Microphone Noise Testing

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
First you need a dead silent room! hard to do that!

Well, I don't know any literature, I would first start replacing the capsule for an equivalent circuit, in condensers with a cap matching the capsule capacitance would be enough, for other mics figure out, RL circuits for dynamics may do the job, maybe a cap is needed but hard to know the value, doing an impedance measurement as done with speakers and then trying to match that may be the way to go. Wit this you will be able to determine the electronic noise generated by the circuit, but even then it will be really low and would be hard to amplify it to a measurable level without adding significant noise. For the capsule, I don't know, I don't know any way to get down to unmeasurable low acoustic noise environment, or at least low enough for some mics around, which only has a few dBA SPL. But I think the initial approach, for the electronic noise, would give a reasonable approach, since, in absence of acoustic perturbation, the capsules should behave as the electronic equivalent.

I'm curious, why are you trying to measure that?

JS
 
Joaquins,

I am testing ribbon microphones with active electronics.

First you need a dead silent room! hard to do that!
I have a room that is a fire vault that has LPs as well as tapes on three of the four walls. Suspended in the middle of the room is a pressure cooker with foam on the inside of that. So it is reasonably quite.
Well, I don't know any literature, I would first start replacing the capsule for an equivalent circuit, in condensers with a cap matching the capsule capacitance would be enough, for other mics figure out, RL circuits for dynamics may do the job, maybe a cap is needed but hard to know the value, doing an impedance measurement as done with speakers and then trying to match that may be the way to go. Wit this you will be able to determine the electronic noise generated by the circuit, but even then it will be really low and would be hard to amplify it to a measurable level without adding significant noise.

We have two test jig for measuring the head amplifier, for normal production testing we use a low .2ohm resistor that simulates the ribbon resistance. We also have a test jig that matches the impedance of the ribbon but this is not normally used. The test jig is then plugged into a grace M101 for amplification and P48 supply.  The output is measured on an Audio Precision. I have an excel spreadsheet that calculates the noise factor of the Grace by its self then the noise factor of the complete system, and by subtracting the two noise factors it give the noise for just the head amplifier. 

I did not setup this excel file the previous chief engineer did, and I am the new guy that has been tasked with figuring out a way to do production testing for the noise of the microphone. My boss wants: spectral, unweighted, CCIR quasi peak and an EIN dB A noise measurement on each microphone.


 
I don't know how the CCIR quasi peak is done exactly but you will  find it anywhere. For all the other you need the noise, with proper timing, maybe 2s RMS, on each band, to do spectral and unweighted that will do the job, for EIN dBA you need the value per band, apply the A curve and apply the sensitivity probably at 1Pa to the reference value, which is 2*10^-5 Pa. So, you have XdBu @ 1Pa,1kHz (94dB SPL=94dB A), Then, for your noise which will be YdBu you can know the pressure. Let's say you have -30dBu sensitivity (I don't know if I'm even close to your value) and -100dBu noise.
-30dBu----->94dBA
-100dBu---->EIN[dBA]=24dBA (my example is crap!)

For the unweighted you still need a filter, it doesn't make any sense to have a value affected for the 1MHz noise you may have in your mic, since it will die soon, at most I would take a 100KHz bandwidth, and maybe filtering some of the low side, at a couple of Hz maybe, both of which will be probably done by your preamp anyway.

I would stick with the replacement of  the capsule for production test, I don't know the final EIN of your mic, for acoustical measurements usually you want to keep your measurements over 10dB of the noise floor, an "enough" silent room to measure 7dBA EIN is -3dBA (NUTS!) which is typical figure for low noise condenser microphones, for ribbon probably you will have more, some Royer are around 18dBA EIN I guess, not well specified.

JS
 
negative dB values are very real, no worries, specially when looking at third octave bands... a trap for young players when adding (log sum!) bands is to 'forget' that you can have negative dB values.
At work I use a excel sheet for STC measurements. in very low (background) noise environments and when doing only one or two background measurements the averaging in excel would add 0dB on all 1/3 oct. bands with empty cells (excel assumes 'zero' for an empty cell, the sheet was prepared to average four measurements). this would upset the average log sum... summing 1/3 oct. bands with 0dB each will sum up to 14.7dB.....

see here: http://www.sengpielaudio.com/calculator-spl30.htm

- Michael

 
a soBer Newt said:
I am testing ribbon microphones with active electronics.

