What surprises me is that the phase stays flat while the response dips at 10KHz. Correct me if I'm wrong but this is not minimum-phase behavior.
Or maybe it is more accurate to time align the impulse to the HF as in the attachment? Then the phase is mostly zero until just before 10KHz. I wonder which alignment is more useful. Perhaps this is more realistic since no frequencies will be arriving before the impulse.
Or maybe it is more accurate to time align the impulse to the HF as in the attachment? Then the phase is mostly zero until just before 10KHz. I wonder which alignment is more useful. Perhaps this is more realistic since no frequencies will be arriving before the impulse.
Attachments
Last edited:
REW lets me set T=0 to a position on the impulse. I use the group delay plot to set that point at either of the two positions the group delay is flat. When I do that the phase changes. I wonder if some of this just has to do with convention. Some people may set T=0 at the root of the impulse, in the middle, or at the peak. I think the group delay makes the most sense as an indication of where T=0 should be.
Here is a picture. I can't think of a less obstructive way to hold a capsule.
Here is a picture. I can't think of a less obstructive way to hold a capsule.
Attachments
Last edited:
The literature says you need to support for the capsule coming out the back. Apparently it doesn't take much even with that to affect the response (check the link I provided earlier).
Also most of what I have read indicated that the initial point of change is the correct reference. What is somewhat crazy is that he acoustic center of the microphone changes with frequency. Its not a lot but the kind of thing that the metrologists obsess over. It may also explain what you are seeing with phase vs. time. This link may help even though its talking about MEMS microphones https://www.knowles.com/docs/defaul...icrophones---theory-and-practice.pdf?sfvrsn=4
Also most of what I have read indicated that the initial point of change is the correct reference. What is somewhat crazy is that he acoustic center of the microphone changes with frequency. Its not a lot but the kind of thing that the metrologists obsess over. It may also explain what you are seeing with phase vs. time. This link may help even though its talking about MEMS microphones https://www.knowles.com/docs/defaul...icrophones---theory-and-practice.pdf?sfvrsn=4
I've mentioned the acoustic center in several posts. With the damping applied to the trafo, the impulse is clean enough to clearly see the different acoustic centers. Attached. Is this not a useful indication of where the time center of the impulse is?
Once the calibration is applied to a measurement, the acoustic center is corrected in the measurement, so technically it should not matter if you set the acoustic center 10cm in front of the microphone for the calibration, the group delay in the measurement will be flat, just offset by 10cm. The question is where you specifically think the acoustic center is, if it is different from the result given by the FFT. You may set the acoustic center to be the face of the microphone, so you can locate the microphone face 1 meter away, but the FFT may disagree by a few mm.
If I use the bamboo skewer for all my measurements (extremely likely unless something better appears) then the calibration is valid for my usage. And anyone else could easily use the same setup for cents.
Once the calibration is applied to a measurement, the acoustic center is corrected in the measurement, so technically it should not matter if you set the acoustic center 10cm in front of the microphone for the calibration, the group delay in the measurement will be flat, just offset by 10cm. The question is where you specifically think the acoustic center is, if it is different from the result given by the FFT. You may set the acoustic center to be the face of the microphone, so you can locate the microphone face 1 meter away, but the FFT may disagree by a few mm.
If I use the bamboo skewer for all my measurements (extremely likely unless something better appears) then the calibration is valid for my usage. And anyone else could easily use the same setup for cents.
Attachments
I can even move the FFT window over the spark event E-field, which appears 162us before the acoustic wave hits, so I know that according to the speed of sound, the microphone is 2.19" away from the spark. How much positional accuracy do you want? How much certainty do you have in the speed of sound?
Attachments
The Knowles paper seems to agree with setting the time center to the earliest group delay.
What this means is that it makes sense to set the group delay to the the flat HF region. When I do this I get the phase in post #261. The phase is negative and that is expected, but it is lower than expected.
My thinking is that the holes in the face of the capsule may affect the acoustic impedance of that space at wavelengths that approach the width of the capsule. As a restriction they would convert pressure to velocity just before the membrane.
#264 said:
What this means is that it makes sense to set the group delay to the the flat HF region. When I do this I get the phase in post #261. The phase is negative and that is expected, but it is lower than expected.
