What about laying a speaker on it's side, covering it with cling film ( saran wrap to those in the land of Colt Seavers, the Dukes of Hazard and Knight rider ) sprinkling some sand on and making Chladni plate patterns. Probably wouldn't learn anything, but you'd get some pretty patterns an tell were the braces are, like when you see a bonnet ( sorry, hood ) of a car on a snowy day and you can see the brace under the bonnet ( hood ).
Well, if a panel isn’t actuated in a serious way, testing modal behavior might be a bit of overcomplicating things. But then again, we DIY-ers often don’t have a clue which panels we have to pay attention to. And I must confess, I do tap box panels occasionally. My feelings about a build count also...
I was thinking of using accelerometers. But if you use a microphone, you’d have to sum the sound pressure of a lot of measuring points to get a reliable idea about the sound power. As you do know of course.
Last edited:
I don't post much anymore or even read many posts. But I saw this thread and though I might comment. In the far field the sound pressure radiated by the driver's cone, at frequencies where the wave length is greater than the driver's diameter, is directly related to acceleration. Thus, assuming no cone breakup, if you measure the acceleration of the cone and compare it to a far field SPL measurement, any deviation of the SPL from the cone acceleration would represent sound radiated from other sources, such as the enclosure surfaces.
But rapping it with your knuckle won't be misleading?
And who says you can't do both?If you really want an easy way to learn about panel resonance and what braces do and don't do, build some enclosures out of foamcore. By preference the cheap, thin foamcore from the Dollar Store. It buzzes and sings like crazy, and adding braces moves that around, but it's hard to kill. If not foamcore, then 1/8" plywood works well.
Go to SparkFun website pick up $10 3axis accelerometer with buffered voltage outputs. Then to Dataq instruments and pick up starter kit for $60 bucks. Download Windaq software for free. Run data acq and then perform FFT function. Works quite well. You will need to play around a little to get accustomed to software. I used same thing in characterizing coater heads in industrial high tech app.
An impulse (knuckle rap) test provides less but largely complete information rather than partial information. This is what makes it less misleading and prompted my comment. However, this assumes a person understands what is being measured and the physics of things like modes, free and forced responses, etc... which I now suspect is not widely the case on this forum. So you may well be correct.
It is curious how patchy knowledge about the engineering of speakers seems to be among DIY folk. Passive crossovers for example are generally well understood with plenty of resources to build knowledge, the behaviour of dipole speakers is fairly well understood supported by resources like John Ks website, but something as fundamental as the cabinet seems to be remarkably poorly understood by most with widely held but false views on bracing, damping, isolation and how a speaker cabinet vibrates and radiates sound. Unlike other topics there seems to be almost no reliable source of information on the topic only a small number of incomplete and often partially incorrect and/or misleading ones. A case for a community project perhaps?
I do concur. Problem is that cabinet response on mechanical and acoustical stimulus isn't one-dimensional. Engineering hasn't evolved that much further than the cited Danish article also, which essentially boiled down to the brute force approach. Soundproofing engineering in other industries (car industry notably) uses this too. With all due respect, this is way out of bounds for most hobbyists (wasn't it you who sought for an usable BEM package?).
Having said that, there are a lot of 'rule of thumb' measures we can take and which work pretty well, in a way that the remaining problems become underlying to other issues (e.g. cabinet cavity damping or port damping). Be it that even in this ballpark things get mixed up (felt pads anyone?). So yes, let's do that.
I bought this one:
https://www.mouser.dk/productdetail/measurement-specialties/ach-01-03-10?qs=tcRzwG6oasZJpW836F3eAw==
Can work with the low voltage from a 3,5mm jack in any soundcard. Then just run REW and see what comes out of it.
I found a clear resonance in my cabinet, which could be temporarely fixed with EQ. But the cabinet i flawed in some way. But now I can document what is going on.
