How do we know that "flat" is flat?

[mmdiss]

Hello everybody, please be prepared for the dumbest question you might have read recently (sorry).

I've searched this topic on google without success (probably I've searched wrong, please point me in the right direction!).

What is a flat loudspeaker? Well, you say, it's a loudspeaker that sounds flat to a calibrated (flat) microphone. What about a flat microphone, how do you know it's flat? Of course, it's flat because it's recording flat when exposed to a flat loudspeaker... and back again...

So, how do we know what is flat?? What is the reference for flatness?

I mean, this should be a basic question. I can imagine a few possibilities:

1. we have a way to produce sine waves at different frequencies with the very same amplitude
2. we have a perfect source of pure white(pink?) noise
3. we don't have any of the above, but we can calculate the deviation between the actual sound and the theoretically perfect signal of points 1. and 2.
4. we just know what is flat, and the flat reference is a sort of average flatness for different listeners

Points 1. and 2. would be very desirable, and I would bet on these, however, I couldn't find anything like that (again, I might have searched wrong). I would settle also for 3. but, again, I couldn't find any evidence of that.

Point 4. is plausible but not very desirable, since everybody might have a slight ear difference and this would limit the ability to calibrate by ear (what if my ears don't respond exactly to equal loudness? I might be targeting the wrong curve).

Please help me in understanding this! It was fantastic when I discovered that EQ could be used to fix the audio (It was really a surprise after years of "knowing for sure" that EQ was just additional "distortion"). And this was possible just by targeting the equal-loudness contour (eventually adjusted by the Harman target). Later on, I started questioning myself whether I was just calibrating my audio devices to a consistent (but possibly wrong) curve, so I need answers

 

[sergeauckland]

Firstly, flat is a relative term. Flat used to mean +-3dB between certain limits, typically 20Hz-20kHz. More recently, flat would be +-1dB 20Hz-20kHz, and for measurement purposes, one would really want flat to be +-0.1dB between 20Hz - 20kHz, maybe even 1Hz -100kHz depending on the application. I'm happy if my electronics measures +-1dB 20Hz-20kHz, which is inaudible on music and speech programme compared with +-0.1dB or less. Even the old standard of +-3dB is hard to hear on speech or music although it depends on the speech or music and how the +-3dB goes. Gentle variations are more audible on speech and music than a sharp peak or trough.

What all this means is that the term 'flat' can mean whatever you want it to mean, dependingon how its measured and the application.

As to microphones, they are calibrated with a special device that goes back to physical units like air pressure. These devices in turn are calibrated against tighter tolerances and physical units, until everything goes back to a National Standards Laboratory and definitions of the metre, kilogram, second and coulomb. Up until recently, the metre was arbitrarily defined as the length of a metal rod held in Paris, the kilogram to a lump of metal also, I think, in Paris. More recently they have been defined in atomic terms, as the primary standard metre or kilogram would lose a few atoms every time it was handled, so not a fixed standard, and the second and the coulomb have for a long time been defined against atomic standards, not affected by the passage of time or handling.

Getting back to loudspeakers, a loudspeaker can be measured using a calibrated measurement microphone either in an anechoic chamber, (very large, very expensive and still not as accurate as modern methods) outdoors, (weather dependent mostly) or more recently using computer simulations of anechoic conditions using clever mathematics that can remove ambient effects.

S.

 

[NTK]

Your question is not dumb. This kind of questions is fundamental to the field of metrology.

Here is an article from NIST on their new laser based microphone calibration and is described in this paper. Their technique covers the range of 250-1000 Hz in their study. Other frequencies are covered using other methods.

Testing 1-2: New Laser-Based Microphone Calibration Measures Up

Researchers at the National Institute of Standards and Technology (NIST) have conducted the first demonstration of a faster and more accurate way to calibrat

www.nist.gov

 

[abdo123]

Your question is not dumb. This kind of questions is fundamental to the field of metrology.

Here is an article from NIST on their new laser based microphone calibration and is described in this paper. Their technique covers the range of 250-1000 Hz in their study. Other frequencies are covered using other methods.

Testing 1-2: New Laser-Based Microphone Calibration Measures Up

Researchers at the National Institute of Standards and Technology (NIST) have conducted the first demonstration of a faster and more accurate way to calibrat

www.nist.gov

Are cheap Class 2 microphones we buy for our room correction fall under this? Seems too expensive to do it individually to each one.

 

[NTK]

Are cheap Class 2 microphones we buy for our room correction fall under this? Seems too expensive to do it individually to each one.

I don't think so. NIST (USA) and its counterparts in other countries typical serve to calibrate the test instruments of commercial testing (or calibration) laboratories. These instruments then serve as "secondary standards" and are used to calibrate individual instruments. This provides the "traceability" of the calibration back to the national authority (e.g. NIST traceable calibration).

Metrological Traceability: Frequently Asked Questions and NIST Policy

Antonio Possolo, Sally S. Bruce, Robert L. Watters, Jr.

www.nist.gov

 

[sergeauckland]

Are cheap Class 2 microphones we buy for our room correction fall under this? Seems too expensive to do it individually to each one.

