Measuring microphone differences (somewhat scientifically?)

[lightningbefore]

I'm slowly trying to build these great DIY mics shared on this forum to see which I like most on my singing voice, which I understand is generally subjective to me.

While building some mics, I've realized it would be interesting to test the differences say before and after a capsule replacement or vs. a similar microphone to "see" what's different in addition to listening to it. This post from @kingkorg U67 circuit also got me thinking more about this.

So I found another newbie post online about how to even start measuring frequency response of a mic in general: How to Test Microphone Frequency Response | TSP

I was able to get through that doc successfully (I think) using my Sonarworks calibrated reference mic, Audacity, and REW. This was my first result with playing pink noise, with the mic pointed at the speaker 1 meter away, and 1/3 of smoothing.

I have a few questions:

• Did I do the measurement right?
• I'm not really even sure what I'm being told by the graph, most mic frequency graphs seem to hang more around 0db? I know I'm checking SPL here so it explains why it's up around 45db but I guess I thought pink noise through a reference mic would be more "flat"?
• Is this useful or is it more just showing the reflections and frequency buildup of my room?
• How do I setup this test and compare another mic in the same graph?
• Anything else that would be useful to change or check?

I'm also curious if it's worth it going down this path more or "to see" or if I should just stick with Kingkorg's suggestion in that linked post of simply putting the Sonarworks mic next to the mic(s) I want to compare and get the same take and just listen to them, and then do the same again with whatever component is swapped out.

Any feedback or suggestions would be appreciated, I tried searching the forum for similar posts and it seems like something that would be asked a fair amount but I'm still pretty confused.

 

[Bo Deadly]

Measuring mics is actually really hard to do. The biggest problem is resonance in the room. Bascially every room has reverb. So you're going to get reflections mostly in the bass region but you can get "standing waves" in other places depending on the size of the room and how parallel the walls are. In my bedroom I can clap and hear a rapid echo that is several pulses long. This is because the walls are perfectly parallel and they're plaster so they're very smooth and hard. Normally this sort of measurement would be done in an anechoic chamber like this:



The next best thing would be to take measurements in a field in the middle of no-where where there is no road noise and such.

 

[user 37518]

Measuring mics is actually really hard to do. The biggest problem is resonance in the room. Bascially every room has reverb. So you're going to get reflections mostly in the bass region but you can get "standing waves" in other places depending on the size of the room and how parallel the walls are. In my bedroom I can clap and hear a rapid echo that is several pulses long. This is because the walls are perfectly parallel and they're plaster so they're very smooth and hard. Normally this sort of measurement would be done in an anechoic chamber like this:

View attachment 104241

The next best thing would be to take measurements in a field in the middle of no-where where there is no road noise and such.

What Bo Deadly posted is exactly how is still being made, you need a very good anechoic chamber like the one displayed in the picture, you also need an omnidirectional source, which is some sort of polygon with speakers mounted on each of the faces. It is not something one would do in a regular studio tracking room.

 

[ccaudle]

What Bo Deadly posted is exactly how is still being made, you need a very good anechoic chamber like the one displayed in the picture, you also need an omnidirectional source, which is some sort of polygon with speakers mounted on each of the faces. It is not something one would do in a regular studio tracking room.

That is true for full frequency response measurements, but using time gating you can get reasonable mid to high frequency measurements. See for example the measurements which @kingkorg has been making. As he has pointed out the polar response measurements are suspect, but the on axis measurements above a few hundred Hz agree relatively well with more formal measurement results.

 

[kingkorg]

Measuring mics is actually really hard to do. The biggest problem is resonance in the room. Bascially every room has reverb. So you're going to get reflections mostly in the bass region but you can get "standing waves" in other places depending on the size of the room and how parallel the walls are. In my bedroom I can clap and hear a rapid echo that is several pulses long. This is because the walls are perfectly parallel and they're plaster so they're very smooth and hard. Normally this sort of measurement would be done in an anechoic chamber like this:

View attachment 104241

The next best thing would be to take measurements in a field in the middle of no-where where there is no road noise and such.

This could have been true for the period when b&w images were the thing.

Sorry for the sarcasm, had to do it. We have come a long way since then, and no you absolutely don't need anechoic room, or even quiet environment.

Even big names use anechoic for R&D stages mostly, QC, certification, calibration files are done in regular rooms.

Resonance is not a problem simply because it will be the same with dut and reference mic. And most microphones don't really vary that much in lows anyways.

Any noise will be discarded by the software simply because the software didn't generate it at specific frequency at specific moment in time. And by perticular sorcery the software knows when and where sound came from

 

[kingkorg]

I'm slowly trying to build these great DIY mics shared on this forum to see which I like most on my singing voice, which I understand is generally subjective to me.

While building some mics, I've realized it would be interesting to test the differences say before and after a capsule replacement or vs. a similar microphone to "see" what's different in addition to listening to it. This post from @kingkorg U67 circuit also got me thinking more about this.

So I found another newbie post online about how to even start measuring frequency response of a mic in general: How to Test Microphone Frequency Response | TSP

I was able to get through that doc successfully (I think) using my Sonarworks calibrated reference mic, Audacity, and REW. This was my first result with playing pink noise, with the mic pointed at the speaker 1 meter away, and 1/3 of smoothing.
View attachment 104226
I have a few questions:

• Did I do the measurement right?
• I'm not really even sure what I'm being told by the graph, most mic frequency graphs seem to hang more around 0db? I know I'm checking SPL here so it explains why it's up around 45db but I guess I thought pink noise through a reference mic would be more "flat"?
• Is this useful or is it more just showing the reflections and frequency buildup of my room?
• How do I setup this test and compare another mic in the same graph?
• Anything else that would be useful to change or check?

I'm also curious if it's worth it going down this path more or "to see" or if I should just stick with Kingkorg's suggestion in that linked post of simply putting the Sonarworks mic next to the mic(s) I want to compare and get the same take and just listen to them, and then do the same again with whatever component is swapped out.

Any feedback or suggestions would be appreciated, I tried searching the forum for similar posts and it seems like something that would be asked a fair amount but I'm still pretty confused.

Please ignore the method by the seasonal podcast shared. I have no idea how they came up with that.

Place reference mic on axis at 20-40cm away from speaker between tweeter and woofer. Do a sweep sine measurement in REW. Now do the same at exact same position with the mic you are interested in(dut). Sub milimeter accuracy required when it comes to positioning!

Once you get both curves simply in trace arithmetic section divide the dut curve by the reference mic curve. This will get rid of speaker and room anomalities to an extent. Use smooting up to 1/12 for reliable but prettier curves. There is much more to it, but enough for somewhat scientific requirements of your op.

EDIT.
Dont put up both mics at the same time, one will affect the others FR in close proximity. Measure one at a time.

 

[gyraf]

..I used to do this outside at night, on our roof - speaker pointing upwards, flying the microphone a couple of meters above, pointing down. For everything other than noise measurements (that we don't really care for, do we) this is fine

..and oh: so is audiomatica's clio. Keeps you from freezing during winter work, and is trustworthy down to some 500Hz at least, right there at your workbench

/Jakob E.

 

[JohnRoberts]

image of mic measurement rig

Back in the day such measurements were made outdoors.

JR

 

[Bo Deadly]

Resonance is not a problem simply because it will be the same with dut and reference mic. And most microphones don't really vary that much in lows anyways.

Any noise will be discarded by the software simply because the software didn't generate it at specific frequency at specific moment in time. And by perticular sorcery the software knows when and where sound came from

Yeah, I can see how that would work. That's definitely what I would look into if I got into testing mics. But if the room was really small without any acoustic treatments, I have to think that this technique would have trouble with rejection.

This is all possible for DIY now because you can get a decent reference mic for cheap. Last I checked there are electrets that are kinda awesome. I'm surprised they're not used more professionally actually. Is it because of SPL issues or because nobody want's to sell a mic that has a capsule that cost $3.59 USD even if it is some fancy mems thing?

 

[kingkorg]

Is it because of SPL issues or because nobody want's to sell a mic that has a capsule that cost $3.59 USD even if it is some fancy mems thing?

Probably noise. Small diaphragms usually have terrible noise specs.

 

[k brown]

With the pretty dramatic exception of the Primo EM273 and EM200; I still don't how they do it.

Those little guys have no business being as quiet as they are!

 

[ccaudle]

if the room was really small

You have to keep in mind the reflection time when setting the measurement gating time, which will directly set the lowest frequency you can accurately measure.
So for sure living in the US in the suburbs would be better than living in a tiny flat in the middle of London for this type of measurement. Probably having a day job at a school or church or performance venue where you could access a big auditorium in the off hours would be even better, but even reasonable size rooms can give you decent results in the midrange and up.

I'm surprised they're not used more professionally actually.

