Measurement microphone specs and distance needed for measuring speaker distortion

[Mr. Haelscheir]

In short, how can we know if a measurement microphone is suitable for measuring speaker distortion with respect to frequency at some reference down to, say, 0.01% THD?

Forgive any misconceptions I have in the following:

Not all measurement microphone manufacturers publish their distortion measurements. Taking Earthworks for example, in advertising "140dB SPL rating without distortion" and "20dB SPL equivalent (A weighted)" self-noise, does that mean that the microphone's distortion products will remain below the 20 dB self-noise floor until the microphone starts receiving more than 140dB SPL? If so, am I right to assume that with a sufficiently large input signal (the microphone sufficiently close to the transducer after setting the reference level at the reference distance) and sufficiently low ambient noise, these Earthworks microphones could measure up to 120 dB SNR or amplifier-level 0.0001% THD, only limited by the microphone preamp's and whatnot's own noise and distortion? Now, Brüel & Kjær specifies a "3% distortion limit" coincident with the stated dynamic range", whereby if the distortion curve is similar to the one shown in https://www.isemcon.com/datasheets/EMX7150-US-r04.pdf, then it is perhaps rather around 20 dB below the max SPL that the distortion products start rising above mic's self-noise, suggesting a maximum measurable SNR of 100 dB or 0.001% THD (though that PDF seemed to show a floor of around 0.15% THD which is well above the self-noise at those levels).

Furthermore, in setting the reference level (e.g. 86 dB) at the reference distance (e.g. 1 m), is it valid to afterward move the mic closer to increase the maximum measurable SNR?

 

[Joachim Herbert]

How did you come up with 0.01% THD for speaker distortion?

 

[staticV3]

How did you come up with 0.01% THD for speaker distortion?

Good speakers can reach down to about 0.03% THD, so wanting your measurement setup to be capable of at least 0.01% for some headroom seems perfectly reasonable to me.

 

[Mr. Haelscheir]

How did you come up with 0.01% THD for speaker distortion?

Staring at the juicy low parts of Genelec's THD+N versus frequency charts...

 

[Joachim Herbert]

I own a pair of these. Never thought they are below 0.03% THD, but they obviously are. Looking at bass and lower mids its more like 1%, though. Thank you for pointing this out anyway.

To translate between dB and % see this handy calculator: http://www.sengpielaudio.com/calculator-thd.htm

 

[fineMen]

My mini Panasonc capsules with bootstrap-amp (feedback) generate some distortion for sure. I can easily guestimate how big it is with changeing the distance to the D.U.T. Up to about 100dB it is too low to argue with it like < 0.05% which isn't audible anyways.

In bass the limit is determined by the necessary elastic movement of the diaphragm of the mike.

 

[Mr. Haelscheir]

I am so far hoping that an Earthworks M23 with Focusrite Scarlett 2i2 will suffice considering that the M23's THD in https://www.mtg-designs.com/tips-tricks-tests/measurement-mic-tests/hd-measurement-mic-tests was the only noise-limited mic measured, unless there are better alternatives in that price range.

 

[Blumlein 88]

I am so far hoping that an Earthworks M23 with Focusrite Scarlett 2i2 will suffice considering that the M23's THD in https://www.mtg-designs.com/tips-tricks-tests/measurement-mic-tests/hd-measurement-mic-tests was the only noise-limited mic measured, unless there are better alternatives in that price range.

I don't know what to think of his results. Here is Umik 1 measuring at a few inches registering a little above 90 db SPL. I show it in dbR to match MTG views. You can see the 2nd harmonic is much lower than MTG found.

Also surprisingly low was a ribbon microphone. Used under the same conditions.

The similarity in the shape of the curve vs frequency makes me think the speaker was the main factor in the distortion.

 

[fineMen]

In short, how can we know if a measurement microphone is suitable for measuring speaker distortion with respect to frequency at some reference down to, say, 0.01% THD?

In the old days my granny would listen to a valve powered radio. Distortion of less than 1% were considered exceptionally good. Today with lowest distortion high gain op amps you can't no longer count (and memorize) the zeros following the decimal point. All the trouble is hidden behind massive feedback, so that one won't tell a reasonable from a not-so good model by its distortion figures anymore. Amplifier distortion became a fetish in the 70s (last century) as soon as the numbers went down, and so could be used in advertizing, batteling the competition. It went to even inventing 'new' forms of distortion, that were, sure enough, an unforgivingly brutal threat to the music.