Easiest thing would be to just build a good cabinet to begin with. But you live and learn
https://www.mouser.dk/productdetail/measurement-specialties/ach-01-03-10?qs=tcRzwG6oasZJpW836F3eAw==
Can work with the low voltage from a 3,5mm jack in any soundcard. Then just run REW and see what comes out of it.
I found a clear resonance in my cabinet, which could be temporarely fixed with EQ. But the cabinet i flawed in some way. But now I can document what is going on.
Easiest thing would be to just build a good cabinet to begin with. But you live and learn
Clever. But do you have the nerve to sit for a blind A-B test to see if your fix is audible?
Now, I might point out that any EQ might be audible, whether instituted to repair a cab vibration or otherwise. But it might reveal that you the vibration isn't audible (although if you can hear the EQ, again might not be the vibration really).
B.
Now, I might point out that any EQ might be audible, whether instituted to repair a cab vibration or otherwise. But it might reveal that you the vibration isn't audible (although if you can hear the EQ, again might not be the vibration really).
B.
Wow! Very interesting! Makes total sense.
Getting the itch again, or is this a one off?
And do we even have measuring equipment available that is precise enough to separate out the wheat from the chaff, as it were?
Measuring cabinet signature - B&W approach
What is the sound radiating from a speaker cabinet? How much, and at what frequency? At what point does it become audible with program material?
It is a question that comes up repeatedly, and the same points and counterpoints are rehashed. … and no wonder, it is a complicated subject, with too little data. I think that there are two reasons we struggle with understanding the unwanted sound which may or may not radiate from a cabinet.
(1) The cabinet sound is very difficult to directly measure. The sound from the driver(s) usually overwhelms the radiation from the cabinet, so it is hard to segregate the two. Augerpro’s thread has shown that it is possible, but only by constructing dedicated experimental boxes. It does not seem to be possible to take an existing speaker system and measure it in-situ, and simply determine how much of the sound output is due to cabinet resonance.
(2) It is even more difficult to assess cabinet signature with real program material. Swept sinewave testing can certainly reveal resonances, but this testing does not distinguish audible from inaudible sounds. The amount of energy in program material is maximum in the 200 – 300 Hz range, and it decreases rapidly above that. It decreases below that as well, but not as rapidly. It takes energy to excite cabinet resonance. There is evidence that high Q cabinet resonances are less audible because program material is less likely to excite them. One of the many cabinet design philosophies is to make cabinets stiff enough so that resonances are pushed up high (where there is less energy), and simultaneously making those resonances higher Q so they are more difficult to excite. I myself have used this approach, and I advise others to use it as well… But how effective is this, really? It is hard to say without some kind of measurement, and we just don’t have any using real program material. The prototypical BBC approach is to dampen resonance and lower the resonance frequency because low frequency cabinet resonances are less audible, according to some research. How effective is this…? Again, it is hard to say.
This AES paper from B&W may offer an interesting new way of measuring cabinet radiation. This was posted by Augerpro in post #97 of this thread
https://www.diyaudio.com/forums/mul...ybe-easy-bracing-question-10.html#post6371314
What B&W did, which was very innovative, was their test method of isolating cabinet acoustic radiation from the much larger driver radiation: They first demonstrated that above Fs, the cone is decoupled from the motor structure. Then by analysis they showed that the virtually all of the force transmitted to the cabinet comes from the motor alone, transmitted through the driver chassis. This knowledge allowed them to cut the cone away from the voice coil. The motor, voice coil, and spider remain intact. They were doing this to validate their FEM and other numerical methods, but it is a very interesting technique.
Now they have a driver which produces very little acoustic radiation, but which can induce fully representative vibration forces into the cabinet. This solves what I believe is the biggest hurdle with measuring cabinet radiation.
So now we can take an existing speaker system and test it. It will cost us a sacrificial driver (nothing is free it seems), but we can get a pretty definitive answer as to what kind of sound is coming from the cabinet.