I think the reasons you don't see measurement capsules used for recording more often are probably due to:

• the calibration methods for measurement microphones usually require sealing to the mic, so it only works for omni, not directional patterns. Omni mics are useful for music recording, but don't tend to get used as often as cardioid.
• small capsules have worse signal to noise ratio, so sometimes are too noisy for music recording (obviously application dependent)
• if the capsule is not intrinsically flat response, but relies on a calibration file for measurement use, that may limit where it is useful for recording.

There are some places where having a low cost omni capsule available opens up some possibilities which would have been really expensive previously, such as spherical array microphones for surround sound recording.

I don't know what kind of capsules they use, but the Earthworks omni mics are kind of what you mention, omni capsule, but then they have a fancy machined housing, and they have to assemble and test it, so it still ends up costing $400, because a not particularly low noise omni microphone has a bit of a niche market.
Earthworks TC20 page
Note the 26 dB SPL(A) equivalent noise floor.

I have also heard other manufacturers say that if you buy the cheap capsules there are enough variations in performance both new and over time that you have to gently bake them for a little while to pre-age, then test them all and throw out some for not meeting tight enough specs.

 

[Bo Deadly]

• small capsules have worse signal to noise ratio, so sometimes are too noisy for music recording (obviously application dependent)
• if the capsule is not intrinsically flat response, but relies on a calibration file for measurement use, that may limit where it is useful for recording.

I'm not familiar with measurement mic capsules. What I was referring to are electret mics like you might find on Mouser for $5. Someone posted specs here one (might have been a fancy mems type) and the SNR was far better than what is required to make a mic that captured the lowest practical ambient noise. And my experience with electrets in general is that their frequency response is almost perfectly flat which cannot be said for even the best condenser capsules.

But maybe I'm wr-wr-wrong ...

 

[Chrisfromthepast]

Measurement mics are sometimes externally polarized, but the ones discussed so far in this thread are electrets.
I haven’t had the occasion to use a MEMs mic connected to a phantom powered mic pre. Maybe you’re onto something?

 

[k brown]

Small cardioid electrets with flat response are very rare; again the Primo EM200 is a remarkable exception.

Small omni electrets that are flat are more common.

When Earthworks started out they were modifying Panasonic capsules; since those were discontinued, I don't know what they're using.

 

[Bo Deadly]

I just tried and failed to find the thread about the capsule someone posted about. It was maybe a year or two ago. All I remember was that it was not a conventional electret like the Panasonic ones, it was bigger, had a really low self noise and SNR was remarkably high. It was a new part at the time. It was on Mouser and had a datasheet. If I had an infinite amount of time and resources, I would search mems mics on Mouser.

 

[ccaudle]

And my experience with electrets in general is that their frequency response is almost perfectly flat which cannot be said for even the best condenser capsules

That might be a secondary effect of the small size. If the capsule is otherwise designed well, smaller should have less frequency response deviation due to membrane size vs. audio wavelength issues. Small size has intrinsic SNR reduction compared to larger capsules, and non-intrinsic manufacturing difficulties (just meaning that as you get smaller it can be hard to keep tight tolerances in terms of a percentage of size in low cost manufacturing).
If you are satisfied with the SNR you can get at a particular size, then you can look out for one with the frequency response, both on axis and off, that you can live with.

Perhaps improvements in manufacturing capabilities (more precise and consistent backplane manufacture) are allowing better performance than before and some good devices have escaped DIY notice so far.
I keep an eye out in these forums for anything interesting getting mentioned, but I don't go out of my way to see what is new at the vendor and distributor sites, so maybe some capsule intended for voice activated smart devices or phones has stepped up the features lately.

 

[cyrano]

MEMS mics?

I've seen some ruler-flat curves on data sheets. Not always easy to get in small quantities tho...

 

[ccaudle]

MEMS mics?
I've seen some ruler-flat curves on data sheets.

Any pointers to something recent? I have not looked in a long time, the last time I did the devices were high noise, and limited frequency response. Stated and obvious use case was for cell phone microphones, and the claim to fame was around manufacturing process, not acoustic performance.
Quite probably I have missed some advancements in that area because I just stopped looking after the early devices did not seem to have any music recording application.

 

[cyrano]

The same goes for my search. A while ago and abandoned because of noise figures.

But recently, I came across some people building mic arrays and they were rather enthusiastic about recent ones. Especially sensitivity and noise. Of course, there's no mention about musicality.

Unfortunately, I can't remember the type numbers. I'll see what I can dig up...