This is really impressive Avi! Would you mind sharing what kind of hardware the local ingestion + detection is running on? 

It's a fun experiment to think if this could be expanded to any live stream happening on any platform (Instagram, TikTok)

Hi Tessa, thanks so much for the update!!! I love that there is an interactive website now. This is such a valuable website, I'm happy to see the updates ;) - Liz

Actually, on the subject of acoustics, the Raspberry Pi based sound localization system I developed has been running continuously since 2023 writing to a 256GB SD card :-)
 

https://github.com/hcfman/sbts-aru

I submitted it for addition to that list a few years ago. Should be there also I guess.

I have three of these running around my house. Off power though because I can. Actually I use one of them as a time server for all my computer equipment because it maintains microsecond time accuracy continuously.

This is a great resource!

Hi Barbara. 

If you are able to power the Audiomoth externally from a lithium-ion source, the lithium-ion batteries can be used down to -20 deg C. They can't be charged below 0 deg C though so if you were thinking of a solar charging setup as well, then you'd need to look at the temperature range you're planning to use them in. 

In regards to the battery life for continuous recording, I found this information from here:
 

The recording lifespan of the AudioMoth on one set of batteries has previously been measured only for a subset of the possible configuration settings. Hill et al. reported the battery life of the AudioMoth using 3000 mAh lithium batteries for some common configurations [15], reporting that an AudioMoth could record for 115 days recording at 8 kHz, the lowest sample rate, for 30 s every 5 min. The developers also reported the AudioMoth lasted 9 days recording nonstop at a 48 kHz sample rate. While the AudioMoth configuration app provides estimates of battery life for any chosen configuration settings, these estimates have not to our knowledge been validated empirically.

Even derating the battery life by a factor of 2 due to cold temperature conditions would seem to get you over the finish line. One experiment you can try is to record continuously in a refrigerator and look at the recording duration. Typical refrigerators are around 2 deg C which could approximate conditions you mentioned. 

If you will be using stock Audiomoths with no modification for rechargeable batteries, then I'd recommend using Energizer Lithium AA batteries which have some of the highest battery capacities for disposable batteries. Those are airplane safe if they go in your carry-on luggage. 

Hope that helps.

Akiba

Hi Julien,
I use this mic for my raspberry pi setup and it works well picking up bird species singing in our garden while the mic and raspberry sit under our roof tiles.
Greetings,
Robin

Hi — I came across this thread and wanted to chime in on the microphone selection question, as it touches on some tricky hardware trade-offs.

For bird monitoring in the 50-20kHz range, the MEMS route is actually more accessible than it might seem. The ICS-43432 from InvenSense is a solid choice — it's an I2S MEMS microphone, so it connects directly to the Raspberry Pi's I2S interface without needing a sound card at all, and it has good sensitivity and low self-noise in the relevant frequency range. No plugin power needed, no USB complexity. A simple breakout board (SparkFun or Adafruit have them) and a few lines of configuration and you're up.

If you want to avoid I2S configuration entirely and still keep costs reasonable, the ICS-40720 you mentioned is also excellent and can be used with a simple circuit — a basic op-amp buffer stage is straightforward to build and eliminates the need for a sound card entirely. I can sketch out a basic circuit if that would help.

The Audiomoth uses the SPM0408LE5H-TB partly because of its very low noise floor, but for a Pi-based system the ICS-43432 gets you most of the way there with much simpler integration.

What's your target deployment environment — indoor, outdoor fixed, or field portable? That affects the enclosure and weatherproofing approach more than the microphone choice itself.

Hi — I came across this thread just joining the community. I'm an electronics and embedded systems engineer with 30 years of firmware and hardware design experience, mostly on STM32 and similar ARM Cortex-M targets.

The UAC2/Android issue is a known pain point. Android's UAC2 support has been inconsistent across vendors and API levels — it was only partially addressed in Android 5.0 and many OEMs never fully implemented it. Phil is right that if TinyUSB only exposes UAC2 descriptors, most Android devices will simply ignore the device.

A possible path forward without abandoning TinyUSB: some versions of TinyUSB do allow building a composite device or overriding the audio class descriptor manually. If you can post the relevant section of your usb_descriptors.c (specifically the audio function descriptor and the bcdADC field), I can take a look and see if there's a way to force UAC1 compatibility at the descriptor level without breaking the existing hosts.

Alternatively — and this might be the simpler path — have you considered a small STM32F0 or STM32G0 as the USB bridge? They have mature HAL USB audio support with UAC1 out of the box, cost under £1 in volume, and would leave your Pico firmware untouched.

Happy to help either way.

Love this advice down to the chip level 😀