Step 1 would be to make a frequency sweep of the speaker with a fairly high voltage, something like 5 V. This FR would establish the baseline. Step 2 we cut the cone out of the driver, leaving the dust cap and spider untouched. We need to replace the mass of the cone, so we add a little mass to the dust cap (silicone caulk) until the Fs is back were it was originally. Step 3 we make a sine sweep at the same voltage level. Since there is no cone, the driver will make little sound, but the full vibration energy of the motor will remain, and it will be transmitted into the cabinet. Resonances will now be easily revealed and measured. If, for example, a strong resonance at 500 Hz produces 85 dB, and the baseline measurement of the intact driver was 95 dB at 500 Hz, then we know the cabinet is 10 dB below the driver. Step 4 would be (ouch) buy a new driver.
It will be necessary to rotate the cabinet and take FR from all four sides and above. I would conservatively assume that maximum output from any direction is the representative signature at that frequency.
What about program material? Well it would be interesting to play real program material and test the assumption that high frequency high Q resonances are rarely excited. This technique would make that pretty straightforward. The theory is that an 800 Hz high Q resonance is much less intrusive than a low Q (well damped) 200 Hz resonance. Testing a high stiffness cabinets and lower stiffness CLD cabinets with real program material will reveal the truth, at least for those two cabinets and that program material…
Thoughts?
J.
What is the sound radiating from a speaker cabinet? How much, and at what frequency? At what point does it become audible with program material?
It is a question that comes up repeatedly, and the same points and counterpoints are rehashed. … and no wonder, it is a complicated subject, with too little data. I think that there are two reasons we struggle with understanding the unwanted sound which may or may not radiate from a cabinet.
(1) The cabinet sound is very difficult to directly measure. The sound from the driver(s) usually overwhelms the radiation from the cabinet, so it is hard to segregate the two. Augerpro’s thread has shown that it is possible, but only by constructing dedicated experimental boxes. It does not seem to be possible to take an existing speaker system and measure it in-situ, and simply determine how much of the sound output is due to cabinet resonance.
(2) It is even more difficult to assess cabinet signature with real program material. Swept sinewave testing can certainly reveal resonances, but this testing does not distinguish audible from inaudible sounds. The amount of energy in program material is maximum in the 200 – 300 Hz range, and it decreases rapidly above that. It decreases below that as well, but not as rapidly. It takes energy to excite cabinet resonance. There is evidence that high Q cabinet resonances are less audible because program material is less likely to excite them. One of the many cabinet design philosophies is to make cabinets stiff enough so that resonances are pushed up high (where there is less energy), and simultaneously making those resonances higher Q so they are more difficult to excite. I myself have used this approach, and I advise others to use it as well… But how effective is this, really? It is hard to say without some kind of measurement, and we just don’t have any using real program material. The prototypical BBC approach is to dampen resonance and lower the resonance frequency because low frequency cabinet resonances are less audible, according to some research. How effective is this…? Again, it is hard to say.
This AES paper from B&W may offer an interesting new way of measuring cabinet radiation. This was posted by Augerpro in post #97 of this thread
https://www.diyaudio.com/forums/mul...ybe-easy-bracing-question-10.html#post6371314
What B&W did, which was very innovative, was their test method of isolating cabinet acoustic radiation from the much larger driver radiation: They first demonstrated that above Fs, the cone is decoupled from the motor structure. Then by analysis they showed that the virtually all of the force transmitted to the cabinet comes from the motor alone, transmitted through the driver chassis. This knowledge allowed them to cut the cone away from the voice coil. The motor, voice coil, and spider remain intact. They were doing this to validate their FEM and other numerical methods, but it is a very interesting technique.
Now they have a driver which produces very little acoustic radiation, but which can induce fully representative vibration forces into the cabinet. This solves what I believe is the biggest hurdle with measuring cabinet radiation.
So now we can take an existing speaker system and test it. It will cost us a sacrificial driver (nothing is free it seems), but we can get a pretty definitive answer as to what kind of sound is coming from the cabinet.
