Spark calibration of microphones

Member
Joined 2004
Paid Member
Well sort of. . In talking to them they said they calibrated their spark setup with a Gefell 301. You have done some work around the spark calibration and I remember you saying its not a perfect impulse. What is it? I'm looking at setting up something similar and interested in any helpful suggestions.

Even with the spark you still need to compensate for the acoustics in terms of acoustic losses etc. Small at audio but big at 100 KHz.
 
Last edited:
Well sort of. . In talking to them they said they calibrated their spark setup with a Gefell 301. You have done some work around the spark calibration and I remember you saying its not a perfect impulse. What is it? I'm looking at setting up something similar and interested in any helpful suggestions.

Even with the spark you still need to compensate for the acoustics in terms of acoustic losses etc. Small at audio but big at 100 KHz.

You need to look up N wave theory, there are many references. The point is the sparkgap is smaller than any acoustic source you could make so the wavefront can be computed from first principles. For relative intensity there should be no need to calibrate, the problem is that at the high end there are corrections needed for spark energy and at the low end you need to have no reflections. I wasn't too hard to set up a jig where you could get down to 500Hz or so and just over 20kHz with and ordinary piezo spark like a grill or oven starter. It compared better than +- 1dB to a calibrated 1/4 capsule. Over most of the range I would say +-0.25dB.
 

obh

Member
Joined 2016
Paid Member
Or use a spark coil from a car, 12 V from a computer PSU/transformer/temporarily disconnected car battery etc, a 555 timer connected to a driver transistor driving the HV transistor capable well over 1000 V (the primary sideof the coil, at least the ones I have played with, shoots horrendous voltage spikes every time it's flywheeling!) connected to the coil, provides for a bit better predictability of the spark repetition and frequency.
 
Last edited:

obh

Member
Joined 2016
Paid Member
To calibrate a microphone solely by a spark source seems to be a somewhat
difficult task.

Acoustic energy in sparks from two spheres seems to be highly repeatable
though with an deviation of 3% of breakdown voltage, see http://downloads.hindawi.com/archive/2014/980913.pdf. Energy released is 1/2cv^2 if path to source is
non-inductive. And I think I read somewhere that about 1.2% of that goes to acoustic energy.

Yuldashev et al https://acoustique.ec-lyon.fr/publi/karzova_jasa15b.pdf writes that it is very difficult to measure
the actual peak pressure of initial shock wave - they measure 10% uncertainty while using high-speed camera to measure sub-microseconds rise time. (Microphones not up
to task). So it seems difficult to use the spark energy as direct input to numerical
models of N-shape propagation?

Another method was to take microphone measurements at different radius
(between 1-3 m from source) and then using some math to compute
the N-wave and from there the frequency response of the microphone.
But would this be feasibly with a 192KHz sound card?
 
Just my 2c of amateur pov thoughts that, spehre's should not be used but rather very pointy electrodes like two needles, this should reduce significantly the randomness from where the spark will leap and thereby varieties in acoustic energy, secondly one could also try evaluate using a relatively very small gap down to only a single millimeter, that I think would also reduce significantly air imperfections affect and sensitivity to drag on the spark quality and consistency, a small gap will of course also produce a much lower acoustic output allowing one to keep the DUT mic much closer to the gap without the risk of saturation, even as close as only a few single centimeters away from the gap (although EMI might be of a concern too, needs to be checked), which in turn also should reduce the damping effect on higher frequencies of the acoustic signal produced by the spark as air has a low pass filter effect attenuating higher frequencies more than lower frequencies, plus perhaps a few other variables as well.

What is left is to solve is how to figure out the absolute acoustic energy measured, but the relative output still can tell us, for instance, if the above assertion holds true producing a spark of higher consistency and fidelity by means of an empiric test, and if so then serve as a suitable source for calibrating the mic.
Again, these are my non-educated "guesses", would be good to hear from others who knows better as I find the "spark calibration" interesting for my own DIY project as well.
 
Last edited:

obh

Member
Joined 2016
Paid Member
When electrode is sphere with radius larger than gap, then breakdown voltage is more consistent.
Measuring High Voltage In Millimeters (and Other HV Probe Tricks) | Hackaday.

sphere_length_voltage_chart.jpg
 
Member
Joined 2004
Paid Member
Lots of interesting ideas for spark generators. I happen to have this box, a Honda (really) impulse noise generator for testing AC power stuff. It has the right parts to make a pretty serious spark. In the middle is a vacuum HV relay and seems to be good for 20 KV and 10 or 20A at least. My current plan is to modify the relay drive so I can run it from my software to trigger it whci will enable much better data acquisition and analysis.

Also current plays a major part in how loud the spark is. its really energy in to energy out. And I can attest that a 100 uS spark from a big surge generator (3KA) can be really loud. I hope to get something not quite so serious. I do want a decent SNR for the testing.

I did look up the Taser like device. It can be had for as little at $15 and it could be interesting but not triggerable Ais seems. I will also get a gas grill sparker but again not triggerable.

My current plan is to use a surface gap spark plug for the spark (illustrated below) so I can have a stable predictable gap with no local resonant cavities. It seems Mercedes has shielded connectors for the spark plugs to control EMI and I'm chasing one of those as well so I can minimize the inductance in the loop.

Its a somewhat involved project. It should be interesting. If it can be transferred to a simpler solution using cheap parts (the grill sparker is $8 or so) that would be great.
 

