What is the most important tool when working with speakers? Of course it’s a measurement microphone. Be it a restoration project or a whole new speaker system – without actually measuring the sound field, you are working in a blind. Also this measurement should be accurate and correspond to a reality. But how do we know that? And here comes the most important question: what is measuring the measurement microphones?
Above is my reference measuring microphone. It’s a DIY effort from 10 years (or so) ago. Nothing fancy, just inherently very linear WM-61A capsule, tube and XLR connector. WM61A is long discontinued now, but it still pops-out sometimes on eBay from reputable Japanese sources. Everything else you see labeled “WM61A” on listings these days is a Chinese junk.
I did modified capsule as proposed by late S. Linkwitz to make it work as source follower and it did lower distortion significantly, albeit only at higher SPL levels. Pre-amp is mounted directly in XLR connector and it’s a simplicity itself. Just a two op-amp gain stages. Probably one is more than enough, but I wanted to be able to regulate gain slightly. So putting a noisy potentiometer in a feedback loop of 1st stage doesn’t seemed to be a good idea, hence the two stages. Output is not balanced, as this is not needed for a couple meters of signal cable. Then 3 XLR pins are enough for ground, signal and power supply.
I share your skepticism. Internet is full of these DIY mics. Everybody and his dog has one. So what’s so special about it? And when does it become “reference” ? Well, it becomes “reference” when you can reference it to some standart.
Meet Brüel & Kjær Type 4189. It’s a 1/2 inch IEC 61672 class 1 microphone with Type 1706 pre-amp. Together they cost probably about the same as used minivan. I had an opportunity to beg&borrow this puppy for a day. And boy did my palms got sweaty when handling it. It came with it’s own suitcase and fresh calibration file.
I used it to measure Audax TW034XO tweeter flush mounted on a IEC 268-5 standard baffle with a size of 2.025m x 2.475m. And this became my “reference” measurement. Next, I made the same measurement with my WM-61A mic. The difference between two measurement’s is my new WM-61A calibration file.
At that time I didn’t knew it, and just thought that what I’m doing “makes sense”, but this is actually a well known substitution method for microphone calibration. It seems that even big boys like Earthworks are calibrating like that. There is a wonderful paper by them “How Earthworks Measures Microphones” with all the details and more insides. Really recommended reading.
As you can see, datasheet has a much more optimistic frequency response. But there is a good correlation with wrinkles at 4kHz and 7kHz. These do look suspiciously like diffraction artifacts from a capsule edges. Or maybe capsule cavity resonance. Can’t tell for sure. Mine are a little bit lower though. High frequency sensitivity increase is something that all (or at least the ones I saw) 1/4″ electret capsules exhibit on axis. So overly my calibration response has a perfect sense.
All these shenanigans were taking place somewhere around a year 2014, some 6 years ago. And I thought to my self – how well does my calibration holds ? I mean, it’s usual practice in metrology to calibrate measurement equipment at least once a year. Unfortunately I no longer have access to that sweet B&K mic, so it has to be something readily available.
I was choosing from Dayton EMM-6 and Mini-DSP UMIK-1. These two are the most obvious candidates in a “reasonably priced” factory calibrated mics category. I don’t like the idea of 3 meters of USB cable so EMM-6 won. I have read dreadful stories about “how bad” the factory cal files were, so I was quite anxious to see how mine turned out.
Not that bad after all! And it’s quite fresh also, marked as 10/2019. Of course, as always – the devil is in the details. And we can’t see any, because of the enormous 80dB graph range. Let’s download the calibration file and see whats going on there closer.
OK, so there are still lots of measuring artifacts in this frequency response. I mean, all these little wrinkles and peaks – they can’t be real. It’s basically impossible to make microphone capsule that bad. There can be numerous reasons for them. Read my holder influence page to get an idea on how sensitive these measurements are.
So after applying heavy 1/3 octave smoothing – we get something usable. Again, two distinct valleys @2 and 5kHz. That’s even lower then WM61A (12mm vs. 6mm mic outside diameter, but both 1/4″ capsules). Inherent high frequency on axis increase is here too. Seems reasonable. I don’t know about that 50Hz-1k range though. Is it real or just an artifact? Anyhow, I wouldn’t split hairs because of it. I mean, we are talking here about 0.6dB of absolute error! And if you really want to get in a ±0.5dB precision range and have guaranteed repeatability – it will cost you. How much? The usual answer is – more then you can afford if you ask this question.
So I repeated the same measurement of the same Audax TW034XO tweeter on the same baffle as 6 years ago with EMM-6 microphone this time. And above you can see a final result. It’s a difference from my reference WM61A mic and it’s within ±0.5dB. What was surprising is how consistent my WM61A mic capsule and TW034XO tweeter has stayed over these years. This tweeter is in my main system and is working daily. But when I nulled the two measurements, made with the same WM61A mic, just 6 years apart – it was basically a straight line. I don’t even see any point in posting it here.
So the takeaways would be:
As you have probably noticed by now, all calibration frequency responses I made here starts at 1kHz. And there are couple reasons for that. First of all, my reference measurement was done with a dome tweeter. And although it’s a 1″ dome, I wouldn’t trust frequency response below 1kHz to be very consistent. Next, a good 1/4 inch capsule should already be very linear from 1kHz down to it’s fc. To test this assumption, I have build a pressure chamber.
I will only show here the final result, which is a full frequency calibration curves of both mics.
EMM-6 is actually better in a low-end region then factory calibration would suggest. I measured it’s -3dB fc to be at 14Hz instead of factory suggested 20Hz. Not a big deal really.
Usually microphone sensitivity is measured in mV/Pa@1kHz. That is RMS voltage in milivolts that microphone is producing when subjected to 94dB of sound pressure at the frequency of 1kHz. Why 94dB? Because remember that decibels is just a ratio to some kind of the reference. When talking about Sound Pressure Levels we assume the reference of 20μPa. That is considerate to be the absolute threshold of human hearing. So 1Pa to 20μPa has a ratio of 1:5000. What else has the same ratio? Right – 94dB.
For the most part I was using an SPL meter to level calibrate my whole measuring chain. This is a ±1.5dB device, but I think it’s more then enough. That’s because we are rarely interested in absolute measurements when working with audio. What we are looking is relative differences. My pressure chamber experiments provided opportunity to measure my microphones sensitivity more accurately.
|Dayton EMM-6||Panasonic WM61A||UNI-T UT353 SPL Meter|
|7.35 mV/Pa@1kHz into 1kΩ||57.2 mV/Pa@1kHz into 1kΩ||94.8db (+0.8dB)|
So it turns out that this cheap UNI-T SPL meter is not far from the truth. But EMM-6 sensitivity seems to be much less then 10mV/Pa as stated in a calibration sheet. I decided to investigate it further.
I also have made a 1kHz level calibrator that I published in a DIY section. With this last addition I finally have my “acoustical measurement kit” complete.
I guess I will still use trusty WM61A as my main mic, but EMM-6 will still be useful as a backup. When things start to measure funny it’s always good to have a second mic for quick evaluation. For the next upgrade stage I plan to make a portable USB audio interface geared towards measurements, so stay tuned.