Are we starting that same game with speakers now?

Again and reiterated, harmonic distortion is at some point completely masked by the musical content that needs its own harmonics to simply exsist! In bass a grand total of hundreds (!!) of percent of harmonics is not to be accepted, but to be appreciated. Even tens of percent of HD won't shift the sonic impression of a bass instrument, let alone the room's reverberation, that will introduce strong comb filter effects, altering the effective distortion figures, and--you name it, the relations of the musical harmonics to the base tone. That room effect will yield the very same effect, a shift in the sonic signature, as harmonic distortion does.

Another is intermodulation. IM is too complicated to be communicated to the general public. Hence, no interest.

Coming back to microphones, better measure the intermodulation. I did, it is way higher (w/ my Panasonics) than HD.

 

[DS23MAN]

Don't try to do distortion measurements with those cheap 6mm capsules:

Microphone Distortion Comparison

In this blog post I compare distortion performance between four different microphones. From left to right I have the following microphones. Dayton UMM-6 I've used the UMM-6 for about a decade now and it has served me well. It is the only USB microphone in this test. The microphone retails for...

josephcrowe.com

Even a cheap dynamic mic is better!

I bought a very expensive NTI M2215 ( same class as B&K and Grass).

https://www.nti-audio.com/Portals/0/data/en/Measurement-Microphones-Specifications.pdf

Does 153 db at 3% distortion. Something about the self noise in microphones. It is related to the noise in a space and it's own generated noise. For example a mic with 25db specs can't measure below 25 db noise in a space. Believe me 25db in a space is very,very, very low. In night at a very rural area you maybe expecting 30db without wind outside.

 

[fineMen]

Don't try to do distortion measurements with those cheap 6mm capsules:
...

I advocate the contrary. Get yourself a measuring microphone, even if it has to be cheap for obvious reasons. Reasoning:

with music and speech and noises and what have you, distortion products as IM and HD below 60dB are barely detectable by even trained individuals (exceptions apply), at least when listening for recreational or educative purposes other than the infamous 'critical listening' as kind of a contemporary sport

speaker's defects won't show in distortion figures, proven by the effort that Klippel takes to detect e/g rub&buzz, or they show absolutely clearly like a spike in all(!!) HD at a ratteling point

Conclusively an, admittedly well chosen, Panasonic mini capsule like WM61 (?, the Linkwitz type for 2$) will suffice.

As an example I'm absolutely able to show the IM of a KEF R3 coax (tweeter within midcone) to be too low to be relevant @-60dB and less, but can also show that only the Doppler effect limits the mid's performance with higher volumes. I don't know of any of the owners of fancy microphons to share such nasty knowledge. Must be the reason that it needs to start out ...

 

[DS23MAN]

It will not. Those cheap capsules can't stand high sound pressures and have very high noise floors. They don't have a "dynamic" range.

 

[Mr. Haelscheir]

Yes, I am avoiding especially cheap capsules or anything that clearly could not attain the desired SNR when moving the mic closer to the driver amid a 40 dB noise environment. I do remember Joseph Crowe's measurements, but it would be good to know if the Earthworks M23 could achieve the target performance without having to dish out as much for something in the price range of the ACO.

 

[staticV3]

Those cheap capsules can't stand high sound pressures and have very high noise floors.

I use a cheap Primo EM272 10mm Electret capsule as a headset microphone.
It has a self noise of 14dB SPL A-weighted and a maximum input level of 119dB SPL.
It is by far the quietest mic I own.

 

[Mr. Haelscheir]

Regardless of what cheap electret capsules can do, I think the question is of whether amid a 30 to 40 dB background noise, the microphone could still exhibit no more than 0.01% of its own harmonic distortion at the 120 dB fundamental (I suppose this could be relaxed by 10 to 20 dB depending on the actual levels of the noise within the frequency bands of importance) at X distance below the reference distance needed to assess driver distortion down to 0.01%.

