We did some tests with a colleague who works with Audiomoths. I work with Song Meters, so we thought we could do useful comparisons.
We made a weatherproof case for the Audiomoth which consisted of a plastic box (Snack Box) of 9.5 x 5 x 2.5 cm, covered inside with synthetic foam for padding (see photos in Comparing 2 Audiomoths with 1 Song Meter). We drilled a hole in the plastic box on top of the microphone and covered it with a Gore GAW112 acoustic vent, which is used on almost all of the Wildlife acoustics microphones because it impedes ultrasound transmission only little. We attached another layer of foam to the lid of the box to prevent the PCB touching the lid and added silicone to the microphone corner of the Audiomoth to prevent contact of the PCB with the vent to reduce vibration transmission.
We compared the sensitivity for ultrasound of the bare Audiomoth to the sensitivity of the Knowles SPU0410LR5H-QB microphone attached to the Song Meter SM2Bat+ (this is a custom microphone). Our Knowles microphone was inside a tubular microphone housing, and covered with a GAW112 acoustic vent and a wind screen from Wildlife acoustics. We programmed six Audiomoths to record continuously at a sample rate of 384 kHz, with a high gain, and we recorded at 192 kHz with the Song Meter. We emitted sounds of different frequencies with a portable loudspeaker (1, 2, and 6 kHz, SoundCore Anker) and the Wildlife Acoustics Ultrasound Calibrator which emits chirps at 40 kHz. Sounds were emitted at distances of 2, 4, 8, 16 and 32 m in front of the devices. We equalised the loudness of the SM recordings with the Audiomoth using the nearest unclipped recording of ultrasound by amplifying it.
Comparison with SM2Bat+
We found that all our new Audiomoths record sound very similarly. The recordings were synchronized without gaps, and showed similar spectrograms. The sensitivity of the microphones and the ambient noise levels too were visually identical. This is good news, it shows the device quality is very consistent.
The self-noise level of the bare Audiomoths was slightly lower than the self-noise of our packaged Knowles microphones: the 40 kHz chirps at 32 m to the front were barely visible on the SM recordings and clearly visible for the Audiomoth, but it is a slightly unfair comparison since the Knowles microphones we used had been used in the field, and were covered with a vent and windscreen. So: to be interpreted with caution.
Audiomoths record several undesirable harmonic artifacts when sound level is high, but well before clipping. This is of relevance to bat researchers because fidelity of bat call recording will be low. We suggest that the other Knowles microphone we used might be better suited.
Comparing different protection levels
We repeated this test to compare the sensitivity of 2 bare Audiomoths (when used uncovered), 2 inside the plastic bag provided with the product, and 2 inside the case we built for the devices, at different audible and ultrasonic frequencies. We also tested the directivity of sound reception, by emitting sound in front and perpendicularly to the microphones. Unfortunately, the measurements for the bag setup are less precise as one unit did not record, so we had only one recording for that setup.
Immediately apparent: For ultrasound (40 kHz) at 8 m, signals from bare Audiomoths were saturated, those inside cases were not, while signals from Audiomoths inside the plastic bags were not saturated already at 4 m.
More in detail: We measured the mean sound amplitude for each unit (bare/plastic bag/ case) using ten identical, not saturated 40 kHz chirps. For audible sound, we used two 0.8 s selections for each frequency and device. We used the “Plot spectrum” tool in Audacity and averaged the measurements.
For ultrasound (40 kHz): The plastic bag attenuated sound on average 22.15 dB compared to bare units. Our case attenuated sound at 40 kHz on average by 13.95 dB compared to bare units and were thus distinctly better than the plastic bags, which was not unexpected, but noticeably worse than the bare units.
For audible sound: at 1 kHz, the bag did not attenuate sound noticeably, cases by 7 dB. At 2 kHz, the bag did not attenuate sound noticeably, cases attenuated sound by 5 dB. At 6 kHz, the bag attenuated sound by 16 dB, and the cases did not attenuate sound noticeably.
This is the price we pay for weatherproofing with our simple design, and it leads to a reduction in detection range.
At 40 kHz, with bare units, chirps at 32 m could all be detected (weak); with cases, very few chirps could be detected at 32 m and they could be easily detected at 16 m; within the bag, the signals were weak at 16 m and none could be detected at 32 m.
(Plotting sound transmission curves would allow to interpolate the exact extinction distance for each design.)
Audio signal shapes were visually looking relatively unaffected by the case, which is good news. The transmission loss caused by the case could probably be reduced if the gap between the case and the PCB was filled by a tube or better, with a horn.
UPDATE: For more info about directionality, see this thread.
We compared the amplitude of 4 chirps perpendicular to the Audiomoths against 4 chirps in front of them, for each unit, and took the average reduction in sound level to express directionality.
Sound pickup at 40 kHz is quite directional. With bags, there is a 8 dB reduction from the front to the side (90 degrees). The bare units are actually attenuated more (12 dB reduction from front to side), maybe the plastic bags' uneven, slightly protruding surface could help in picking up sound from the side but there was only one unit and variability is high so I prefer not to speculate too much. The cases led to higher directionality (19 dB reduction).
Sound pickup in the audible range was only checked visually with spectrograms, not measured (I can provide files). In theory, it should be less directional than in the ultrasound range anyway, and indeed, we did not see a difference between the bare units and those in bags or cases, but signals from the front are always visibly stronger than those from a perpendicular source, which is also to be expected.
In order to obtain stereo recordings, more than one device facing opposed directions should be used (this is shown in the "Comparing 2 Audiomoths with 1 Song Meter" thread).
To conclude, due to the high reduction in sound cased by the plastic bags, proper acoustic vents are highly recommended when you record bats, and due to the narrow sound pickup should be borne in mind. For recording birds, if sound quality is not critical, the plastic bags seem to do the job.
I just updated the post with new measurements for the audible frequency range and directionality and for more clarity.
That's all very interesting. I am not sure that you need to use them on high gain, at least in the UK - it might be necessary to turn the level up a tad if looking for very high frequency bats, but the full noise spectrum is detectable on low gain so you shouldn't be missing any detectable signals at that. Since the dynamic range is only 12-bit you need all the headroom you can get or, as you noticed, you get distortion and aliasing. For identification with a Mk 1 eyeball this is not a too serious problem, but probably messes up auto-identifiers, and makes it difficult to describe signals accurately.
Hi Kevin, Thanks for that. It looks really useful for anyone using AudioMoth to record bats. Alex