Attachments

  • Honda impulse Generator.jpg
    Honda impulse Generator.jpg
    63.3 KB · Views: 472
  • Surface Gap Plug.PNG
    Surface Gap Plug.PNG
    36.8 KB · Views: 481
When electrode is sphere with radius larger than gap, then breakdown voltage is more consistent.
Measuring High Voltage In Millimeters (and Other HV Probe Tricks) | Hackaday.

I did read the article, I see that he uses a CRT HV supply, most likely a diode-capacitor multiplier, such device ramps up slowly and may be contributing to the inaccuracy such as the build up of the ionization cloud in front of the tip of the HV electrodes which can be sensitive to things such as breathing on it but also variation in mains voltage supply, however, with respect to the accuracy I can't see much in that article which would be of concern for us if using a spark circuit as I described earlier using a spark coil where such device is producing the spark dictated by the timer, not when the condition for the spark to leap by itself is seen as fit.
 

Attachments

  • sphere_length_voltage_chart.jpg
    sphere_length_voltage_chart.jpg
    91.2 KB · Views: 430
I think it is important to note that every time you spark you are blowing craters in the electrodes and oxidizing/carbonizing the surface there. I think that should count toward some uncertainty. I wonder about generating the spark in an inert atmosphere?

Paschen's law shows that spark formation is a statistical process, so I would think that volts per gap would show a correlation with uncertainty, assuming perfectly consistent voltage supply.

In the references given, I see that short (say 1mm) spark gaps are necessary to give a measurement BW above 20KHz. Furthermore higher energy sparks are also lower BW, simply because it takes longer to discharge more energy from a given tank circuit. Bigger isn't always better, depending on what you want to do.

Today I suspected my microphone was bad as I measured 2 speakers, one used and one very unused, and both of them had lost sensitivity in the treble since the last time I measured them.

So I retrieved the old Ford V6 coilpack I had and triggered it with a 9V battery. I can get spark this way with gaps under 1mm. Across the gap I have a 560pF 1KV ceramic capacitor. It takes a number of tries to get a good spark as you tend to get trains of sparks instead.

With my EM172 microphone and 192KHz motherboard sound, the impulse response I get is remarkably similar to figure 2a in this reference:

http://www.sea-acustica.es/WEB_ICA_07/fchrs/papers/nla-02-004.pdf

However while that is presumably a direct pressure measurement, what I measured was a doublet which I had to convert to an impulse with a LP filter. I then fed that into REW.

So I don't know why my measurement looks like a doublet (N-wave) while theirs looks like the impulse form of that doublet. The blue line is the doublet, the black line is the impulse response (abusing REW by importing the doublet as audio data). Note that I have the same T=25uS as in the 1974 reference, which corresponds to figure 5.

I've tried this in various incarnations using tasers, grill sparkers, and now a Ford coilpack. These results are very indicative of what I've seen every time. The main thing to watch out for seems to be that your sparker is generating single sparks and not spark trains. You can hear the difference, the spark trains fizzle out in the room rather than a normal sounding spark reverberation.

I don't really know how useful this is and I don't have a reference mic. But I'm dumb enough to try spark calibration. The rising response under 2KHz is due to white noise. Here is the REW file:

Spark Calibration for DIYAudio - Google Drive
 

Attachments

  • Doublet-vs-impulse.png
    Doublet-vs-impulse.png
    27.5 KB · Views: 397
  • EM172-spark-response.png
    EM172-spark-response.png
    32.2 KB · Views: 155
  • EM172Z1 July 2015 (1).pdf
    102.4 KB · Views: 77
BTW, one advantage to the coilpack sparker is that the output is balanced. So the electrostatic wave cancels out directly in front where your microphone will be. Maybe not perfectly balanced depending on the internal construction but I would think it would still have to be an improvement.

This is important because you can move the microphone closer to the spark to get a better signal to noise ratio with a weaker spark, without needing to add some kind of shielding screen between them which may ruin the idea of calibration.
 
Member
Joined 2004
Paid Member
You can use a magnetic pickup to get an idea of the energy across the gap, both time and level. A few turns of wire oriented to pick up the current should be enough, although a current probe is better, and then you will have a better view of the spark duration. I expect the current profile will be a better analog of acoustic energy generated by the spark. The shunt cap may be too big. Also the components of the spark gap may be vibrating. Lots to explore.
 
The most troubling issue right now is the rising response below 2KHz. So far any time I've seen other kinds of response variations, it's been due to pulse trains more than anything else.

I've had this issue with the >2KHz response since the first experiments with different spark sources. I put a metal screen in front of the microphone and it's still there. It's possible to make it worse with different arrangements but I don't think I have gotten it below 2KHz with this microphone. I wonder if it could be an absorption type effect in the electret capsule or a settling time effect of the internal circuit.

The other possibility is that the rising response is actually part of the microphone response, but I don't think so.
 
Member
Joined 2004
Paid Member
A big limitation of impulse testing is the really low SNR for low frequencies. One test would be to generate a electronic pulse into the preamp with a similar level and time as the spark output from the microphone. You can then see if the self noise of the system is the LF limitation. Usually sparks are only used for high frequency testing. You are in that classic trap of looking for an independent reference to see if the HF is matched to the full response. The only way I know is an electrostatic actuator which is not applicable to ECM mikes.

I may be able to lend you a calibrated mike if that would help.