That is completely incorrect. These axial aberations are caused by diffractions at the mouth. At some frequency the diffracted wave arrives at the measurement position out of phase with the direct wave (it has a different path length) causing a partial cancellation. There is no "phase flipping" of the tweeter and the on and off axis data are not "out of phase".
Yea Bill, thanks for that confirmation and the end note! Thats true of a lot of things. Take nonlinearities for example. A Klippel "type" system is great for measuring the nonlinearities that you can't hear (2nd, 3rd, etc.), but it doesn't do very well on the ones that you can (> 4th).
So we can put harmonic distortion toward the bottom of the list with phase for an audibly accurate speaker unless extreme or of high order?
http://www.gedlee.com/downloads/Distortion_AES_I.pdf
http://www.gedlee.com/downloads/Distortion_AES_II.pdf
Dr. Geddes, care to discuss impulse response?
Dan
http://www.gedlee.com/downloads/Distortion_AES_I.pdf
http://www.gedlee.com/downloads/Distortion_AES_II.pdf
Dr. Geddes, care to discuss impulse response?
Dan
People frequently fall into this trap.
These tests go back to a good paper by Peter Fryer of Celestion. He found that when you doubled Q the resonance peak had to be 3dB higher to be equally audible. Now lets assume we are talking about a physical system, either driver resonances or cabinet resonances. Usually if you vary Q you do it by modifying damping in some manner. (If you change mass or stiffness you would be changing resonance rather than Q, right?)
Now if you reduce damping enough to double Q the height of the resonance will rise by 6dB. It needs to be 3dB higher to have "audible parity" but it rose a further 3dB higher than that. Under these real scenarios of mechanical systems, higher Q will be more audible, the level at resonance will rise twice as fast as the perceptual requirement.
Generally, I can only imagine that electrical systems can vary Q without changing resonance level as Toole describes.
David
Earl, yeah, the waterfall thing was a disappointment. I had already thought the waterfalls were more flash than dash because of their susceptibility to changes in formatting being able to make them look good or bad, but thought they had more audible meaning than they apparently do (or don't). I had just previously developed another waterfall technique that calculated a 3-D plot of all the 'shaped toneburst energy storage' responses, like those from Linkwitz' burst tests, but done at all frequencies at once from an impulse response. Sadly, that one doesn't do much better on detecting the low-Q resonances, either. (It looks pretty neat, though, if that counts for anything)
Bill
I keep alluding to this, because I do not have the time to pursue it, but have you ever looked at Prony methods. I think that they would be ideal for impulse responses for exactly the reasons that you describe.
I keep alluding to this, because I do not have the time to pursue it, but have you ever looked at Prony methods. I think that they would be ideal for impulse responses for exactly the reasons that you describe.
In the test done in my program (and I presume, in Toole's tests), the response is altered by obtaining two signals, one unfiltered and one filtered through a 2nd order bandpass of a given Q, and summing them with various gains applied to the filtered tone. The amplitude of the response peak (or dip) is the controlled parameter in that test. The decay time of a tone applied at the center of the bandpass increases with Q, which is what stands out on a waterfall (as opposed to simply looking at it on a frequency response plot).
If anyone wants to play with the program to hear the effects of mixing in resonance tones, send a PM, and I can email a copy.
If anyone wants to play with the program to hear the effects of mixing in resonance tones, send a PM, and I can email a copy.
This is all true. The question is what happens in natural systems where Q is varied by adjusting damping. Double the Q and the peak level (at resonance) goes up 6 dB. It becomes 3 dB more audible. Halve the Q and it drops 6 dB which is perceptually 3dB better.
The damping of resonances is still a good thing.
David
When looking at waterfall plots, and different measurement systems alwyas come up with different shapes for some reasons, one needs to look at the rate of decay over time per frequency at good time resolution, and also how the peaks shift in frequency with time. It is very useful, but as with any other measurement methods, it's best used with other measurements in the diagnosis process.
If it is a wave guide then what I said about phase flip may not apply. The appearance is very similar to what a 1" dome tweeter does with no phasing lens. Any non-lens dome tweeter will flip phase based on its' diameter.
Still I would like to measure the time response of the transducer.
If a high Q resonance is in the driving system such as a dome tweeter or woofer that resonance will always be stimulated by the driving energy passing through the resonance mechanism. If a high Q resonance is in a coupled resonator such as the wall of an enclosure energy can only transferred to the resonator from an energy source emitting energy at the same frequency or a sub-harmonic of the resonant frequency.
In the case of the dome I had an Accuton ceramic tweeter with a 14dB resonant spike in the response at 1758Hz. Because that existed not as a coupled resonator but as part of the driving system it was always stimulated by any energy passing through the system and it always was very audible.
On the other hand a box wall with a Q of 30 which is not tuned to one of the equal tempered standard scale pitches will not likely even be audible in music. It will be audible when stimulated by the correct frequency. So there may be high Q resonances in a box wall that are off frequency and not be a problem. If the resonance of the wall is centered on a standard pitch frequency then it will probably be audible in music.
Dan, Im not sure what you are trying to figure out with the impulse?
It really is just a raw driver measurement that is used to create a whole bunch of more meaningful plots.
If you get use to looking at impulses measurements I guess you will start to notice different decays in drivers and you will start to notice when reflection start to show up, showing you where to set your gate.
Right now I believe you only have to know where to set the gate on the impulse then you need to start looking at FR plots on and off axis, you can then look at CSDs or Wavelets (found on this forum).
What software are you using to create your impulse measurement?
Heh, next time please take care to know what graphs you are referring too. Earls data mostly are wave guides.
As for Accuton ceramic tweeter, it's necessary to look at the waterfall plots at two scales. 0.4ms time range and 4ms time range, also look at impedance plots. It's also helpfull to look at near field response starting at 1mm, 5mm, 1cm 5cm, 10cm. If you have this kind of data, I'd be glad to look into it.
At the 1977 AES Convention in Paris, Carsten Thomsen summarized the main arguments and
misconceptions about phase, by pointing out that phase is an engineering unit - and thus not
necessarily something we can hear. But the influences of phase errors are many and clearly audible by
giving time smear, poor transient response, overload due to phase errors, distortion of distance
perception, confusion of stereo image, and change of tonal quality.
From here: Multidimensional Audio
by Henning Moller, Bruel & Kjaer
found here: http://www.linkwitzlab.com/multidimensionalaudio.pdf
misconceptions about phase, by pointing out that phase is an engineering unit - and thus not
necessarily something we can hear. But the influences of phase errors are many and clearly audible by
giving time smear, poor transient response, overload due to phase errors, distortion of distance
perception, confusion of stereo image, and change of tonal quality.
From here: Multidimensional Audio
by Henning Moller, Bruel & Kjaer
found here: http://www.linkwitzlab.com/multidimensionalaudio.pdf
The reflections I get. What a good impulse looks like and what a bad one looks like is what I don't get thankfully. Otherwise generated clean in doors data is a guess.
I use REW. Good software w/ a new version in the works and beta version released.
Thanks Doug!
Dan
I had assumed the polar plot exposed what type of speaker it was.😱 and that you were just thinking 1 inch driver.
Interesting notes on resonance.
Thanks,
Dan
- Status
- Not open for further replies.