A poor (wo)man's microphone measurement equipment

[Roman Beilharz]

Kind of like with distorted electric guitars, it is extremely easy to hear differences in frequency response simply because all the frequencies are there the whole time and you get all of them. With vocals for example, it can be tricky because a lot of HF where most the differences are found are transient in nature, and short lasting. I like drum tracks as well as i get long tail of cymbal decay.

Alright, that makes perfect sense now.

I use these takes with Eq match software (CurveEQ, Fab Filter) to match similarly constructed microphones if i need to. Say u87 and a 100$ k67 based mic with similar headbasket. You would be surprised how close you can come. If i suddenly need a matched pair, this helps big time.

Clever! Of course; quickly done . Thanks for sharing!

To a certain extent. A lot of mythology spills from guitar amp world. However, here we use tubes in pretty linear region, with for example vocals under 1% thd. So how a certain tube saturates is really not important. It is sometimes questionable if you could really tell the difference between a FET and tube, presuming everything else being the same.

I thought so, but wasn't too sure about it. I can easily distinguish mics using transformers from transformerless designs; differences between transformers seem harder to tell. So the sound of my U47 build seems more related to 1. an effective marriage of headbasket and capsule (design) and 2. the transformer than to any other variable in the game.

Most of the differences in sound come when you start swapping tubes due to some differences in purely electrical properties. For example plate resistance of a tube can interact differently with following capacitor and transformer. You could get some resonances in the low end, or frequency response could change. Also bias point of the tube could be different. But that is no magic, or mojo. 220pf range capacitor in Elam 251 will shift cutoff point big time depending on internal resistance of the tube. Which can tremendously impact the response of the high end. So certain tube might sound "warmer" than the next one. However no magic here, just stupid change of the frequency response of the circuit.

Yep, makes sense. Thanks again.

I am anal when it comes to empiricals when i build and discuss gear. How else would you clone a certain mic. The very second i start doing creative work i start thinking intuitively, and forget about every measurement i ever made.

Seems to be the way to go. Changing hats for different modes - great, if you can pull it off. Working at it

Cheers

Ro

 

[Mr. Moscode]

What speakers do you folks use when doing your mic tests? Cone, electrostatic, coaxial?
What brands and models?

I've been very curious about this.

Garcias

 

[kingkorg]

Coaxial are preferred for this purpose.

 

[Mr. Moscode]

I can see that. What do you trust to be flat? Make and model? What's your pref?

 

[kingkorg]

I can see that. What do you trust to be flat? Make and model? What's your pref?

I'll repeat again what i wrote before. "You can use almost any speaker, the imperfections of the speaker are compensated by a known flat reference microphone."

 

[MicUlli]

I'll repeat again what i wrote before. "You can use almost any speaker, the imperfections of the speaker are compensated by a known flat reference microphone."

This statement is completely true for comparing mics of identical directionality. Typically one uses an omnidirectional pressure sensitive mic as reference. If your device unter test is also an omni it is only important to place it at the same coordinates as the ref mic (every cm counts!).
Another influence factor might be that the stimulus signal is at least 30 dB higher than all noise components (room noise, electronic noise). A log swept sine is for example better than pink noise.
Things become complicated when the DUT is a cardioid or fig8 and the ref mic is omni. Then the sound source distance (and shape) may influence the frequency response (proximity effect). Also standing waves may play an important role (omnis see nodal points, pressure gradients see antinodes at least on axis) so windowing the impulse response becomes important. That's why IEC 60268-4 demands a plane wave field..
I also want to point out that single diaphragm pressure gradient mics behave different with respect to double diaphragm.

 

[kingkorg]

Which is again why i take measurements outdoors if i want them precise. You get anechoic measurement, and you can forget about the issues mentioned.