With a known calibrated measuring microphone to calibrate the source, it should then be easy to derive a calibration file for any microphone just by running a sweep or impulse and letting the software (something like REW, ARTA or whatever) effectively do the sums. I did this manually for my measuring mic, comparing it to a calibrated Earthworks mic and putting the values into a spreadsheet which calculated the calibration file. I just did it for the 31 third-octave frequencies as microphones tend not to have rapid variations, but REW etc will do the same at hundreds of points within a few seconds. It will take longer to unpack the microphone, mount it, pack it up again and print off the calibration file / store the electronic version, than it takes to actually run the sweep.

S.

 

[Doodski]

Hello everybody, please be prepared for the dumbest question you might have read recently (sorry).

That is a very very good question and the answer affects every part of your life from buying weighed bread to of course mic'ing your speakers.

 

[abdo123]

With a known calibrated measuring microphone to calibrate the source, it should then be easy to derive a calibration file for any microphone just by running a sweep or impulse and letting the software (something like REW, ARTA or whatever) effectively do the sums. I did this manually for my measuring mic, comparing it to a calibrated Earthworks mic and putting the values into a spreadsheet which calculated the calibration file. I just did it for the 31 third-octave frequencies as microphones tend not to have rapid variations, but REW etc will do the same at hundreds of points within a few seconds. It will take longer to unpack the microphone, mount it, pack it up again and print off the calibration file / store the electronic version, than it takes to actually run the sweep.

S.

That's hardly feasible when each production cycle has 1 million microphones to calibrate. but @NTK explained that these laser techniques calibrate the calibrators and the calibration trickles down to class 2 microphones.

 

[abdo123]

buying weighed bread

That's no longer a concern as we unified Plank's constant a few years back

 

[sergeauckland]

That's hardly feasible when each production cycle has 1 million microphones to calibrate. but @NTK explained that these laser techniques calibrate the calibrators and the calibration trickles down to class 2 microphones.

Really? I would be surprised if there were 1million people in the world that even knew what a calibrated microphone was, let alone be buying one.
In any event, whatever calibrator is used, it's still a largely manual process to fit the microphone into the calibrator and take it out again.

S.

 

[abdo123]

@NTK I have a question, is the 250Hz to 1000Hz limit there because most microphone become directional at this point? and the laser inherently cannot calibrate the off-axis response because it won't have access to the moving diaphragm?

For a second i thought this could easily replace anechoic rooms for the purpose of speaker calibrations, but then i realized the laser must always be 'in sight' of the diaphragm.

 

[abdo123]

Really? I would be surprised if there were 1million people in the world that even knew what a calibrated microphone was, let alone be buying one.

when you consider that each AVR now comes with a calibrated microphone i wouldn't be surprised if they're produced in bulk by one company or few companies in the world. If we include stuff like music and movie production the number rises even more substantially. The microphones used to record vocals and instruments and stuff may not need as strict requirements as something that is used to measure speakers but they do need to be within a reasonable range 20Hz to 20KHz.

 

[mmdiss]

Here is an article from NIST on their new laser based microphone calibration and is described in this paper. Their technique covers the range of 250-1000 Hz in their study. Other frequencies are covered using other methods.

Testing 1-2: New Laser-Based Microphone Calibration Measures Up

Researchers at the National Institute of Standards and Technology (NIST) have conducted the first demonstration of a faster and more accurate way to calibrat

www.nist.gov

Thank you! This was something I was looking for. However, I'm just wondering what to think when the article says:

• By repeating this laser test with a "reference" microphone -- one that has been previously calibrated -- researchers can more accurately and easily calibrate microphones than they could with more traditional methods.

Do I understand right that the laser is just a faster way to perform a calibration AFTER you already have a calibrated reference microphone?

Also, a question about the "reciprocity calibration". I understand that, with a microphone and a loudspeaker, I can "easily" let the speaker play some audio and correct what the microphone is picking up to match it with the source signal. The problem is, I can't know what part of the correction is attributable to the microphone and what part is due to the speaker. Am I understanding correctly if I say that, the reciprocity calibration is based on the same principle PLUS the consistency/similarity of the microphones/loudspeakers (ideally, one should have three identical devices) PLUS some kind of symmetry between a device working as a microphone compared to the same device working as a loudspeaker? It seems to me that without all these conditions we would be in the same situation I described with a "general" loudspeaker and a "general" microphone. Is a microphone "symmetrical" when operated as a loudspeaker? (meaning voltage in equals to voltage out)

EDIT: I did not realize that NTK posted two links, the paper clarifies my point:

• This approach utilizes the fact that the magnitude of the pressure sensitivity of a reciprocal transducer is the same regardless of whether it is used as a receiver of sound or a source of sound.

 

[kongwee]

I just manually EQ since I doing DAW all the time. I tend to boost bass more for listening. Off EQ if doing DAW.

