The kind of data I would like to see in a microphone is a minimum of something like this:
Reference:
iSEMcon EMX-7150 review from
Production Partner.de, 06/2012 Issue
Here we can see the 0.01% mark is 20dB down from 1% and 30dB down from 3%
Reference:
iSEMcon EMX-7150 review from
Production Partner.de, 06/2012 Issue
Here we can see the 0.01% mark is 20dB down from 1% and 30dB down from 3%
Last edited:
Even with very expensive microphones you don't always get this data.
In fact, with most of these very expensive microphones you also just get a "calibration"* file of one-axis at a certain temperature and humidity.
* Technically this is not a calibration even, but a verification that can be used as a simple compensation.
If we really get serious and pedantic, calibration involves a whole lot more.
I have only never seen that for audio, except for very state of the art, serious and expensive physics research centers.
A calibration process will costs you at least 700 bucks that way.
My colleagues and I have been investigating THD in microphones decade ago. I carried out theoretical modelling of THD vs SPL based on the your mentioned B&K document and my colleagues measured mics THD experimentally. The match was surprisingly accurate.
https://forum.vegalab.ru/showthread.php?t=66583&p=1945801&viewfull=1#post1945801
X-axis is SPL in dB; Y-axis is harmonics in "%", dots are experimentally measured values of THD in Panasonic WM-61.
NO - it's 0,1% 20dB down from 1%! Cause - 2nd order function ...
3rd harmonic goes down faster/rises faster. But condenser microphone capsules are dominated by their first harmonic.
There is a special 1/2" microphones from B&K which has a few 100mV sensitivity and has very low self noise (-2dBA is only theory), I used these for noise measurements for a project. But really expensive, don't own one.
A big membrane cardioid condenser ... a microphone which is not linear in frequency response on axis, extremely unlinear off axis AND doule unlinear at low frequencies cause of the proximity effect? You can't EQ that.
We do a noise measurement here - you need an average of all directions (diffuse field corrected) to do so. It is simly not possible to do this with a cradioid microphone - your measurement isn't compareable to any other measurement or measurement rule.
Don't use that microphone for a measurement ...
USB powered, you need a noiseless PC/Laptop and/or put it in a different room. That's what I do.
Just add averaging!
Microphones have a lot of low frequency noise - A-weighting helps to mask that ;-) but of course makes sense for such low noise signals.
They are very robust, no worries. The main source of many measurement mics noise is often not the input fet but the capsule itself! This is different with a big capsule like 1" studio microphone or very noisy electronics like these cheap measurement mics.
As soon as you have some regular trafic somewhere near your house you are done with that. Windows are very bad at low frequencies and trafic produces a lot of that. Our hearing threashold is pretty high for these frequencies - but even A weighting will not rescue that situation.
As soon as you switch on any device with a fan ... computer, audio analyser (these are silly loud!), power supply or multimeter ... you are done. When you are sitting in the same room and just move a little.
It is actually easy to HEAR 10dBSpl - our ear can do that with no problem. And as you say, in a silent night you can hear amazingly far and finest sources.
But as soon as you want to do functional measurements on a regular basis with equiment and humans involved ... <10dBSpl(A) is becomming pretty hard to do.
I did a lot of measurement setups with microphones - you pretty quick get measurement errors and influences between the mics at high frequencies!
Putting one mic next to each other gives you an offset AND influences between the microphones.
I built a laser aided positioining system (sounds great, or? It's just 2 laser diodes aiming at a point in 0,5m distance 🤓) which helps to find the same spot with different microphones (about +- 1-2mm) - then you can do a correlation measurement but are still time variant.
How much is the low frequency fall of of your mic at 1m distance? As the noise rises at low frequencies you get significant less dynamic range <100Hz!
You could try to catch a used AKG C480 with omni capsule, these go sometimes for "cheap". They have pretty low self noise (spec at 11dBSpl(A) but normally 1-2dB higher) and are at least omni. With some HF EQ these would be useable measurement microphones. (I worked for AKG many years ago, still have a few of these gems).
