A poor (wo)man's microphone measurement equipment

[kingkorg]

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

Amir at ASR has posted many measurements of a lot of speakers, some of which are very inexpensive. Some of these are not ideal as studio monitors as they have different quirks related to dispersion, hyped response in one or another way. However many of them have very smooth responses without any crazy dips or bumps. Even if not ideally flat, they are great for 0° measurements as the irregularities in FR will be compensated by reference mic.

 

[ct.waveform]

Even if I know much about electronics my acoustic knowledge is a little lacking.

I hope this remark wasn't prompted by my earlier reply. I assure you it wasn't intended as a piece by piece destruction, though I see it could look like that. I have for some while been thinking about doing a YT video itemising the facets of driver frequency response, and how they originate, to help DIYers choose their drivers. So the list, which is meant to be helpful, was mentally quite close at hand.

Much of what has been discussed here touches on quite advanced areas of acoustics. In the case of the headphones it's radiation resistance, which is something I have seen PhDs get completely wrong (and in one instance be awarded a doctorate for getting entirely wrong! ) Headphones can get an extended response because they pressurise the chamber at low frequencies and notionally this rise at 12dB/Octave continues all the way to DC, cancelling the normal high pass of the driver. Of course there are losses and leaks in practice, but it's still quite a thing! People have used this principle, measuring *inside* the cabinet, to get very accurate measurements of what the driver is doing mechanically, both in its pass band and a long way into its stop band.

 

[kingkorg]

The devil is in the details. Neumann's QC "anechoic" chamber is rather tiny. I purposefully put the anechoic under quotes. They also utilize the floor.
Screenshot of How it's made - Neumann u87.

 

[ct.waveform]

The devil is in the details. Neumann's QC "anechoic" chamber is rather tiny. I purposefully put the anechoic under quotes. They also utilize the floor.
Screenshot of How it's made - Neumann u87.

I have to say I'm a bit at a loss as to why anyone would build an anechoic chamber these days (or really at any point since KEF built their 8m cube for their Eureka project in the early '80s). We aren't reliant on continuous sine wave measurements any more, and if we want to use them then there's always outdoors. Aside from noise questions, an anechoic chamber (which is never as perfect as no walls at all) could be seen just as an excuse not to get cold.

 

[MicUlli]

Regarding the frequency response measurement of (hyper- or super-) cardioid SDCs there may exist a powerful method for estimation of the low frequency behaviour. Here is a short explanation:
SDCs have an omnidirectional part O (sound pressure component) and a directional part D(phi, jw, d) (pressure gradient component). Overall response is

FR = H(jw)*(O+D(phi, jw, d)).

phi is sound incidence angle, jw is complex radiant frequency and d is distance from sound source.

D has a figure8-shape related to phi. Therefore

D(0°, jw, d) = -D(180°, jw, d)

If we measure the two frequency responses FR(0°) and FR(180°) and add them together we get

FR_add = 2*H(jw)*O

This means that we can compare FR_add directly with our omnidrectional measurement microphone. Of course this is a simplification but it gives a very close matching. Give it a try

BR MicUlli

 

[MicUlli]

I promised to contribute some basic aspects regarding mic measurement and will start with a basic requirement: Calibrate your audio interface!
Although this seems to be a trivial task one (but of course not the only) approach can be found in the appendix

BR MicUlli

 

[ricardo]

MicUlli, please don't be put off by the naysayers but continue to detail how you measure mikes. I have my own methods based on Prof. Angelo Farina's swept sine but as they involve having Windows XP with Virtual PC running Win 98 running a DOS window, they are not really suitable da 21st century
One of my requirements is that curves must match a measurement on a B&K 2307 chart recorder but it's difficult to get the steam to run it today

My $0.02 is that you should use a Quasi-Anechoic method allowing you to get truly anechoic results down to 200Hz in an empty domestic room and make a good estimate of performance below that using Benjamin's method.
AES E-Library » Extending Quasi-Anechoic Measurements to Low Frequencies

For the ONLY good explanation of Quasi-Anechoic, you need to read the CLIO manuals, version 10 or earlier. I highly recommend CLIO for an inexpensive acoustic measurement setup, especially for production testing.
How Do You Test Your Mics?
If you know of a better explanation, please post.