All one needs is reasonably flat studio monitor, take measurement outdoors, make the measurement at the point where proximity effect stops, typically about 40cm. Known flat 100$ range mic like Umik-1. Measure DUT, measure reference mic at same exact position, divide DUT measurement by reference mic and you will get a precise result comparable to any commercial anechoic response.

If you are not satisfied with ugly jagged line, and want baby bottom smooth response, don't despair, you don't need one unless you want to publish the measurements somewhere or do this professionally. Don't smooth the measurement because you need to see al the quirks which point to potential issues with either DUT or technique used. There is a misconception that smoothed responses are in line with what we are able to hear, and how we hear. However we are not after this, we can hear it already, we are chasing things we are not able to hear in order to point to issuues, performance, and learn from it.

You can also do this in a regular well treated studio room, but forget about low end precision, which you don't even need because there's nothing crazy going on there anyways with condensers. All the character is in the mids and the high end.

I am not necessarily adressing @MicUlli in this post, just explain to average hobbyist how to do it in simplest terms, and all that is needed. Ulli seem to know everything he needs to, so no point for me to "teach" him anything. Not like he would appreciate it anyways. I again suggest him to make detailed description with preferably pictures of his approach before fishing for something he obviously lacks.

Chasing IEC, and other standards won't get you anywhere, unless you are after commercially viable measurements. Many manufacturers have to obey these, yet you get rubbish products and specifications.

There is absolutely nothing more to know about measurements that can't be found on the two YT channels i shared.

Contrary to what Ulli said before, there is quite a lot money to be earned in this area, especially R&D. Otherwise "Listen", and Amir form ASR would be in a serious trouble. These measurements are not just about mics

 

[ct.waveform]

Everything you need regarding basic and advanced measurement techniques can be found on these two channels. I, selfishly, am hesitant compiling them in a simple thread, and servin them on a plate because of earlier mentioned reasons. Probably two most underrated YT channels ever.

https://m.youtube.com/@ListenInc
https://m.youtube.com/@AudioScienceReview

I like an awful lot of what Amir and Audio Science Review does but, boy, are they fixated on measurements being the be-all and end-all of Audio! There are at least half a dozen ways in which the signal can be corrupted which don't show up in a steady state measurement, before we even get to the distribution of harmonics. It's misguided to the other extreme from the subjectivist position of listening being the only test - and any batshit improbable mechanism being the cause of why something sounds good or bad.

I haven't come across ListenInc so will try and have a look at it later tonight.

 

[kingkorg]

I like an awful lot of what Amir and Audio Science Review does but, boy, are they fixated on measurements being the be-all and end-all of Audio! There are at least half a dozen ways in which the signal can be corrupted which don't show up in a steady state measurement, before we even get to the distribution of harmonics. It's misguided to the other extreme from the subjectivist position of listening being the only test - and any batshit improbable mechanism being the cause of why something sounds good or bad.

I haven't come across ListenInc so will try and have a look at it later tonight.

Well we are talking measurements here. Example. When i change capsule spacer thicknes by couple of microns, or thickness of a diaphragm both of which might introduce change of 1db at 16khz, i have nothing else but measurements to go by.

Going by ear has lead many mic "gurus" to make products that are all over the place QC wise.

Also if i make a measurement of a 10yo mic, and brand new one bought in another part of the world, and they measure exactly the same, that has a lot to say about a certain company making them.

 

[Roman Beilharz]

What speakers do you folks use when doing your mic tests? Cone, electrostatic, coaxial?
What brands and models?

I've been very curious about this.

Garcias

I am using a Sky Audio Verdade 2.1 system (2 way, magnetostatic tweeter with 8" woofer) with KS digital B88 sub. But coax systems offer the advantage of being able to position your mic more freely and still get coherent signals.

 

[ct.waveform]

What speakers do you folks use when doing your mic tests? Cone, electrostatic, coaxial?
What brands and models?

I've been very curious about this.