My thinking is that the holes in the face of the capsule may affect the acoustic impedance of that space at wavelengths that approach the width of the capsule. As a restriction they would convert pressure to velocity just before the membrane.
Here is a curve for the Primo EM264 cardioid bare capsule at 11" away. Again a gradual rise in the treble appears in the datasheet but is nowhere to be found in my measurements. My measurement could be hiding the rise, but then that does not explain why the EM272 and EM258 have about the same response.
Thinking about minimum phase.
When REW generates a minimum phase curve from the response, the phase is -90 degrees at the corner frequency of the capsule. This is equivalent to shifting the acoustic center 2mm, and places both group delay plateaus after T=0.
So we have 2 different ways of deciding on the position of T=0 for phase. We can place it at an acoustic center where group delay is flat, or we can place it so that the response obeys minimum phase.
In the case of the EM258 the only difference between minimum phase and acoustic center is a time delay distance of 2mm.
Thinking about minimum phase.
When REW generates a minimum phase curve from the response, the phase is -90 degrees at the corner frequency of the capsule. This is equivalent to shifting the acoustic center 2mm, and places both group delay plateaus after T=0.
So we have 2 different ways of deciding on the position of T=0 for phase. We can place it at an acoustic center where group delay is flat, or we can place it so that the response obeys minimum phase.
In the case of the EM258 the only difference between minimum phase and acoustic center is a time delay distance of 2mm.
Attachments
Here are some measurements from the B&K manual. It was created in 1966. The later manuals are much less informative. In any case the Phase and impulse response are from an electrostatic actuator. Still they are valid reference points.
"3. Phase Characteristics and Pulse Responses.
Typical pressure response phase angle characteristics of the quarter-inch
microphones are shown in Fig. 1.13. The fundamental resonance of the diaphragm in vacuum occurs at about 50 kHz, but when air at normal
pressure is present between back plate and diaphragm the center of the
diaphragm is virtually blocked and the first resonance appears at a highe1
mode around 70 kHz.
Fig. A.6 in Appendix shows some examples of pulse responses."
"3. Phase Characteristics and Pulse Responses.
Typical pressure response phase angle characteristics of the quarter-inch
microphones are shown in Fig. 1.13. The fundamental resonance of the diaphragm in vacuum occurs at about 50 kHz, but when air at normal
pressure is present between back plate and diaphragm the center of the
diaphragm is virtually blocked and the first resonance appears at a highe1
mode around 70 kHz.
Fig. A.6 in Appendix shows some examples of pulse responses."
Attachments
That is probably the most useful. There aren't many situations where that distance and time difference would be meaningful.
Now that you have the process down can you describe the steps such that I and others can reproduce them? I have the spark. I need to know how you are recording and post processing the recordings.
Now that you have the process down can you describe the steps such that I and others can reproduce them? I have the spark. I need to know how you are recording and post processing the recordings.
Mathematically the peak of the impulse response is T=0 so you would be correct, that is reducing the problem to the ideal case an impulse at 0 gives a zero phase for all frequencies. Here the impulse response is computed by integrating the doublet and flattening the mid region phase should be very close and not matter here. For use for this application it's not likely a tiny difference could matter.
The first figure in the Knowles article simply shows that a straight line phase proportional to frequency is a delay and the second figure shows what is simply a minimum phase response with a single high pass pole (90 degrees) going to zero phase (any delay removed) until resonance.
If you want to do exactly what I am doing you will be making a spark train with the sparker and letting my bash script process the recording into a single averaged file in Linux, with guidance from the scatter plot to minimize error. The rest is inspection and windowing in REW.
PM me your email address and I can send you the script, when we get that working I can show you the process of wrangling the impulses in REW, but you probably only need help learning the interface.
The minimum phase setting places T=0 nearly at the starting root of the impulse, which is where the doublet would be if the microphone had a perfect response. The HF limit of the microphone places the group delay behind the face of the capsule. So when mic engineers say the acoustic center "moves forward" at HF they might mean it moves closer to the membrane, not in front of it.
There is some confusion here, what is the starting root of the impulse? Any true delay is non-minimum phase (group delay and phase delay are not the same thing). If you model a BW limited impulse as an ever decreasing in width Gaussian pulse the only setting that is minimum phase is with the peak at T=0.
