DIY microphone calibration

[Randolf]

The Minidsp UMIK-1 is a good and easy to use USB measurement microphone. However I needed to make time aligned measurements of individual drivers and export them into VituixCAD for crossover simulation. The recommended procedure is to perform time aligned measurements in REW via loopback interface, which is hard to achieve with an USB mic. Therefore, I ordered a Focusrite Scarlett 2i2 3 gen sound interface and a Behringer ECM8000 microphone. The ECM8000 comes with no calibrations so I needed to create them. You can send your microphone to a calibration service. I did so some time ago for my UMIK-1. Hifi Selbstbau in Germany offered a 2 orientation calibration for 40 Euro. The result in comparison to the Minidsp calibration was:

Upper graphs show the 0 degree calibrations (with 1db offset to show the differences) and lower graphs show the 90 degree calibrations. So the Minidsp calibration was already reasonable good in my case.

Every microphone calibration requires a calibrated reference microphone, once you have one you can calibrate microphones by your own. This is the procedure I used:

1. You need a reasonable loudspeaker (which at least is able to reproduce the frequency range) in my case a KEF Q 350.
2. Position the speaker reference axis (usually tweeter) at half room high.
3. Use REW to perform an on axis measurement at 1m distance and time gate it at 4ms to eliminate room reflections. Perform this measurement with calibrated reference mic (using its calibration file) and uncalibrated mic (without calibration file) for 0 and 90 degrees mic orientation. Adjust the SPL e.g. at 1khz so that both mics measure equally loud. Try to position both mics as similar as possible.
4. Perform a near field on axis measurement of the woofer for 0 degrees mic orientation with calibrated reference mic and uncalibrated mic. Again, try to position both mics as similar as possible.
5. Use REW trace arithmetic “A / B” (A=uncalibrated mic measurement, B= calibrated mic measurement) to calculate calibrations for all 3 measurement pairs.
6. Use REW trace arithmetic “Merge B to A” to merge near field calibration into the the 2 other calibrations.
7. Finally apply Psychoacoustic smoothing to the 2 calibrations.
8. If everything looks reasonable use “Export measurement as text” to export both calibrations. Select “Use custom resolution: 96 PPO” and “Do not include phase in the export”. You now can use these exported files as mic calibration files.

I have done so once with my UMIK-1 to compare my calibration with the original one and with the EMC8000:

The upper 2 graphs show the original 0 degree and my generated calibration (with 1db offset to show the differences). The middle 2 graphs show the same for 90 degree orientation. The lower 2 graphs show the generated calibrations for the ECM8000 (with 10db offset).

Since the described procedure relies on rather stable DIY near field and 1m time gated measurements only is seems to be robust enough to generate reliable enough microphone calibration data.

Final measurements of the Q 350 with UMIK-1 and ECM8000 now show practical identical results. Furthermore, with the Focusrite Scarlett 2i2 + ECM8000 a loopback time aligned measurement of the complete speaker, separate woofer and tweeter now shows that summation of woofer and tweeter (Trace arithmetic A +B) perfectly matches the complete speaker measurement. So the result can be exported into VituixCAD for proper crossover simulations.

 

[staticV3]

Nice guide!

Finally apply Psychoacoustic smoothing to the 2 calibrations.

Why would one apply psychoacoustic smoothing to such calibration measurements?

No psychoacoustics are at play here.

 

[DVDdoug]

However I needed to make time aligned measurements of individual drivers

I've never used REW but the UMIK-1 should be OK.

There is ALWAYS latency (delay) through the computer but it's the same for all frequencies and all of the drivers. It's the timing differences that can cause phase and other issues.

And you don't need a calibrated mic to measure time differences.

Yes, you might get lower latency with the Focusrite, especially if REW supports ASIO drivers.

 

[Randolf]

Nice guide!


Why would one apply psychoacoustic smoothing to such calibration measurements?

No psychoacoustics are at play here.

Thanks, no particular reason for that, I just did the smoothing with a sense for proportion. In particular there was a little bit of ripple at very low frequencies from the 2 near field measurements I liked to remove.

 

[Speedskater]

an old Earthworks paper:

How Earthworks Measures Microphones
By Alex Khenkin, Director of Engineering, Earthworks
How to measure microphones and the implications relating to measuring loudspeakers

https://recordinghacks.com/pdf/earthworks/how-earthworks-measures-mics.pdf

 

[Curvature]

It's also possible to calibrate using a pressure chamber: https://www.mvaudiolabs.com/diy/microphone-calibration-pressure-chamber/

I would probably use the OP's method, but I would use a frequency dependent window, close distance to the speaker, and REW's EQ function for the traces.

 

[Randolf]

I've never used REW but the UMIK-1 should be OK.

There is ALWAYS latency (delay) through the computer but it's the same for all frequencies and all of the drivers. It's the timing differences that can cause phase and other issues.

And you don't need a calibrated mic to measure time differences.

Yes, you might get lower latency with the Focusrite, especially if REW supports ASIO drivers.

My initial setup was DDRC-24 + UMIK-1. With this I could not get a properly time aligned separate measurements of woofer and tweeter which would sum correctly. To be more precise I once succeeded half a year ago by using some alignment functions in REW (cannot remember how I exactly did it that time), anyway I could not get it to work again. Since the REW, CLIO, ARTA and SoundEasy guides (linked in the VituxCAD documentation https://kimmosaunisto.net/) explictly mention to use a 2 channel measurement with loopback time alignment and avoiding USB mics I finally decided to go that way, which worked out of the box. Maybe there still is some way to get DDRC-24 + UMIK-1 + REW separate driver measurements correctly time aligned. If so it would be good to know how. In general REW supports "No timing reference", "Use loopback as timing reference" and "Use acoustic timing reference" measurments and "Time align " and "Align IR start" manipulations.

The overall goal is not to just measure time difference, my goal is to measure frequency response of individual drivers both a) correctly time aligned and b) with correct SPL, since both is required for crossover simulation. For b) of course we need a proerly calibrated mic, which is this thread about.

 

[asteroth]

Complete noob here, I want to buy a Behringer ECM8000(+ the cable) or a Superlux ECM999 or anything else, on the cheap side that you suggest.
And borrow a SoundID Reference Measurement Microphone to perform the calibration of the mic that I will buy.

I have this soundcard Behringer UMC204HD: https://www.behringer.com/product.html?modelCode=0805-AAS

Do you have any suggestions or do you see any problems with my plan?

 

[FERCSA]

Nope just follow OP's guide.

I also bought the ECM8000 and created a calibration file in REW using my umik1 as reference.

So just go ahead and have some fun. I use this method since 2024 when I bought my own studio interface.

 

[sergeauckland]

I calibrated my ECM8000 against an Earthworks mic but using the 31 third-octave frequencies and old-fashioned level measurements rather than REW. Maybe I should redo it using REW if I can again get hold of a calibrated microphone.

S.

 

[Randolf]

Complete noob here, I want to buy a Behringer ECM8000(+ the cable) or a Superlux ECM999 or anything else, on the cheap side that you suggest.
And borrow a SoundID Reference Measurement Microphone to perform the calibration of the mic that I will buy.

I have this soundcard Behringer UMC204HD: https://www.behringer.com/product.html?modelCode=0805-AAS

Do you have any suggestions or do you see any problems with my plan?

Looks good to me, Merry Christmas.