Garcias

FWIW, if we are talking about referencing one mic against another (known) mic, I just use a tweeter mounted in 2pi with no crossover. Your limits (the first reflection) are the same as with a full range speaker and you can know the T/S parameters of the tweeter in their entirety, if that comes into the assessment later on. The main differences will be above 10 kHz anyway and most 1" tweeters still have plenty of output at 1kHz, which is another area where things might differ. You can easily see differences in sensitivity, even low down, and you only have one axis to consider (not multiple paths). I use 7deg off axis as my principle measurement axis as "on-axis" is kinda useless and insignificant in the scheme of things. As you would imagine, it's also far more repeatable. The O-A response represents a tiny dot in the room-sized target of concentric circles that the tweeter is radiating into and is nothing significant in terms of power in the room. I have been lucky enough to have had Clio almost since it came out, but I expect Holm Impulse or REW can do something just as useful.

I can't think it's terribly important what speaker/driver you use, in general, but one of the nice things about using a tweeter is that you know it'll be pistonic in that low range. It's basically theoretical, unless it has a second chamber, which they don't always get to behave as a single compliance. You may not get very much below 600Hz, but you won't have any mechanical anomalies to worry about. (Do feel free to shower me with papers on the non-linearities of tweeters at LF ) Nonetheless, I can think of half a dozen horrors that a midrange and bass unit could introduce. So, yes, just use a tweeter.

As for brands, I think that's less important than choosing a tweeter that's smooth. (Or choose one you think you are going to want to use.) It'll be important that it doesn't have a phase plate, and for smoothness you really will be better with a silk dome. SEAS, who are friends of mine for full disclosure, do a 26mm silk dome which I think goes down quite low - say 750Hz?. It's kinda near textbook. That's what I would choose in re-referencing mics. Or, if you are on a budget, their 19TFF is a jewel. Almost no one has noticed how good that tweeter is. Of course Scan Speak is better known for low resonant frequencies in their tweeters, but most of their Classic tweeters struggle to get beyond 15kHz (which doesn't stop them sounding nice) so you lose some of the top end in searching for more signal lower down.

 

[Mr. Moscode]

Thank you for the info. I keep thinking and electrostatic panel would most likely measure flat. I have not done f response testing on them. I have Quad ESL 63s.

I guess if you confine the tests to a tweeter at 1K and above that would tell you a lot. Phase issues don't matter, i assume, if you are near field to the tweeter. How far away would you put the mic?

And a known calibration mic would be essential.

 

[MicUlli]

Lot of useful comments have arrived yet

I fully agree using a tweeter. I myself prefer a SB26CDC-C000-4 (SB Acoustics) as main sound source for FR measurements. It is mounted in a pyramid like paper construction (to avoid too much edge reflection) and placed on the floor. The pyramid construction serves as a wave guide to the floor. This leads to a nearly ideal 2 Pi radiation. The tweeter is driven by a small self built amplifier (don't laugh, its based on the old TDA2030 chip). The tweeter exitation voltage is 1,6 V RMS and i get an SPL of approx. 84 dB for frequencies higher than 700 Hz (in 1 m measurement distance). At low freq. there is a rolloff with
40 dB/decade below 700 Hz.

The trick: reducing the measurement distance to 10 cm produces 104 dB SPL. Even at 20 Hz SPL is 42 dB. This is the way for me measuring omni mics against my reference mic. Room modes/reflections are very small because the walls are at least 2,4 m away.

BTW: My ref mic is a calibrated ECM8000 (one of the older ones, with transformer and WM61 capsule). At the beginning of this century i did a comparison against B&K 1/2" ref mic in an anechonic room at Ruhr Uni Bochum (Germany) and use the deviations over frequency as cal file in REW. From time to time i check for sensivity with a calibrator and found the mic to be very stable (< 1 dB in lifetime yet).

