how to find frequency response of microphone

hello I want to find the frequency response of a microphone. I give the input signal to my speaker and it produces a specific SPL with specific frequency. on the other side, I read the microphone output voltage. but I wonder which value of SPL is required to find the frequency response of microphone? should I fix the SPL and just change the frequency and read the microphone output? if so, which SPL value I must choose? should I fix the SPL to 94 dB SPL?

you can assume that I have a really precise reference microphone and I can find out the SPL of speaker by means of this reference microphone. I also don't really care about direction and distance because I use the reference mic so the only thing that matters to me is SPL. also assume that I afford very precise speaker too and don't care about it's response. I appreciate your help

thanks

• If you have a reference microphone then you can do it with substitution method, Did it myself many times. Preferably, you want to have a class-1 reference microphone and a way to record the signal off it via audio interface. First, place the ref mic close to the speaker and play the sweep (I suggest REW for that). Then remove the reference mic and in the exact same position place the unknown microphone. Measure with sweep again. Divide the two responses and you will end up with the response of your microphone under test. You can use any playback level. 94 is good for sensitivity at 1k.
– jojek
Feb 19 at 13:13
• Also, I think you might be confused. Sensitivity is measured for 1k sine wave @ 94 dB SPL (1 Pascal). So for analog microphones, this might give you values such as 12.6 mV/Pa, basically what's the voltage RMS when you expose the microphone to 1 Pascal of acoustic pressure. Alternatively, sensitivity can be represented in decibels relative to 1 volt (dBV). So the above example would become -38 dBV of microphone sensitivity. Now, the important thing is that you don't have to measure the sensitivity at exactly 94 dB SPL, in fact you can use other levels and simply scale the result.
– jojek
Feb 19 at 13:58
• So for example, if you measured the same microphone with 114 dB SPL, 1 kHz sine wave you would get 125.9 mV = -18 dBV. Then, in order to find the sensitivity, just subtract the difference between 114 and 94 (20), which will give you -38 dBV (12.6 mV/Pa). So, as long as you know the exact level of your excitation signal, you can easily calculate the microphone sensitivity. As for frequency response, it can be measured at any level. Some will graph it so that 1kHz is exactly at 0 dB, whereas others will shift it so that 1kHz is exactly at the dBV level equal to senstivity. It's all the same.
– jojek
Feb 19 at 14:04
• Yes. As long as it's not too loud (causes clipping/distortion) or too quiet (drowns in the self-noise) then you will be fine. Usually, when I am working with cheaper speakers, I would play the sweep at 74 or 84 dB SPL and scale the result. No need to drive it so loud. And like I said before, make sure that you know the level for 1kHz , other frequencies don't matter - you will get offsets from the frequency response. Even speaker doesn't need to be perfectly flat - you gonna remove its effect when subtracting the measurement of the reference microphone.
– jojek
Feb 19 at 14:18
• I certainly would NOT consider an ADMP401 as a reference microphone for anything but that really depends on the requirements for your application. Feb 19 at 16:03

This is unfortunately a difficult problem.

1. A microphone doesn't have a single frequency response. It has a different one for each direction of incidence. Especially for second order microphones (dipoles, cardioids, etc.), the frequency response also depends heavily on the distance and radiation impedance of the source. So first you need to decide: what distance and do you want on-axis, frontal averaged or diffuse field response
2. Measuring the microphone directly requires an EXTREMELY well behaved loudspeaker. Most microphones are very linear and quite flat, but most speakers are definitely not. A measurement grade speaker is hard to come by and very expensive.
3. If you really just want the microphone response you will also need an anechoic chamber that covers your frequency range of interest. Otherwise the response of your acoustic environment (room, walls, furniture, etc.) will dominates the microphone response. Buying a a few of those will seriously strain your relationship with the CFO of even a mid-sized high tech company. Don't ask me how I know :-)

Your best shot is to find a good reference microphone that has the same directionality type of the microphone you want to measure. Than excite a loudspeaker with pink noise and measure the power spectrum for both the reference mic and the your test mic at EXACT the same location and orientation. Average over many positions, rooms and a few different speakers.

The frequency response of your microphone is the difference of the mic's power spectrum to that of the reference microphone.

• you can assume that I have a really precise reference microphone and I can find out the SPL of speaker by means of this reference microphone. I also don't really care about direction and distance because I use the reference mic so the only thing that matters to me is SPL. also assume that I afford very precise speaker too and don't care about it's response. I appreciate your help Feb 19 at 13:01
• I actually have another question. when a company releases a frequency response of a microphone, they give us a diagram that it shows the sensitivity in different frequency. I wonder which input SPL they use in all frequency ? do they change the SPL and frequency concurrently ? or the just fix SPL at 94dB? Feb 19 at 13:20
• You can use any SPL as long as your signal to noise ratio is good and you are not overdriving anything. The actual measurement on the production line depends on the brand and type of microphones. High quality measurement microphones are typically measured using a coupler. More "normal" microphones are measured in a test both against a known reference. Feb 19 at 16:06