Step 1 would be to make a frequency sweep of the speaker with a fairly high voltage, something like 5 V. This FR would establish the baseline. Step 2 we cut the cone out of the driver, leaving the dust cap and spider untouched. We need to replace the mass of the cone, so we add a little mass to the dust cap (silicone caulk) until the Fs is back were it was originally. Step 3 we make a sine sweep at the same voltage level. Since there is no cone, the driver will make little sound, but the full vibration energy of the motor will remain, and it will be transmitted into the cabinet. Resonances will now be easily revealed and measured. If, for example, a strong resonance at 500 Hz produces 85 dB, and the baseline measurement of the intact driver was 95 dB at 500 Hz, then we know the cabinet is 10 dB below the driver. Step 4 would be (ouch) buy a new driver.
It will be necessary to rotate the cabinet and take FR from all four sides and above. I would conservatively assume that maximum output from any direction is the representative signature at that frequency.
What about program material? Well it would be interesting to play real program material and test the assumption that high frequency high Q resonances are rarely excited. This technique would make that pretty straightforward. The theory is that an 800 Hz high Q resonance is much less intrusive than a low Q (well damped) 200 Hz resonance. Testing a high stiffness cabinets and lower stiffness CLD cabinets with real program material will reveal the truth, at least for those two cabinets and that program material…
Thoughts?
J.
Last edited:
The B&W paper is poor. It came up earlier w.r.t. the simulations which contain a gross systematic error in failing to correctly predict the easiest largest scale resonance of the speaker nodding on its feet. One can speculate how but if this is wrong what else is?
I don't think the measurements presented are what you seem to think. If you look at the photo of the drivers without the cone the suspension remains and is obviously going to radiate a lot of sound. Less than the cone but far more than the cabinet. If you look at the presented measurements they are not taken with a microphone but a laser vibrometer. They measured the deflection over the baffle and the side panel and integrated to get the SPL. This is the normal way in industry (and in practise likely the only reasonably accurate and reliable way) to obtain the cabinet radiation. There is of course no need to mess about removing driver cones to do this.
Conclusion: the intended experimental approach was so flawed the results could not be presented but the authors wanted to publish something regardless. Their supervisors/managers let them despite what it might portray about the technical competence of the company.
My comment was just a thought. May not be practical. As noted the problem is that the acoustic output from cabinet surfaces is generally well below the driver's output. Right there, if that's the truth, it's not a big issue. But if it were significant, using a 2 channel measurement system you could attach the accelerometer to the driver and then use it's signal as a calibration signal for the SPL. The SPL measurement has to be anechoic. But given that both mic and accelerometer would be pretty flat at low frequency the result should be a flat curve at low frequency. Any sharp discrepancies in amplitude would be an indication of cabinet resonances, or so I would hope. Well, you would still have the baffle step in there too.
But if you really want to isolate panel resonances the way to do it would be with a laser scan of each surface. You could generate a plot of x-y position on the surface with amplitude that evolved as a function of frequency, f(t). Capture it on video and then play it back to watch the resonances evolve (in time) as the frequency changes. Like looking at the surface of a lake as the ripples change.
Maybe a simpler way would be sprinkle salt on the panel and see what happens. As soon as I though of that I figured someone on YouTube would have done it. Sure enough, here is a video showing doing that. Amazing Resonance Experiment! - YouTube
Doesn't give amplitude info but shows panel modes.
But if you really want to isolate panel resonances the way to do it would be with a laser scan of each surface. You could generate a plot of x-y position on the surface with amplitude that evolved as a function of frequency, f(t). Capture it on video and then play it back to watch the resonances evolve (in time) as the frequency changes. Like looking at the surface of a lake as the ripples change.
Maybe a simpler way would be sprinkle salt on the panel and see what happens. As soon as I though of that I figured someone on YouTube would have done it. Sure enough, here is a video showing doing that. Amazing Resonance Experiment! - YouTube
Doesn't give amplitude info but shows panel modes.