I mean the start of the impulse where it departs from zero. The HF BW limit spreads the impulse out across later time, so the initial impulse tends toward the beginning of the mic impulse response. When T=0 is set for minimum phase, it's placement on the impulse is about where the doublet should be.
Group delay includes both minimum phase and non-minimum phase behaviors, so the question is whether the minimum phase delay from the mic response counts in the position of the acoustic center. Did anyone ever define whether the acoustic center was minimum phase or not? Because the peak energy occurs later than the physical delay due to the mic response.
If the mic response were perfect, the distance between the mic and source would be the distance between the peaks of the impulses when emitted and when captured. But the group delay of the mic response means that the peak is a bit further than that. Setting T=0 for minimum phase places it where the peak would be if the microphone were perfect.
It seems this question may have even stumped the AES at one time:
Determination of Acoustic Center -
Techtalk Speaker Building, Audio, Video Discussion Forum
When the step in group delay is so obvious in the mic response, it's hard not to think that the acoustic center could be accurately determined.
Group delay includes both minimum phase and non-minimum phase behaviors, so the question is whether the minimum phase delay from the mic response counts in the position of the acoustic center. Did anyone ever define whether the acoustic center was minimum phase or not? Because the peak energy occurs later than the physical delay due to the mic response.
If the mic response were perfect, the distance between the mic and source would be the distance between the peaks of the impulses when emitted and when captured. But the group delay of the mic response means that the peak is a bit further than that. Setting T=0 for minimum phase places it where the peak would be if the microphone were perfect.
It seems this question may have even stumped the AES at one time:
Determination of Acoustic Center -
Techtalk Speaker Building, Audio, Video Discussion Forum
When the step in group delay is so obvious in the mic response, it's hard not to think that the acoustic center could be accurately determined.
I guess this is sort of the same as asking, if a subwoofer has a 1ms group delay, is the acoustic center 13.5" behind the sub? That would mean that the acoustic center is always on the opposite side to the listener. So does the acoustic center change depending on the position of the listener or not?
It's tempting to say the obvious answer is no, that's absurd. But if we're talking about a midwoofer in a 2-way speaker, the group delay of the woofer still matters for the alignment. If you take the group delay as part of the response, then you must be guessing the woofer's phase by deriving it from the response, unless you can measure absolute phase.
Part of this is merely semantic, as a speaker can still work whether we define the acoustic center as minimum phase or not. But in the discussions I've seen, that part of the definition seems to be missing.
It's tempting to say the obvious answer is no, that's absurd. But if we're talking about a midwoofer in a 2-way speaker, the group delay of the woofer still matters for the alignment. If you take the group delay as part of the response, then you must be guessing the woofer's phase by deriving it from the response, unless you can measure absolute phase.
Part of this is merely semantic, as a speaker can still work whether we define the acoustic center as minimum phase or not. But in the discussions I've seen, that part of the definition seems to be missing.
Last edited:
I understand that the phase is flat when the impulse is centered.
It's just odd that if the acoustic center is minimum phase then it must be locked at the diaphragm of the mic, with out any movement, and it is not the center of the impulse.
Whereas if the acoustic center is defined as the center of the impulse, then the mechanical and electrical group delay of the microphone always places it on the opposite side as the sound source. And it is no longer minimum phase. And it no longer obeys the speed of sound.
It's just odd that if the acoustic center is minimum phase then it must be locked at the diaphragm of the mic, with out any movement, and it is not the center of the impulse.
Whereas if the acoustic center is defined as the center of the impulse, then the mechanical and electrical group delay of the microphone always places it on the opposite side as the sound source. And it is no longer minimum phase. And it no longer obeys the speed of sound.
Last edited:
This is obsessive enough for anyone looking for microphone acoustic center's: https://backend.orbit.dtu.dk/ws/portalfiles/portal/4302125/Barrera-Figueroa.pdf In the paper the 1/2" (12.5 MM) microphone starts with its center at almost +9 mm and slowly shifts to -1 mm at 15 KHz. Its not a lot. Its enough to bug those obsessives (.03 dB??). Maybe this can be extrapolated to the 1/4" capsules your exploring? At least is may help explain your measurements.