respect: Every peak is a resonance of something! But what is the fundamental? What do we mean by fundamental? Maybe even that is a little ambiguous. Does it mean the lowest frequency where that is meaningful SPL output? Does it mean the lowest frequency natural frequency of the "panel"?
My best guess is that the five big peaks in the impedance curve are the resonaces associated with:
- the exciter resonance
- "pistonic" mode of panel in canvas frame
- first torsional mode of panel
- bending mode of panel in one direction
- bending mode of panel in the other direction.
See below for a correlation between the SPL at one meter, close mic SPL and Impedance Test for the panel in my examples with the Full Surround 3M Indoor. For every peak in the impedance curve there is a corresponding peak in the close mic SPL curve. There are some correlations to the SPL at 1 meter, but not nearly as much.
And below the overlay of the Tap Test and Impedance plots for the same build. The tap test shows no exciter peak at 27 Hz, (test performed without exciter) Also, the presence of the exciter shifts the panel fundamental up from about 54 Hz (tap test) to 61 Hz (Imp), but the remaining resonances are perfectly aligned for all practical purposes.
Eric
Thanks Eric, thanks Christian!
I'll fix the exciter in place to see if it makes any difference to the anomalies below 68hz.
Eric, on close mic measurements, do you line up the mic with the tap point?
I'll redo the measurements and then present at 1/6 resolution.
Hello André,
Do you have link to the source of the XT measured on acrylic? I had a look to their site but I didn't find it? Thank you.
About the previous topic : relation between the measurement, I didn't notice it was a canvas panel. The use of the different views remains but the physics behind is probably different from the plate based DML. Up to now, I haven't found a paper explaining how it works and its basic physics.
Christian
Andre,
Yes, usually. And if you want to see the correspondance with the close mic (sweep) test and the impedance test, tap and mic the spot where the exciter is (or will be).
But I'll add that one of the great advantages of the tap test over the others is that it is much easier to move your tap/mic position than it is to move your exciter. When you tap a particular spot you will see resonance peaks in your results only for modes that have antinodes in the vicinity of your tapping spot. Likewise, in the Impedance and close mic tests, you will only see resonance peaks for modes that have antinodes in the vicinity of your exciter location.
Some may consider this to be a purely academic excercise, but if want to find "all" the resonance frequncies of your panel build, you really have to probe around by moving the drive point (tap or exciter) to different spots on the panel. Which of course is easier to do with tapping than with an exciter. Also, as I mentioned previously, moving the exciter does shift the frequency of the lowest mode(s) slightly, so if you try this by moving the exciter around, when you find a "new" resonance at a slightly different frequency after moving the exciter, it's hard to know if it's a truely a different mode, or of it's the same one but just shifted a bit. When you tap (with no exciter, or without moving the exciter), the resonance modes don't shift around. So anything that pops up at a different frequency from previous test at different tap locations is truely a different mode.
Of course, you can never really find them all, because there are (likely) hundreds of them, and at high frequencies they are so close together so as to be indistinguishable. But below 500 Hz or 1000 Hz or so, the reonance mode are far enough apart that you really can "find" all of them by tapping various spots on the panel. Good places to tap include the very center, a corner, the center of the long and short sides, the center of each half (i.e. 0.25/0.5 and 0.5/0.25).
Eric
Christian,
I can't say that I have or have not seen that consistantly. I almost always put suspension around most of the perimeter, so I don't make the comparison with less suspension very often. That said, see below the distortion results (SPL freq sweep, 1 meter) for my five test cases with the carbon/balsa plate. Highest distortion seems to be the one with the four point suspension, and the ones with Full Surround have the lowest distortion. But I don't mean to say that this would always or even usually be true. I did not take particular notice of the distortion so it is possible that in the Free Hanging or Four Points tests that there was some source of distortion that I could have addressed. but the distortion wasn't so clearly audible in any of them that I felt the need to investigate/adress it for the purposes of these tests.
I have borrowed Steve's tape method for free plates on occasion, particularly for tap testing. I especially like that it doesn't damage face of the plates. I didn't use it in these tests, but for no particular reason.