Measuring directional mics is done at a distance of approx. 30 cm. In order to rise SNR 8 sweeps are performed and averaged (REW does it for you). For cardioids this leads to a 6 dB/oct rise with 3 dB point at 95 Hz (and must be compensated).
Edit: Windowing the IR (12,5 ms) leads to very flat meaningfull FR down to 80 Hz.

 

[ct.waveform]

Thank you for the info. I keep thinking and electrostatic panel would most likely measure flat. I have not done f response testing on them. I have Quad ESL 63s.

I guess if you confine the tests to a tweeter at 1K and above that would tell you a lot. Phase issues don't matter, i assume, if you are near field to the tweeter. How far away would you put the mic?

And a known calibration mic would be essential.

You probably could use ESL 63s without trouble but I wouldn't get carried away with the idea of electrostatics being flat. People imagine they're weightless and nudge the air just the right amount, so are therefore THE perfect speaker. Sadly not. I'm not suggesting their anomalies matter to the sound necessarily, but their normally large surface area means they suffer badly from combing both on and off axis. That's about the last thing you want in something you're hoping to use as a repeatable reference. Repeatability and consistency are the two most desirable qualities here, so any sort of rapid fluctuation in response is undesirable. The design of the 63s as a point source does get over precisely this problem (at least I assume it does, though I have never measured one) but I don't think the same goes for any other electrostatic.

A tweeter, which incidentally should be flush mounted (and ideally set in a large smooth wall), should give you at least an octave below it's main resonance. That SB tweeter is 690Hz so we're looking at getting down to at least 350Hz with a single measurement. At that frequency you are approaching the limit of your first reflection anyway, so you are not losing much (say 12-15dB in signal) over having a bass unit or bass-midrange. Fine tweeter though I imagine it is, this one is not really ideal for this task. The phase plate will probably give you quite noticeable amplitude and phase shifts for small changes in angle near 0 deg, and the very rigid material is probably quite spiky underneath the 1/3 octave smoothing. Of course you can still come away with a usefully functional mic response but you will have lost some information in the smoothing. However, especially when budget is included, wanting to use a tweeter trumps choosing the ideal tweeter for this task.

If you want to get meaningful results below 300Hz then we're starting to look at nearfield measurement with known pistonic drivers - or I might try measuring *inside* the (sealed) cabinet - but then it gets a bit messy knitting it all together. Of course this information can be carried around either in the tweeter or in the mic. A FR of a given tweeter from, say, SB, taken on a calibrated mic in known conditions essentially calibrates your mic when you repeat the measurement. Of course there are too many calibrations from other calibrations doing the rounds these days and it would be nice if someone like SB, who engage with the DIY community, offered calibrated data with some of their tweeters at a small premium for exactly this purpose. I don't think I count a UMIK, mentioned somewhere above, as a calibrated mic (any more than I do my own Audiomatica WM61 based mic). I have heard of JLI International helping out projects with calibrated capsules, which is extraordinarily helpful and is another possibility, but I don't know if the lone DIYer can get the same. But their capsules are a good place to start (though their website means a LOT of leafing through pages unnecessarily; it's pretty infuriating). They do a very tidy equivalent to the WM60/61A and they have some nice 12mm capsules and I imagine they are the source for the AP measurement mic capsule.

I do all my measurements at 2m, and have done for decades! I don't really have much truck with measurements at 300mm or 400mm (except doing 2pi measurements), or any distance where you can visualise the trigonometry of multiple sources adding up at the microphone. Doing compromised measurements because they don't think they've got enough room is kinda the hallmark of the DIYer - and the first way in which the DIYer handicaps himself. With the kit we've got available today, there's no reason for anyone not to be able to measure just as well as someone like KEF. It might not be so effortless or as flashy, but there's no reason for the data to be any less good.

For this mic application I'm not sure how helpful it would be here to use a short distance, except perhaps with a cardioid or a dipole, but you can probably get away with it. Even then I think one would be best served by blocking up the holes and measuring two responses which you can then add together synthetically. This can then tell you how close to an omni plus a dipole it is, and you can check that by placing it at various distances.

