For a while now, I have been trying to find specifications for the amplification gain of AudioMoth.
As for the SM4 for example, one can find the exact dB declarations. If anyone can point me to that for the different AudioMoth configurations, that would be great.
These specifications would help by estimating sound pressure level for example.
Estimating the end-to-end sensitivity of a microphone and recording system can be done (A) by knowing each component's gain and frequency response or (B) "empirically" end-to-end. Note that Audiomoth configuration allows 5 gain settings. (A) is hard because I don't know the pre-amp gain and I'm not sure it's published anywhere (also note that the hardware components and the way the gain is implemented in hardware have changed over AudioMoth versions). I'm sure Alex Rogers from Open Acoustics Devices would be able to answer for a specific version and setting.
On the other hand, (B) is easy. I have measured the end-to-end sensitivity (i.e., calibrating digital dBFS to dB SPL) empirically. I found the offset from the real-world dBA of a pink noise signal to the A-weighted digital signal's DBFS to be the following for each of 5 gain settings: {0: -99.5, 1: -95.4, 2: -89.2, 3: -84.6, 4: -82.2} Don't take these measurements as exact though: these measurements did not use anechoic chambers or professional standardized coupling (pistonphone) and may not be very accurate. Also, ARUs decrease in sensitivity over a multi-year lifespan (https://pdfs.semanticscholar.org/8d23/6f5f3db0bc0f5a94b1123c806562da5ff5b2.pdf).
My recommendation is to empirically measure a few of your recorders' sensitivity: play pink noise at a level that does not clip the recording; measure dBA at the recorder with an SPL meter; apply A weighting to the digital signal then calculate dBFS; calculate the difference between dBFS and dBA.
I'll note that the Audiomoth (and any ARU) is not equally sensitive across frequencies, or in all directions, and that these two factors interact. This is especially true when adding a protective housing (so...always?). We've published a paper and online resource with additional frequency response measurements for axes of rotation, and for various housings: https://www.mdpi.com/1424-8220/23/11/5254 https://github.com/kitzeslab/audiomoth-performance
Hope this is helpful!
Tessa Rinehart writes in her AudioMoth guide https://github.com/rhine3/audiomoth-guide/blob/master/guide.md#create-configuration
The gain is the amount that sounds from the microphone will be amplified once recorded. Selecting the optimal gain requires trial and error in your particular field conditions. If the gain is too high, your recordings will clip, creating an unpleasant distortion that can be challenging, if not impossible, to analyze. Alternatively, if the gain is too low, sounds will be faint and hard to hear.
(emphasis added)
My own conclusion: AudioMoth invites you for experimentation and learning by making errors.
Edit: Concerning the gain settings https://www.openacousticdevices.info/support/main/comment/60800bbcbb7c0c001500dc98 gives the following gain values for 1.1.0: [4.33, 7, 15, 25.1, 30].
These values are consitent with Sam's measurement (excluding top two values, where saturation/nonlinearity may play a role).
Edit2: After checking with the EFM32WG reference manual (the processor used in AudioMoth) It turns ot that the gain values should be linear and would result to the followng dB values [12.7, 16.9, 23.5, 28.0, 29,5]. Subtracting these dB values from the observed values of [{-99.5, -95.4, -89.2, -84.6, -82.2] on obtains [-112.2, -112.3, -112.7, -112.6, -111.7], which is at least roughly constant pretty much -18dBV/Pa (Mic sensitivity) minus 94 dB.
