Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Timtape]

Another measurement error is the sine wave generator. That requires a form of speaker. Speaker distortion is a magnitude of error beyond the mic capsule's THD. We end up measuring both with each contributing to the errors.

Thanks Jim for bringing us back to this fundamental point, mentioned more than once earlier in the thread.

If speaker inaccuracies are a problem in calibrating mics they're even more a problem for listening tests. For we must listen via speakers or headphones.

With mic calibration work a special speaker called an "electrostatic actuator" is often used. It's really a high quality condenser microphone  used as a tiny speaker in a special closed coupler arrangement. Apparently it has the least distortion possible in a speaker or driver.

Even with a speaker as a source, with many distortions, two microphones can be used at once in the same coupler. If those speaker distortions are identical in both microphone outputs, and using another speaker, a different set of distortions are identically produced in the two mics, the distortions are quite validly assumed to originate from the speakers, not the mics, and since the distortions are now known, they can be subtracted from the result, revealing just the distortions of the mics. At least that's my understanding of it.

 But we can only "listen" to a mic via a speaker or headphone, with all its distortions which as you say are likely much worse than of the microphone. In a calibration test procedure we can validly cancel out the speaker distortions. Listening, we can't.
 
Once we put all our confidence in listening tests alone, ridiculing formal test procedures, we are tempted to pretend the speakers or headphones for listening aren't there. They seem a bit like the elephant in the room...

[klaus]

Once we put all our confidence in listening tests alone, ridiculing formal test procedures, we are tempted to pretend the speakers or headphones for listening aren't there. They seem a bit like the elephant in the room...

You can land arguments better by dropping your snarky tone towards those who disagree with you.

Speaking of disagreeing, I see two fallacies in the two main arguments you bring forth in your post.

1. You write:

Even with a speaker as a source, with many distortions, two microphones can be used at once in the same coupler. If those speaker distortions are identical in both microphone outputs, and using another speaker, a different set of distortions are identically produced in the two mics, the distortions are quite validly assumed to originate from the speakers, not the mics, and since the distortions are now known, they can be subtracted from the result, revealing just the distortions of the mics.

Whereby you assume that "distortion" is some kind of single-factor event that can be mathematically deduced and removed from the test arrangement as per your recipe.

But I don't regard "distortion" as a linear or single-factor event, instead, it's the sum of all kinds of interrelated, interactive events: phase issues, frequency issues, delay issues, timbre issues, resonances, on and on. The model you propose, to cleanly "swap" out these anomalies and thereby deduce that they must have originated from the speaker is, I believe, incorrect.

But that is just a side issue, in my estimation.

2. The true "Elephant in the Room" as you call it, is your assumption that it makes sense to completely disregard the listener's (subjective, unreliable) judgement, when, after all, the listener is the ultimate target of any audio product. No testing chamber, no mathematical formula, no theoretically idealized approach, nothing ultimately matters but whether the listener likes what s/he hears or not. And that approach- manufacturers making splendid mics by listening, and listening to the listeners- has worked for many decades, giving us iconic mics that have spawned countless iconic recordings.

I find the engineering of audio products that exclude sensual feedback dystopian and soulless. As we know from those who go that route to achieve "accuracy" in their products, they have not found the degree of acceptance in the marketplace one would assume. 

[Timtape]

"...the test arrangement as per your recipe."

"The model you propose..."

It's neither my recipe or my model. It's what I take to be normal procedures in testing laboratories all over the world. If I've read it wrong (it's not my speciality) I'm more than happy to be corrected.

The true "Elephant in the Room" as you call it, is your assumption that it makes sense to completely disregard the listener's (subjective, unreliable) judgement...

Actually I dont.

But to be specific, if a potential customer comes to you, a microphone expert, and asks you to test  his U87 mic whether it is still operating within factory tolerances, what do you do? How do you test it? If you have not the testing facilities do you send it off to a testing facility? What then is your attitude to formal testing of a microphone and why?

[klaus]

(...)if a potential customer comes to you, a microphone expert, and asks you to test his U87 mic whether it is still operating within factory tolerances, what do you do?

We don't need to reach for a hypothetical "potential". For thirty years of "analyzing" someone's supposedly malfunctioning mic, I plug in and listen. I have never come across a case where that type of analysis did not yield a 100% positive result: a mic's defect either shows up by listening, or it does not exist.

How do you test it? If you have not the testing facilities do you send it off to a testing facility?

What "testing facility" did you have in mind? Did you think that Sennheiser, Telefunken USA, AKG, etc. use rocket science when they test a supposedly defective mic that comes in for repair?

I hope this isn't news to you, but, if the mic distorts, you'll hear it, if its noise floor is too high, you'll hear it. If its frequency response is odd, you'll hear it. As I always say: if the mic sounds right, it is right and healthy, if it sounds wrong, then it needs repair. There are no strata of cancerous developments hidden from plain listening, but that certainty of diagnosis sure does not come on day one of dealing in the matter. Neither does brain surgery.

