Raspberry Pi Automated Bird Recording System, Improved

I invested in a better USB microphone to replace the $18 Cmteck desktop unit that I used for my first swag at the BirdNET Pi. I wanted an omnidirectional microphone with a 24-bit, 48 kHz ADC and no in situ signal processing intended to “improve” the signal.

The best-reviewed microphone that fit the bill was a Blue Yeti Nano, with selectable omni/cardiod patterns, which I found going for about $90. Eventually it will need to go into a semi-permanent installation that will afford protection from rain and snow, but I’m not sure how I’m going to do that, yet. Since we are not having any precipitation at the moment, I’ve made use of the Yeti’s sturdy stand and planted it in the snow, near our feeders.

Hopefully none of our various mammalian visitors will want to nibble on the windscreen.

I also did some quick tests to convince myself that this microphone was indeed an upgrade. I began with a recording of a linear frequency sweep between 20 Hz to 20 kHz, played over a bluetooth speaker; this let me compare the frequency responses of both microphones. As I expected, the Cmteck microphone was less flat overall, falling off between 10 and 12 kHz. The Yeti made it out to about 18 kHz before it went off the cliff. I also estimated SNRs by playing bird calls over the speaker at nominal volume and recording them with both microphones. The Yeti provides about a 7 dB SNR upgrade.

So with the comparatively flatter response, I feel a little better about occasionally uploading some of the stronger captures into eBird. Here is a Blue Jay call recorded with the new setup:

This project has been a great deal of fun so far. It has alerted me to birds I’d otherwise have missed, and it is instructive to see what sounds trigger incorrect identifications. I imagine that the BirdNET developers are aware of many of the enviromental sounds that can trip it up, but this could be a good way to find more. The other evening, the unit was picking up a far-off dog that made occasional yips, and interpreted them as the sounds of a Black-crowned Night-Heron. That is far more understandable than how distant sirens are sometimes called Eastern Screech-Owls. A recording from 2 AM this morning was labeled Common Loon; it turned out to be the obnoxious hoots and hollers of some apparently inebriated party-goers down the street. Common loons indeed.

Based on a comparable/better Raspberry Pi kit that is available now, a bill of materials for this implementation is as follows:

CanaKit Raspberry Pi 4 Extreme Kit (includes the Pi, case, fan, power adapter, SD card): $140

Diivoo Weatherproof Electrical Box: $17

Blue Yet Nano Microphone: $90

Youshares Furry Outdoor Windscreen: $9

————-

Total: $256

My Rev 3 version will feature whatever weatherproofing I come up with for the microphone that does not compromise performance. Hopefully this won’t cost much; it might be a simple as a putting it under an inclined board. And it will be interesting to see if the Pi-based system is comparable in price to the Physical Universe Codec that Birdweather.com plans to make available soon.


3 thoughts on “Raspberry Pi Automated Bird Recording System, Improved

  1. Enjoying your posts about your experiences with BirdNET-Pi. I’ve had mine installed since August and have enjoyed all I have learned from it!

      1. Just once, after two months of continuous operation. That was in October and it has been running since. The discussion at GitHub has helped me through some slow-down, with the system falling far behind its normal near real-time detections. 🙂

Leave a Reply to legallyblindbirdingCancel reply