Though I can now do better measurements with Angelo's method than I could with a big anechoic in da previous Millenium, I would want a big anechoic AND Angelo's method if I was making microphones and speakers for a living again.

 

[ricardo]

FR_add = 2*H(jw)*O

This means that we can compare FR_add directly with our omnidrectional measurement microphone. Of course this is a simplification but it gives a very close matching. Give it a try

MicUlli, have you tried this for real? Any examples you can post?
If I understand you correctly, you measure 2 IRs, front & back, and store them separately in REW. Then add them to get FR_add

I should point out this doesn't work so well for LDC cos the 'Fig-8 part' goes to Omni at LF. The Shure papers have details

 

[MicUlli]

Yes i did. See the following pictures.
I used a crappy old dappolito passive speaker in a big room, measurement distance 1m.
The 1st picture shows the raw responses, 1/48 oct smoothing, no windowing, test object Audio Technica AT2031.
The 2nd picture shows AT2031 0° and 180° FR added, ECM8000 ref mic and AT2031 (0+180) divided by ECM8000. It is remarkable how continuous the result seems to be.
The 3rd picture shows the result compared to a 1st order highpass with a corner frequency of 200 Hz. Note how close it comes to theory
The FR is usable up to appr. 1 kHz, above you have to use the 0° measurement...

 

[MicUlli]

MicUlli, have you tried this for real? Any examples you can post?
If I understand you correctly, you measure 2 IRs, front & back, and store them separately in REW. Then add them to get FR_add

I should point out this doesn't work so well for LDC cos the 'Fig-8 part' goes to Omni at LF. The Shure papers have details

I examined the "Shure Paper" (Torio/Segota) in more detail. IMHO adding 0° and 180° FR should also work for cardioid LDCs with double diaphragm. You may also come to this conclusion because cos(0°) = 1 and cos(180°) = -1. Examination of formula (5) shows that you would end up in a constant summed FR independent of frequency. That is exactly omni behaviour.
Unfortunately i am not able to deliver the proof, i never measured LDCs, my profession is related to SDCs..
Probably kingkorg has more opportunities to do measurements..
BR MicUlli

 

[MicUlli]

Here is a document regarding mic measurements (part 1 is electronics) Part 2 will follow after vacation

 

[soliloqueen]

The devil is in the details. Neumann's QC "anechoic" chamber is rather tiny. I purposefully put the anechoic under quotes. They also utilize the floor.
Screenshot of How it's made - Neumann u87.

this is called a hemi-anechoic chamber and it's common in product QC, as there are very few tasks that you need an actual anechoic chamber for. there are also certain tests you can only do with the reflective ground

 

[jp8]

Here is a document regarding mic measurements (part 1 is electronics) Part 2 will follow after vacation

I have studied the microphone measurements document and would like to make some comments and additions to chapters 4 and 5 (Symmetry damping and Mic current consumption).

First, I would like to note that my personal preference would be to use the well-known notion of CMRR , rather than introducing a new notion of "Symmetry Damping". I don't see "Symmetry Damping" being used on the internet in this context. But OK, it's not my document, so you're free to use any terminology you like.

Secondly, it must be noted that CMRR is a system property, so your cable symmetry and especially the Audio Interface also determine the outcome of your measurement. CMRR of cables + Audio Interface should be an order of magnitude better than the expected CMRR of the DUT. It could even be that asymmetry in your DUT is compensated by reverse asymmetry in your Audio Interface and you end up measuring a much too positive value.

I will show you some examples of what CMRR you can expect from a cheap Behringer UMC202HD Audio Interface, a CMRR measurement of the mic input of my "DIY Audio Analyzer" built around an improved UMC202HD and a CMRR measurement of the Behringer B-5. These examples will make clear that the CMRR of your Audio Interface is something to keep in mind when measuring the CMRR of microphones.