Your belief in scientific diagnostics* of defective microphones may stem from your personal unfamiliarity and inexperience with microphones, and therefore may be more informed by images of lab coats, cleanrooms, oscillographs and HP Scientific Calculators. But, just between me and you: most of my colleagues at (companies shall remain unmentioned) do the exact same thing as I do, but they may be a bit more diplomatic about how they sell it, to keep the customer's belief in the superiority of objective science intact.

* OK, I lied. Some customers are so intent in having the diagnosis delivered in a scientific way, that I will actually sweep the mic (shitty methodology, for sure:) using uncalibrated speaker, sine wave, white noise generator etc. but at least it will give the owner a relative ± number of the deviation from standard, which make him happy.  (Not that this was ever necessary for me to diagnose and remedy the underlying issue, which is usually a contaminated capsule or defective capacitor.)

[Timtape]

My understanding is that at least with measurement microphones used for say industrial, medical, legal purposes it is normal, perhaps decreed  by law, to have the mics' calibrations formally tested on a regular basis, and to a  high level of accuracy.

Your belief in scientific diagnostics* of defective microphones may stem from your personal unfamiliarity and inexperience with microphones...

 My own experience of calibration was in the late 70's and early 80's when I worked as a technician in repairing and calibrating  hearing aids and annually repairing and calibrating portable audiometers (hearing testing units) to National Standards. The audiometers were used by school nurses to screen children for possible hearing problems. I took my calibration work seriously as I was trained to, and  adjusted each spot frequency (from 125Hz to usually 6000Hz) as closely as possible to the 0 (zero) reference on the B & K analyzer which was fed by the B & K Artificial Ear condenser measurement microphone.

I know of at least one recording engineer  specialising in acoustic classical recordings, who has his high quality recording  condenser microphones regularly and formally tested to confirm they are still performing within manufacturers' specifications.

 

[Kai]

With mic calibration work a special speaker called an "electrostatic actuator" is often used. It's really a high quality condenser microphone  used as a tiny speaker in a special closed coupler arrangement. Apparently it has the least distortion possible in a speaker or driver

The "electrostatic actuator" isn't an acoustic sound source of any type, but a grid that can be mounted close to the diaphragm of a measurement microphone. It excites the mic by electrostatic force, then a (theoretical) correction curve for the sound pressure (treble) boost caused by the physical dimension of the mic is applied. The result is the the calibrated frequency response published by the manufacturer.

I don't think that harmonic distortion measurement makes sense this way (I might be wrong).

For distortion measurement on mic's the differential method is used.
2 different frequency sinewaves from 2 different soundsources are used at the same time and "mixed" acoustically.
The microphone does produce differential (not harmonic) distortions that are not present in the individual sound sources, which produce harmonic distortions only.
With exciting frequencies of, e.g. 10kHz from speaker one and 11kHz from speaker two, the mic's nonlinearity produces e.g. 11kHz-10kHz=1kHz.
Harmonic distortion of the sound sources don't matter this way.

This method can be used to measure extremly low electrical distortions too, as the (then two) generators don't need to be very low distortion types.

[Timtape]

The "electrostatic actuator" isn't an acoustic sound source of any type, but a grid that can be mounted close to the diaphragm of a measurement microphone. It excites the mic by electrostatic force, then a (theoretical) correction curve for the sound pressure (treble) boost caused by the physical dimension of the mic is applied. The result is the the calibrated frequency response published by the manufacturer.

Thanks for that correction Kai. I admit I didnt quite understand why in photos of the Electrostatic Actuator the "diaphragm" looked more like the mere backing plate of a condenser microphone!

On rereading 6.6  I now see Nedzelnitzky makes it plain. The relevent section is 6.2 where Nedzelnitzky does  mention using a reference microphone as a sound source.

I don't think that harmonic distortion measurement makes sense this way (I might be wrong).

For distortion measurement on mic's the differential method is used.
2 different frequency sinewaves from 2 different soundsources are used at the same time and "mixed" acoustically.
The microphone does produce differential (not harmonic) distortions that are not present in the individual sound sources, which produce harmonic distortions only.
With exciting frequencies of, e.g. 10kHz from speaker one and 11kHz from speaker two, the mic's nonlinearity produces e.g. 11kHz-10kHz=1kHz.
Harmonic distortion of the sound sources don't matter this way.

This method can be used to measure extremly low electrical distortions too, as the (then two) generators don't need to be very low distortion types.

Thanks for that. The heterodyne principle isnt it?

As I said this field is not my speciality so great to have your contribution.

[Kai]

On rereading 6.6  I now see Nedzelnitzky makes it plain. The relevent section is 6.2 where Nedzelnitzky does  mention using a reference microphone as a sound source.

This is another method, as far as I remember you compare 3 microphones this way to determin their sensitivity at a certain frequency, but I never dug deeper into that.

The heterodyne principle isnt it?