 

[fineMen]

Regardless of what cheap electret capsules can do, I think the question is of whether amid a 30 to 40 dB background noise, the microphone could still exhibit no more than 0.01% of its own harmonic distortion at the 120 dB fundamental (I suppose this could be relaxed by 10 to 20 dB depending on the actual levels of the noise within the frequency bands of importance) at X distance below the reference distance needed to assess driver distortion down to 0.01%.

Do I get it right, the noise level is related to the minimum HD that can be detected? Do you take into account that the noise level is integrated over the full frequency spectrum, while HD and IM are usually measured at some certain frequencies, in a narrow part of the spectrum? To put it more visually, to see a line at -60dB 'on spot' within a sea of noise at -60dB 'integrated' should be relatively easy. Only additionally one should also take into account, that noise is frequency dependent. Especially with 'cheap' FET transistors. In case your calculations could be improved.

But I admit, some of my Panasonics run in a feedback design (bootstrap with OP, phantom power), some other use the symmetric Schoeps design with quite similar behaviour. Anecdotally, I'm happy to record the bats in my backyard with either of them.

I already discussed the feasibility of measuring distortion to such low levels for speakers and headphones.

I only wonder how you are going to evaluate the results ;-)

Add.: only to illustrate the case, Panasonic with high self noise (not betterned by feedback) recording in room in presence of environmental noise wind, traffic, children a single sine; 0,1% of distortion no problem and there is still 30dB headroom and averaging isn't used yet

 

[Blumlein 88]

Here is an idea to try and bypass distortion from the speaker. Two different speakers produced one tone each. Recorded with microphone about 10 feet from the speakers. I used 1100 hz and 1400 hz. Idea being no IMD from speaker, but IMD from the microphone will be revealed. Used the front speakers in my theater setup. So some 11 feet apart.

Here is the Umik 1. Couldn't quite get each tone the same level, but close. Used REW in RTA mode with spectrum display using 256k FFT and 8 averages.

Here is an Avantone CK1 SDC with omni capsule. Tried to get levels and position exactly the same. Used a cal file for it derived from published FR charts. I've checked it before and it is a close match for the Umik-1.

Here is a Warm Audio SDC WA84 which is a clone of the old Neumann K84. Don't have cal file for it on this computer, but it hardly needs much of one anyway.

 

[fineMen]

Here is an idea to try and bypass distortion from the speaker. Two different speakers produced one tone each ...

Not much of a difference, right? Noise: no issue at all.

I now think that the original question originates in a misunderstanding. If, presumably, the thinking was like:

I want to measure 0,01% which is equivalent to -80dB. So the noise must be at least -80dB down. But down from which reference? Ask the board about speaker level and distance ...

As already explained, the noise value is given as an integral over all the spectrum, while distortion is a measurement for a few single frequencies. How much noise affects the measurement depends on many factors. It depends strongly on the digitized measurement method. The latter shall be clearly understood which needs a university grade understanding of mathematics in general.

btw: it would be cost saving to start the intended investigation with a cheap microphone first, and only later once the concept is established buy (if still needed) a more expensive device.

 

[Blumlein 88]

Results are so similar on the IMD sidebands I wonder if there isn't a problem. All three measured lower harmonic distortion of the two test tones by a good margin. And very similar making me think they were measuring the speaker's actual distortion. There were peaks and nulls which made positioning the mics tricky. I was able to get pretty close it would appear. If you look at the low end you can see the fridge cut off during the 3rd measurement.

I knew that concerns about the noise were somewhat misplaced having done some such measures previously. FFT and averaging lower the noise floor so you can uncover tones and harmonics and such. I could have used 4 meg FFT's to drop it lower, but it wasn't needed. Moving the speakers closer together further from walls and the microphone closer would have helped clean things up some too.

 

[Blumlein 88]

Thought I would post some simple distortion measures. One speaker. First is the Umik 1. Don't worry about the bad THD+N numbers. Trains, and planes and air conditioners kept interfering. I think it only effected the low end.

Then the CK-1. This was a rare quiet moment.

For something different a Shure KSM 44 in omni mode.

Finally, a CAD M179 in omni mode.

In case your wondering I did all of these at least 3 times and a couple 4 or 5 due to noise issues in the low end. Other than the low end each run produced the same results to within 1 or 2 in the 2nd digit past the decimal place for each harmonic.