 

[ct.waveform]

Well we are talking measurements here. Example. When i change capsule spacer thicknes by couple of microns, or thickness of a diaphragm both of which might introduce change of 1db at 16khz, i have nothing else but measurements to go by.

Going by ear has lead many mic "gurus" to make products that are all over the place QC wise.

Also if i make a measurement of a 10yo mic, and brand new one bought in another part of the world, and they measure exactly the same, that has a lot to say about a certain company making them.

I don't think my position on measurements is anywhere near where you imagine it to be. Nothing should be done by ear alone, especially not with transducers. I'm not even keen on people "voicing" loudspeakers. If it is not working straight off the computer, then tweaking it and pretending you are a guru is NOT the answer. Re-examining your design axioms or priorities is the requirement - and maybe tossing these ones in the bin as you search for a better answer to what is going on.

My complaint about ASR is the blinkered and unscientific view of SINAD being the only needed figure of merit. I don't think we even have any studies on the audibility of ultra-low distortion, let alone a correlation between low distortion and sounding "good". The depth of ignorance epitomised by this white-coated "faith" is tiresome and frustrating. A friend of mine wrote an article on whether there was 1st Harmonic distortion. This made the important point of "where the hell do you think the energy to excite those other harmonics comes from?" and it's fun to think that the fundamental is the one frequency we don't look at when measuring THD. Nor is any consistency tested for. THD measurements as they stand at the moment are an astonishing thing to put your faith in. If we take just one mechanism we know exists, power supplies contaminating the signal, what happens in a THD measurement? We blitz that frequency down to -130dB (which is so thorough that it's the equivalent of ethnically cleansing Britain down to just 20 Englishmen) so we haven't a trace left of what's going on. But we know that 100% of that ripple has been dropped onto the signal because the PSR is a direct overlay of the Open Loop Gain, and is doing no better. And we know that there has to be ripple because the power supply has an output impedance. (And, FYI, your current sources/sinks do near sod all to stop rail voltages getting into the signal path.

Amplifier measurement is an inconsistent and unscientific mess and we really need to start again. Having faith in what we've got is a pretty good sign that someone hasn't thought about it. This will be the same kind of cretin who is adamant that loudspeaker cable can't sound any different (even though he knows they affect the stability of the amp, act as aerials, change the frequency response at the speaker terminals, have no impedance matching, have a frequency response of their own, are asking the amp to deal with frequencies way above what it can't manage, including reflections, and are just one resistor away from the input stage). What kind of "science" brushes all this under the carpet and dismisses it as irrelevant? The amplifier quite literally doesn't know whether a voltage on the speaker cable is an input or an output (but we know that doesn't affect the sound). Yeah, right!

 

[kingkorg]

I don't think my position on measurements is anywhere near where you imagine it to be. Nothing should be done by ear alone, especially not with transducers. I'm not even keen on people "voicing" loudspeakers. If it is not working straight off the computer, then tweaking it and pretending you are a guru is NOT the answer. Re-examining your design axioms or priorities is the requirement - and maybe tossing these ones in the bin as you search for a better answer to what is going on.

My complaint about ASR is the blinkered and unscientific view of SINAD being the only needed figure of merit. I don't think we even have any studies on the audibility of ultra-low distortion, let alone a correlation between low distortion and sounding "good". The depth of ignorance epitomised by this white-coated "faith" is tiresome and frustrating. A friend of mine wrote an article on whether there was 1st Harmonic distortion. This made the important point of "where the hell do you think the energy to excite those other harmonics comes from?" and it's fun to think that the fundamental is the one frequency we don't look at when measuring THD. Nor is any consistency tested for. THD measurements as they stand at the moment are an astonishing thing to put your faith in. If we take just one mechanism we know exists, power supplies contaminating the signal, what happens in a THD measurement? We blitz that frequency down to -130dB (which is so thorough that it's the equivalent of ethnically cleansing Britain down to just 20 Englishmen) so we haven't a trace left of what's going on. But we know that 100% of that ripple has been dropped onto the signal because the PSR is a direct overlay of the Open Loop Gain, and is doing no better. And we know that there has to be ripple because the power supply has an output impedance. (And, FYI, your current sources/sinks do near sod all to stop rail voltages getting into the signal path.