Calibrate the gain for reference output and then disconnect transformer Primary. With a bridge or high quality multimeter with Kelvin probs (we use HP 3457A) accurately measure ribbon DCR (it should be done after the clamps at the point of transformer leads connection), and then instead of ribbon connect to the Pri resistor of found DCR value.

Best, M
 
audiomixer said:
negative dB values are very real, no worries, specially when looking at third octave bands... a trap for young players when adding (log sum!) bands is to 'forget' that you can have negative dB values.
At work I use a excel sheet for STC measurements. in very low (background) noise environments and when doing only one or two background measurements the averaging in excel would add 0dB on all 1/3 oct. bands with empty cells (excel assumes 'zero' for an empty cell, the sheet was prepared to average four measurements). this would upset the average log sum... summing 1/3 oct. bands with 0dB each will sum up to 14.7dB.....

see here: http://www.sengpielaudio.com/calculator-spl30.htm

- Michael

I know they are real, of course, (once in a hearing test I marked -5dB at one tone, I think it was 500Hz, I was surprised, but it was what the professional said, of course not so good at 4kHz...) In any case -3dBA SPL is really low noise as target, it isn't a single 1/3rd oct band, is the whole thing, A weighted but the whole thing, as I said, maybe with 7 or 8 dBA SPL is enough in  this case

JS
 
Nathan,

For testing total noise microphones are generally put in a  small heavy steel sealed enclosure that's vibration
isolated.

For a ribbon microphone noise sources are:

1) acoustic radiation resistance thermal noise
2) pressure noise of damping screens (and any other acoustic/mechanical damping)
3) ribbon resistance noise
4)transformer  winding resistance noise
5) EIN of head amplifier

The radiation resistance of a typical ribbon is vanishingly small.

Pressure noise of damping is typically very small as well, unless the mic is resistance controlled...like an omni pressure ribbon.
(we make those)

Most will come from the transformed ribbon resistance and amplifier.
With typical 300 ohm output impedance of a passive mic the noise density is
2.2nV/sqrt Hz giving .311uV in 20KHz bandwidth.
That gives 17.8 dB SPL equivalent noise  unweighted with a typical 2mV/Pascal sensitivity.

This is in the ball park of the EIN of a good head amp, so total noise will be a couple or so dB worse at best.
The noise adds in quadrature since the two noise sources are uncorrelated.

Often higher ratio transformers are used in active ribbons to relieve the noise requirements of the head amp.
100:1 is often cited, giving output impedances in the 1000 ohm range.

Les Watts
L M Watts Technology
 
leswatts said:
The radiation resistance of a typical ribbon is vanishingly small.

Pressure noise of damping is typically very small as well, unless the mic is resistance controlled...like an omni pressure ribbon.
(we make those)

Most will come from the transformed ribbon resistance and amplifier.

Hi Les,

Yes, that's the reason I proposed to substitute the ribbon with dummy resistance of the same value to measure the noise. For all practical purposes it should give a very good idea without complications of finding or making a suitable measuring chamber.

Best, M

 
Hello Les Watts,
Thank you for the information.

leswatts said:
Most will come from the transformed ribbon resistance and amplifier.
With typical 300 ohm output impedance of a passive mic the noise density is
2.2nV/sqrt Hz giving .311uV in 20KHz bandwidth.
That gives 17.8 dB SPL equivalent noise  unweighted with a typical 2mV/Pascal sensitivity.

Could you please elaborate how you get 17.8dB SPL from 2.2nV/sqrt Hz.

 
a soBer Newt said:
Hello Les Watts,
Thank you for the information.

leswatts said:
Most will come from the transformed ribbon resistance and amplifier.
With typical 300 ohm output impedance of a passive mic the noise density is
2.2nV/sqrt Hz giving .311uV in 20KHz bandwidth.
That gives 17.8 dB SPL equivalent noise  unweighted with a typical 2mV/Pascal sensitivity.

Could you please elaborate how you get 17.8dB SPL from 2.2nV/sqrt Hz.

It means the output of the mic would be 2.2nV/sqrt Hz when exposed at 17.8dB SPL. To get that you look at the sensitivity, in this case 2mV/Pascal.

So, for 2.2nV/sqrt Hz *sqrt(20000Hz)=311µV
311µV/2mV/Pa=155µPa
20*Log(155µPa/20µPa)=17.8dB SPL

Hope that helps.

JS
 
Have a look at the ISO standard 60268-4 Microphones.  Appendix B describes a sound isolation device. It is essentially a large metal can with very thick walls, and supported by springs to reduce coupling of structure-born vibration into the test chamber. Of course, you need not necessarily follow the instructions precisely.