The measurements have been done with a 280mV stimulus across the 100R ground resistor from MicUlli's setup. This 280mV corresponds to 0dBFs on the 200mV calibrated range on my Audio Analyzer and the UMC202HD gain and output level have been set accordingly. So you can read the CMRR value directly from the dBFS scale on all the plots. The CMRR measurements of the UMC202HD and my Audio Analyzer were done using a method described in the R&S UPL Service Manual: apply a signal between ground and positive and negative input terminal through two ~300 Ohm resistors, which match within 0.01% (30 mOhm). I managed to get them matched within 6 mOhm, so I assume my CMRR measurements are quite accurate.

The "CMRR Micinputs at 280mV" plot shows the CMRR of both mic channels of a stock UMC202HD. The glitch just below 20kHz is from the internal phantom power boost converter, a well-known flaw of this Audio Interface. The worst of the two channels has a CMRR of 48dB. The "CMRR Mic input 200mV setting" plot shows the CMRR of my DIY Audio Analyzer. Much better than the stock UMC202HD, even taking into account that CMRR starts to deteriorate at higher frequencies. The last plot shows the CMRR of a Behringer B-5 measured with MicUlli's setup. As can be seen, the CMRR of this microphone is already quite close to the CMRR of the UMC202HD. I haven't measured other mics this way yet, but I can imagine the better microphones will require a better Audio Interface than the UMC202HD to accurately measure CMRR. Unfortunately, CMRR values of Audio Interfaces are often not (or never?) specified, so you'll probably have to measure it yourself as described above with the 300 Ohm/0.01% resistors.

Finally, I would like to make an addition to Chapter 5, where MicUlli describes the measurement of the microphone current. If you do not have his CMRR test setup at hand, you can simply calculate the mic current from the DC voltage measured on pins 2 and 3. Assuming the standard 48.0V no load Phantom Power and 2 x 6k8 current limiting resistors, the mic current is simply calculated by: Imic = (48.0 - Vout)/3400. For many of you, I guess this method does not come as something new, but I'd just like to mention it.

Jan

 

[MicUlli]

Hi Jan,
thanks for your usefull comments on measuring CMRR
You are completely right: CMRR (or symmetry damping) is a system issue and not only mic related. Perhaps it is necessary to check the audio interface first and then compare it with the measurement of the mic under test. This belongs into the category "How well does your audio interface perform". I agree, never found comprehensive datasheet values regarding audio interface CMRR..
BR MicUlli

 

[MicUlli]

Last but not least:
Here comes part 2 of mic measurements
EDIT:
Part 2 is acoustic properties.
The measurement of (hyper-, super-, wide-) cardioids uses the relationship already mentioned in post #45.
Have fun
MicUlli

 

[OneRoomStudio]

As for the electrical testing of the headamp, I'm curious about where/how signal is being injected. As @Khron recently pointed out to me when I was simulating the feedback deemphasis circuit in a U87 circuit, the signal generator needs to be in series with the capsule/capacitor, rather than parallel to ground to see the full effects of the AC feedback on the backplate.

It seems like the same would be true when making these kinds of tests - shouldn't the signal generator go across the backplate/front diaphragm connections (with a capacitor in series that measures the same as the capsule it's replacing)? Like this:

 

[OneRoomStudio]

I'm curious about other's opinions about the setup above. Does it make sense to use this and have the signal in series with the capsule, or is it fine to just ground the signal generator and inject signal into the capsule connection?

 

[Khron]

I'm curious about other's opinions about the setup above. Does it make sense to use this and have the signal in series with the capsule, or is it fine to just ground the signal generator and inject signal into the capsule connection?

Depends on the circuit.

Some Neumanns have that "calibration input" (and mics that share those topologies); most other mics don't.

 

[OneRoomStudio]

Depends on the circuit.

Some Neumanns have that "calibration input" (and mics that share those topologies); most other mics don't.

Makes sense. As a tool to measure "any" mic though, it seems like this setup would yield more accurate results, right?

 

[MicUlli]

It seems like the same would be true when making these kinds of tests - shouldn't the signal generator go across the backplate/front diaphragm connections (with a capacitor in series that measures the same as the capsule it's replacing)? Like this:

This approach should work quite fine. VERY important is a low coupling capacitance between the transformer windings. AND the transformer connection without the cap in series should be connected to the low impedance part of the device under test.
BR MicUlli