Amplifier measurement is an inconsistent and unscientific mess and we really need to start again. Having faith in what we've got is a pretty good sign that someone hasn't thought about it. This will be the same kind of cretin who is adamant that loudspeaker cable can't sound any different (even though he knows they affect the stability of the amp, act as aerials, change the frequency response at the speaker terminals, have no impedance matching, have a frequency response of their own, are asking the amp to deal with frequencies way above what it can't manage, including reflections, and are just one resistor away from the input stage). What kind of "science" brushes all this under the carpet and dismisses it as irrelevant? The amplifier quite literally doesn't know whether a voltage on the speaker cable is an input or an output (but we know that doesn't affect the sound). Yeah, right!

Totally see where you are coming from. I just thought these two YT channels have enough purely technical info on how to perform and approach the measurements. Interpretation of the measurements is whole nother can of worms.

 

[godzich]

Hi,

Yeah, it's a must to use a reference mic when calibrating a small anechoic chamber for flatness. However, would not hurt to start with the best and flattest possible sound source. I have been thinking of salvaging the transducer from a broken pair of studio-quality headphones. We are talking near-field of about 10cm to the mic. Anyone that has done this already? Those small speakers should have a decent frequency range of 10Hz to 25kHz and are reasonably flat. Even the back-side damping could be used. One worry is if there will be enough sound pressure - without distorting...

 

[ct.waveform]

Hi,

Yeah, it's a must to use a reference mic when calibrating a small anechoic chamber for flatness. However, would not hurt to start with the best and flattest possible sound source. I have been thinking of salvaging the transducer from a broken pair of studio-quality headphones. We are talking near-field of about 10cm to the mic. Anyone that has done this already? Those small speakers should have a decent frequency range of 10Hz to 25kHz and are reasonably flat. Even the back-side damping could be used. One worry is if there will be enough sound pressure - without distorting...

I'm afraid almost nothing has a flat response, sadly. Even in regions where you are absolutely certain the cone/dome is perfectly pistonic there are mechanical impedance mismatches with the surrounds or spiders. Then there are fundamental resonances, slopes caused by rising impedance meaning less current, variations of dispersion due to cone or waveguide profiles and even on a 75mm driver the inherent directivity is already rising at a little above 1kHz. (Definitely noticeable and significant at 1600Hz.) And if we are talking about a box - which I assume we'll need for an anechoic chamber - there's the 6dB baffle step rise, which is usually accompanied by a 1-2dB overshoot at the top end to make it even more of an irritant.

But we don't actually need a flat response anywhere - just something smooth-ish. We will be dividing one response by the other, so taking out response entirely. This is easier to see logarithmically, where we are subtracting one from the other and we're only interested in the difference. The requirement then becomes that the two measurements keep the fixed elements identical, or near enough. That means staying away from fully on-axis measurements (which are a special case) and not having much variation with angle and being sure about your distances.

 

[kingkorg]

Those small speakers should have a decent frequency range of 10Hz to 25kHz

That's when they are sitting on your head with proper sealing. Remove it and you get no low end to speak of.

 

[godzich]

That's when they are sitting on your head with proper sealing. Remove it and you get no low end to speak of.

Yeah, you are right - to keep the freq. response I would need to keep the muffle and enclosure - like putting the mic under test inside the "ear" of an artificial test head... Not very convenient... Especially when testing older condenser mics that are huge... Even if I know much about electronics my acoustic knowledge is a little lacking. Back to the drawing board