 

[Rock Rabbit]

In short: a microphone is an "extended" range loudspeaker (membrane resonances over 10 or 20 kHz), the small membrane guarantee nearly flat response in the passband. For calibration there's only necessary to know frequency and displacement of a transducer to deduce the exact SPL value (+diameter & air density), calibration microphones comes with a pistonphone (mechanical fixed amplitude SPL reference typical 124 dB @ 250 Hz), lab devices even include a barometer to compensate for altitude. In the same way you can make a reference for other frequencies (near field reference).
An automatic procedure to generate a calibration sheet only takes seconds!. So cheap "calibration" microphones are impressively flat up to some 10 kHz but notably noisy

 

[FrantzM]

In short: a microphone is an "extended" range loudspeaker (membrane resonances over 10 or 20 kHz), the small membrane guarantee nearly flat response in the passband. For calibration there's only necessary to know frequency and displacement of a transducer to deduce the exact SPL value (+diameter & air density), calibration microphones comes with a pistonphone (mechanical fixed amplitude SPL reference typical 124 dB @ 250 Hz), lab devices even include a barometer to compensate for altitude. In the same way you can make a reference for other frequencies (near field reference).
An automatic procedure to generate a calibration sheet only takes seconds!. So cheap "calibration" microphones are impressively flat up to some 10 kHz but notably noisy

Agreed...
l would add that progress in computing power and cybernetics have made possible commodity-priced precision instruments.
Seriously, the microphone in an iPhone (In many Smartphones as it turned out) seems very accurate for the purpose of measuring SPL or even FR for audio. Here is an article about it, a Google search will yield many more.

https://acousticstoday.org/wp-content/uploads/2017/06/2-faber.pdf

That is something that has dawned on me: Hardware and Software that only well funded labs or individuals could afford are today, routinely available in consumer products with nary an inkling from even sophisticated and knowledgeable users. We carry without realizing it a device more powerful than the Cray I of yore (1970's)... Our smartphones are extraordinary images processing unit.. witness the incredible photos some (most?) smartphones can take or measuring capabilities... or to software that are available for free.. REW, VituixCAD are example among so many others, too numerous to list.. Lidar components that in 2017 (5 years ago for arithmetic challenged people ) cost around 17,000.oo USD ( that is Seventeen thousands...) have found its way into a circa $1000 smartphone (already performing a lot of other things and is more powerful than a 1970's Supercomputer) ... Coupled with a free app you can take accurate 3-D measurements of space and translate these into engineering levels drawings. I use it in my line of work. I used to wait for architectural drawings for IT installations... All I have to do now is bring my iPhone Pro and ... 10 minutes later .. Accurate drawings and 3-D renditions of the space. 3-D scanners used to cost tens of thousands of dollars.. an iPhone Pro (from the 12 and On) and a (free ) App ... Or an iPad Pro... Same results. And you can use these for other purposes among these take incredible photos or calibrate your Audio system or watch movies in HD and listen to music in pristine, accurate sound or ... surf the web while listening to music or ...

I am OT..I'll stop

Peace

 

[Killingbeans]

Sorry for off-topic, but the headline of this thread made me think of this clip from Rick & Morty:

And yes, a measuremet is only a good as the "gauge block" used to calibrate your instrument.

 

[DanielT]

Maybe these experts have a subdivision that handles speakers?

Modern flat Earth beliefs - Wikipedia

en.wikipedia.org

Sorry, I could not help it.

 

[Ozymand1as]

This is why, as previously stated, we now base systems of measure on invariant observations of the universe. Or, invariant in the observable universe at least.

We can measure the Planck constant, the speed of light, atomic oscillations, etc and use these fundamental properties to define kilograms, meters, and time. With these three units defined, you can define pressure. Now that we've defined pressure, we can calibrate instruments that accurately read pressure and so on. Units are weird.

 

[solderdude]

1. we have a way to produce sine waves at different frequencies with the very same amplitude
2. we have a perfect source of pure white(pink?) noise
3. we don't have any of the above, but we can calculate the deviation between the actual sound and the theoretically perfect signal of points 1. and 2.
4. we just know what is flat, and the flat reference is a sort of average flatness for different listeners

There is another issue at play which has not been included. Even when the FR is kind of flat we also have to consider impulse response as we are dealing with music which does not consist of constant tones. Constant tones are easier to measure and reproduce than impulses in music.
Then we also have timing errors (easily spotted in speaker measurements) and when all is perfect we still have rooms/acoustics that can 'f things up.
I understood that the first wave front is quite important for music and that wavefront amplitude may not be the same for all frequencies yet while measuring the FR may well be 'flatter' than the impulse response.

1: desirable but see my remarks above.
2: noise can only be 'perfect' when averaged over a longer time period.
3: to a degree we can compensate for certain aspects at a selected listening spot. It will deviate at another position and that deviation can be substantial.
4: We know what flat is. It can easily be achieved with electronics. Transducers, acoustics and humans involved are the weak points.