It's based on that underlying principle, widely used for FM radio receivers and high quality tone generators.
It's even called difference tone- or intermodulation- distortion measurement, IMD.
For IMD measurement often a combination of a low and high frequency like 60Hz and 7kHz with a level ratio of 4:1 (SMPTE IMD measurement) is choosen, the low frequency modulates the level of the higher one.
More details here:
https://www.ap.com/technical-library/more-about-imd/
In microphones most of us know this effect - during a strong pop noise (from a singer) the higher frequencies go down in level. If the low frequency is removed by a high pass filter the level modulation still remains.
These kind of distortions make the sound rough, they are present in every system with amplitude nonlinearities. The 2-tone method is just another way to measure the same thing.

For us it's important to kow that these types of distortions exist, for mixed signals like music a distorting microphone or amp doesn't only produce "harmonically rich" 2nd and 3rd order distortions, but all kind of "dirt" in form of sum and difference frequencies of ALL frequencies input.
Guitarist use that in their amps, and most of us will know how a 6-note guitar chord sounds through a heavy metal setting. That's why power rock chords (mostly 2-tone chords based on the 5th) are used, they don't produce dissonant tones when distorted.

[Timtape]

Thanks Kai.

Re a  mic such as a quality 1/2" omni condenser from Neumann or DPA, can you say what are typical distortion figures, such as harmonic, IMD etc? What do you understand as the significant distortions (and their typical values) in such a mic generally, using the word "distortions" to include all deviations from strict linearity?

Regards
Tim

[Jim Williams]

There are papers from the AES Journal on this subject of THD of condenser capsules. I rcall one from 20 odd years ago exploring the effects of loading on the THD results.

Lower value loading resistors like the 200 meg ohms used in older tube mics will generate capsule THD in the lower frequencies. The article determined that a load of 10 gig ohms avoided the low frequency distortion many find subjectively pleasing on vocals.

The 10 gig values lowered the capsule THD to .001% at 50 hz.

[klaus]

That's just THD. I regard any deviation from the original as distortion - distortion of the reality I perceive with my ears.   

Except there's no measuring strick for that, yet. And, as this discussion shows, it's not always easy to be heard on that level, as long as there is the belief that objective science has moved the needle towards better mics in the last five decades.

I am not anti-science. But I am anti cherry-picked science. Observation without limitation to methodology needs to be a central part of scientific investigation, especially when an observed phenomenon cannot yet be explained or quantified at the time of observation.

A good example from the audio field: interconnects all sound different, yet that difference cannot be fully explained in terms of capacitance, reactance, resistance, frequency response, etc.

And neither can inadequate copies of specific components, like M7 capsules, BV8 transformers, VF14 tubes, and so on, be sufficiently explained in scientific terms, to the point that their shortcomings could be overcome.

Here, again, and until further notice, ears are trump.

[Timtape]

There are papers from the AES Journal on this subject of THD of condenser capsules. I rcall one from 20 odd years ago exploring the effects of loading on the THD results.

Lower value loading resistors like the 200 meg ohms used in older tube mics will generate capsule THD in the lower frequencies. The article determined that a load of 10 gig ohms avoided the low frequency distortion many find subjectively pleasing on vocals.

The 10 gig values lowered the capsule THD to .001% at 50 hz.

Where very good low frequency performance was needed I understood that the Sennheiser Radio Frequency type microphone (MKH) was an important development.

But apparently conventional microphones including B & K measurement types appear to have accurate response (including low THD) to well below 20 Hz  and yet are not RF types. Can you shed any light on this? Did the input impedance of FETs improve for example?

Regards
Tim

[Kai]

... interconnects all sound different, yet that difference cannot be fully explained ...
... Here, again, and until further notice, ears are trump.

If I can hear it but cannot measure it I do measure the wrong thing - not from me, but I consider this to exactly describe the problem.
Physics just has entered the area of quantum effects, and mic signals are so low that they may well be prone to those - just an idea and I have no evidence for this hypothesis.

I'm in the process of building a setup to measure extremly low distortions in electronic parts.
Currently I can measure down to -155dB and still trying to come down further.
My findings are already very astonishing: there are sources of distortion where you would not expect them and there are things that I always thougt could be a problem that aren't.
Specially connectors so far did not show up anything suspicious.

[klaus]

With "interconnects" I did not mean connectors but the cables that connect audio components like mics, to mic preamps, and so on. (Maybe it's a term not familiar to the audio community outside the U.S?)

If I can hear it but cannot measure it I do measure the wrong thing

Agreed. And, in a truly scientific process, that point would be the start of further exploration, rather than premature speculation, that, what I cannot measure does not exist (how often have I read this on audio forums!)

[Jim Williams]

The late Bob Pease of National Semiconductor did low level distortion tests on then new LME series of opamps because the traditional techniques of adding noise gain and extrapolating the results through an audio analyzer like Audio Precision didn't go down far enough to reveal the baseline distortion.

Bob ended up adding an Agilent network analyzer to the process to determine that opamp's true THD at -155 dbu. The LME49990 (now discontinued for no good reason) did even lower, almost -160 dbu.

This type of test does apply to a capacitor mic's active impedance converter/output drive but does not apply to a physical capacitor capsule as that requires different stymuli.