An important question is to decide how low you need to measure and to also know  the amount of ambient noise in your lab. From those two figures you can calculate the amount of attenuation that the device needs to provide.

From much personal experience I know that by far the greatest difficulty is with measuring noise below 200 Hz.
 
a soBer Newt said:
I am testing ribbon microphones with active electronics.


We have two test jig for measuring the head amplifier, for normal production testing we use a low .2ohm resistor that simulates the ribbon resistance. We also have a test jig that matches the impedance of the ribbon but this is not normally used. The test jig is then plugged into a grace M101 for amplification and P48 supply.  The output is measured on an Audio Precision. I have an excel spreadsheet that calculates the noise factor of the Grace by its self then the noise factor of the complete system, and by subtracting the two noise factors it give the noise for just the head amplifier. 

I did not setup this excel file the previous chief engineer did, and I am the new guy that has been tasked with figuring out a way to do production testing for the noise of the microphone. My boss wants: spectral, unweighted, CCIR quasi peak and an EIN dB A noise measurement on each microphone.
Then what is exactly your problem? The AP is capable of providing all the desired measurements, and the pressure-cooker is probably good enough for isolation. You just need to set-up the AP to restrict the frequency response to 22-22kHz.
Apparently, your boss does not want to extract the noise factor of the electronics alone, just the standard measurements relevant to a thorough QC and a customer may want to know.
 
Speaking about pressure-cooker, wouldn't a vacuum chamber make perfect isolation? or having air or not around will affect the self noise of the mic even if there's no sound pressure over that diaphragm? You need to be sure the mic body or nothing inside is completely sealed and with gases inside (maybe electrolytic caps may have a problem) but if there's vacuum inside the chamber, it's noise, and it's not so hard to make that in a lab environment. The other thing would be the temperature of the components, and without air any component dissipating some power will be warmer than in air, but that's just about making a fast test, if you want you could let the mic temperature stabilize, then apply the vacuum as fast as possible and then make the measurement. Of course this makes no sense in a passive mic where the mic won't dissipate any noticeable power in normal use.

JS
 
joaquins said:
Speaking about pressure-cooker, wouldn't a vacuum chamber make perfect isolation? or having air or not around will affect the self noise of the mic even if there's no sound pressure over that diaphragm?
There would be a significant change because of the absence of Brownian noise due to to the molecular agitation of the air particles. that's the reason why there is noise even when the mechanical/acoustical is supposed perfect. Brownian noise can be nulled by operating at 0°K or 0 Pa, both of which are relatively impractical  ;)
 
abbey road d enfer said:
joaquins said:
Speaking about pressure-cooker, wouldn't a vacuum chamber make perfect isolation? or having air or not around will affect the self noise of the mic even if there's no sound pressure over that diaphragm?
There would be a significant change because of the absence of Brownian noise due to to the molecular agitation of the air particles. that's the reason why there is noise even when the mechanical/acoustical is supposed perfect. Brownian noise can be nulled by operating at 0°K or 0 Pa, both of which are relatively impractical  ;)

How much the noise figure is affected by the Brownian noise. I don't know what pressure level would that be equal too, but I think pretty low, in the order of -20dB SPL on dubious sources online, compared to over 5dB SPL equivalent self noise of microphones, I don't think that would be a problem...

0Pa is not theoretically impossible, in some cases negative pressure is possible, not in gases though. Still, getting pretty close to 0Pa is much more practical than going any close to 0ºK

JS
 
joaquins said:
abbey road d enfer said:
There would be a significant change because of the absence of Brownian noise due to to the molecular agitation of the air particles. that's the reason why there is noise even when the mechanical/acoustical is supposed perfect.
How much the noise figure is affected by the Brownian noise. I don't know what pressure level would that be equal too, but I think pretty low, in the order of -20dB SPL on dubious sources online, compared to over 5dB SPL equivalent self noise of microphones, I don't think that would be a problem...
In any microphone with self noise below 20 dBA spl, acoustical Brownian noise will be a significant part.

A practical microphone has to have 'resistive' damping for flat response.  This is achieved with acoustic & mechanical resistance which is noisy.

The Neumann pdf book by Peus on their website gives details for where you want the acoustical resistance.

You can design a mike with little acoustic resistance damping so it has a huge peak.  If you now electronically EQ the peak for flat response, you have a VERY quiet mike.  That's how the Sennheisers get their LN performance.

I've done an experimental mike like this for Calrec and it was very quiet.  You don't need their RF circuit.
 
Back
Top