Eric
Hanging Free
Four Points 3M Indoor
Full Surround 3M Indoor
Full Surround 3M Extreme
Full Surround Poron 92
Thank you Eric,
Even if it was not the goal of your tests, this seems to be a new indication in the direction of my observations. To be followed.
I don't know your intention for future tests but to include "Steve's suspension". The results I had with the 2 strings were pretty bad and much better with Steve's method.
I also tested a 3M double face tape (under the brand Scotch here) which is I think the 3M Extreme. The results from distortion point of view where not as good as with a D shape weather foam (see one of my post before). Is it possible you have a look to such a solution?
Christian
@Veleric thank you for sharing this, I will at least try the close-mic method next time I mess with the CLD. Since you've shown the impedance, tap test, and close mic results for those 5 samples, can you share the 1/24 FR now? The Poron sure looks nice in those tests but how does it compare in FR?
Pondering as ever, I ask,
1/ how much damping is too much damping (killing the liveliness of the sound), and can it be ascertained from a trace?, and,
2/ with the extent of testing being carried out, esp by Eric and Christian, are we edging closer to a set of principles, do's and don'ts, preferred materials etc etc to produce high performance panels?
Eucy
1/ how much damping is too much damping (killing the liveliness of the sound), and can it be ascertained from a trace?, and,
2/ with the extent of testing being carried out, esp by Eric and Christian, are we edging closer to a set of principles, do's and don'ts, preferred materials etc etc to produce high performance panels?
Eucy
Hi Christian,
I have not used cross-overs in any of these measurements.
I've taken several impedance curves for similar panels in different constructions. The first panel is stiffened Nidaplast by itself, with no damping, and is driven by an 8-ohm driver.
The frequency response and impedance response are taken with the panel hanging free, supported by one edge. No edge damping
Here it is again. This time the highlighted overlay is for a constrained magnet. It cannot move. The measurement is taken with the panel unsupported except by the driver itself:
This is the last one for the naked Np panel: the orange overlay is for a constrained panel this time, and the magnet is free to move. This shows the magnet's resonant frequency:
================================================================
This second set of curves is for a Nidaplast/canvas panel. The Nidaplast is stiffened as above, and glued to the inside of a tightly stretched canvas in a frame. There's a 15mm gap between the Nidaplast and the frame. In other words, the canvas acts like a 15mm flexible surround. This is with a 4-ohm driver, and delivers a similar SPL, except, because of the canvas frame, a better bass and HF. The scales are the same for all curves.
You can see the magnet resonance at 30Hz.
Here's the measurement with the magnet constrained:
And here again, with the panel constrained so the magnet/driver resonance is clearly seen:
================================================================
Lastly, here are the two FR curves overlaid. This is with panels and magnets unconstrained... everything is free to move.
Blue is 4-ohm on a Np panel in a stretched canvas frame.
Red is 8-ohm on an Np panel, unframed, unconstrained.
Last edited:
Andre,
It sure looks like the canvas is providing a significant amount of damping,
Eric
It sure looks like the canvas is providing a significant amount of damping,
Can you clarify, how is the panel constrained? Like you put it flat on a table or something, like with some weight on it?
And for the test with the NP on canvas, with the constrained magnet, how was the magnet constrained? With a slat attached to the back of the frame that holds the canvas?
Eric
Cut out from behind it almost entirely, leaving a 1/4" or 1/2" to glue to. I'm currently listening to some thin polystyrene panels with wood centers done in this fashion.
Yes, panel face down on the workbench and weighted around the driver so there's no vibration.
Correct. Sticky putty (we call it Prestik here) applied between the back of the driver and a brace across the frame. Adjusted so the VC is still in the centre of its travel.
I'm surprised that there's not a massive drop in efficiency with the amount of damping provided by the canvas.
One thing I want to confirm though, is to do all of the tests again using only 4-ohm, or only 8-ohm drivers. I doubt it will make much difference, but it's still good to confirm.
Last edited: