Resonances hidden from waterfall response

While waterfall responses are widely used to detect resonances I doubt they can reveal all of the resonances. Some things come up in my mind:
  • resonances of coil, spider as they might produce little or no soundpressure and/or sound is radiated backwards not reaching the measuring microphone.
  • resonances on the surface of the cone which are out of phase and so sound may cancel before reaching the microphone. At lower freq even close distance of the microphone might not help or only partially.

One could argue that if the microphone cant catch the resonance our ears wont either. But I think all resonance are detrimental as they also alter the sound envelope at the start of the signal, not only at the decay.

Is it true that standard waterfall response fail to 'see' all resonances? Are there better or more sophisticated methods which can? Can the onset of the signal (instead of the decay) reveal resonances not shown by the waterfall response?
 
  • Like
Reactions: 1 user
Also missing is the detection whether distortion or noise. And also the interpretation, whether hear "friendly" or hear "hostile". Especially since hearing perceives material resonances very sensitively. A recognition, whether e.g. metal or wood or stone, is still not feasible by means of "waterfall-eye diagnosis".
Hearing sometimes needs distortion or noise to recognize more easily. Dead-damping is disadvantageous.
So just keep listening instead of looking;-) Knock tests are more productive: knock with different materials and masses and forces on the various components of the speakers. Most people would then perceive their 200 liter loudspeaker cabinet as a droning and flat (wood) sound - which would make them unknowingly look for fat and unclean sounding amplifiers;-)
 
Disabled Account
Joined 2008
While waterfall responses are widely used to detect resonances I doubt they can reveal all of the resonances. Some things come up in my mind:
  • resonances of coil, spider as they might produce little or no soundpressure and/or sound is radiated backwards not reaching the measuring microphone.
  • resonances on the surface of the cone which are out of phase and so sound may cancel before reaching the microphone. At lower freq even close distance of the microphone might not help or only partially.

1. If your mic don't pick it up you cant hear it either, so why bother?
2. This will show up as peaks and dips on the frequency response.
 
One could argue that if the microphone cant catch the resonance our ears wont either. But I think all resonance are detrimental as they also alter the sound envelope at the start of the signal, not only at the decay.

Is it true that standard waterfall response fail to 'see' all resonances?
Basically no, and the answer is in your first sentence.

Obviously, one singe waterfall (one of the many ways to display an Impulse Response) might not catch it all, though. You need to look off-axis as well etc.
Further, the parameters for a Waterfall or CSD or Wavelet or S-Transform plot do affect the visual representation very much so that it is easy to smooth out the data too much etc, resolution being too low, etc. Note that you all most always have to trade off frequency resolution against time resolution.

So in effect, one could say that if you have only one Waterfall plot in a magazine or review it might or might not have catched all relevant resonances in that they are "visible enough"... but it does not mean it's impossible per se. Best is to have the Impulse Response data in raw form (.WAV-file etc) and use your own analysis tool for display, I would suggest the well-reputed free REW tool (Room EQ Wizard). When the raw data is good (signal-to-noise) then you can vary parameters and zoom in at will, change observation viewpoint etc, to best expose small details.
 
Well, you can use an accelerometer for that. You have one in every phone these days.
Yes, accelerometers attached to voicecoils pick up more disturbances, breakups and parasitic resonances than a conventional SPL measurement by mic.
Are you suggesting a phone could be used? I would like to know more about that.

If not, what particular MEMS or Piezo component are you referring to?
 

TNT

Member
Joined 2003
Paid Member
WF is just a series of FR measurements as time passes. So you see the same thing in a WF as in a FR sweep. But more.... i.e. how the FR develops in time. If someone goes on inside the speaker and don't show in a FR/WF measurement, you won't hear it either.... there is really no magic...

//
 
  • Like
Reactions: 1 user
Disabled Account
Joined 2008
Yes, accelerometers attached to voicecoils pick up more disturbances, breakups and parasitic resonances than a conventional SPL measurement by mic.
Are you suggesting a phone could be used? I would like to know more about that.

If not, what particular MEMS or Piezo component are you referring to?
I was thinking of the phone itself. Useful for speaker cabinet measurements, but of course not very small or light objects like the voice coil of driver suspension …..
 
Member
Joined 2004
Paid Member
Monitoring the voice coil won't tell you what is happening at the diaphragm. The Laser stuff is much better for that. However knowing what to do when you see a problem is much harder that even seeing issues. Adding the mass of an accelerometer to the voicecoil will change what it does a little. This stuff is relevant to subwoofers but not for tweeters.

Some software will allow very high resolution watefall plots. Again lots of data but not a lot of intellegence. You would need to know what is important.
 
Member
Joined 2004
Paid Member
You can also see some of those in the impedance plot. Now, what do you do to fix them? Cut and glue the cone? Surface treatment? Different cone construction, thickness, material? Get a driver that doesn't have the problem? Sometimes life is easier if you weren't aware of the problem.
 
No discussion here: fixing the problem is quite an issue. Surface treatment does not do much i.m.o. Cutting (jbruner) or dimpling (Spangler,McKenzie) are much more promising.

I am afraid we will have to wait for diy accessible FEM software, as long as COMSOL costs a mansion and takes 5 years to learn, in order to take conology beyond trial and error.