This was probably to sloppy from my side (sorry, not a native speaker).
THD of condenser capsules is dominated from H2.
H3 and higher only add at very high levels. This works as long as the impedance converter electronics is linear - very often the limit for the complete microphone in high SPL.
As in the graph above, when H2 reaches 1%/-40dB - H3 only has -70dB. Higher order ones would be even lower.
So when you have a dedicated THD value for the CAPSULE like 0,5% at 140dBSPL or 1% at 135dBSPL like above you can calculate from that. But when you have something like "135dBSPL maximum SPL" its very likely the overload point of the amplifier and doesn't tell you anything about the microphone capsule.
At lower levels like 94dBSpl THD of a useful measurement microphone is burried in the noisefloor, that's for sure.
Based on this, to be able to measure down to, say, -80dB H2, the maximum the WM61's can observe is 94dB, and the B&K 4135 is 110dB.
But to accurately measure the distortion of bass below the Schroeder frequency, you either need
1) Anechoic chamber $$$$
2) Klippel NFS $$$
3) Near field, but then your microphone needs to be to observe very high SPL. $$*
4) Ground plane $... but elusive (very large, very quiet outdoor places, but with access to AC power may be hard to find)
*I can see why @Hörnli uses a multiple microphone solution.
An omnidirectional dynamic microphone may be able to handle 160dB without costing 4 figures that a condenser would cost.
Last edited:
D
Deleted member 375592
I fully agree with all of your very professional notes, and I learned from them. Thank you!
I had bad luck with noiseless PC / laptops. Those that inflict lass USB 1kHz noise were too weak to run MatLab. I gave up and chose an interface with a separate power supply.
you can rent those for just $$
Also that list reads a bit like you assume it's all or nothing.
Which is obviously not the case.
You can get the majority of measurements with a much more affordable solution.
Things like (internal) resonances etc, can even be measured with the cheapest microphones out there.
Measuring high SPL is only interesting if you also really want to know the exact maxSPL of a system.
Which is maybe great for just knowing, but in my opinion doesn't hold a lot of practical value.
The same goes for knowing the last significant digits in distortion at those levels.
That's exactly where you CAN'T perform a precise bass measurement ;-) (at least not without corrections and some tricks)
You can build an anechoic chamber WAY cheaper as a Klippel NFS - but you need a small, dedicated room for it.
About 20dB higher as 1m measurement. That's really no problem for a measurement microphone!
And you don't measure 0,001% at low frequencies - we are very happy to stay <1% at normal listening levels. Will be way higher at high levels.
So nearfiled measurements are easy with a mic like M30 or M215.
Reflections need to be far enough that their energy level is low enough to not disturb measured results in your margin of error.
So if you have your walls in 10-20m distance you would probably be fine.
Why would you need 160dBSpl? And how much would the dynamic mic distort at that level?
Get something in the ballpark of an M30 or M215 and be happy. Both cost less as a good Beryllium 3" midrange ... 😎
problem is that they are all over the place, so we come back to the consistency part again.
Very outdated and most manufactures seem to just copy-cat that.
Many years ago, I made my own microphone (the top one)
Also improved the bafflestep a lot. (which seems to be the issue around 10-15kHz with most mics)
Most electret capsule contain a FET as well, putting that one in just a source follower already improved the distortion A LOT.
(this idea came from Linkwitz, that article is somewhere on his website)
I bet using a constant current supply instead of a resistor (which is normally the way of doing things), would also be a good way forward.
I hear what you're saying about the M30 and M215. And I agree, if you imply that H2 or H3 at low frequencies doesn't need to be 0.001%.
But my data shows that a 1cm measurements is MUCH more than 20dB than a 1m measurement.
Here is 31.6cm vs 2cm.
I put a marker at 500Hz, and the difference there is some 16dB.
If I pull my mic back to 1m, it will be at least 16+10=26dB between 2cm and 1m (31.6cm to 1m)
I wonder what is causing our discrepancy between 1cm and 1m?
Last edited: