Monday, 28 January 2019

DodoPi Bat Recorder: "hey, we have a live one!"

Bats flying around in mid-winter! ...whatever next?

It may be relatively rare compared to the number of recordings made each night during the summer...

...but just one recording shows that there are bats out there.

Inspired by the Bats London project, I've been running my new automatic bat recording system for almost every dry night during January (plus a few where the weather forecast was wrong!) in the hope of capturing any bat activity. This is the system that uses a Dodotronic Ultramic 192k USB microphone connected to a Raspberry Pi 3.

Even in the absence of any real bat calls, its been a useful exercise for testing the software and fiddling with the detection sensitivity parameters. However, a few days ago my patience was rewarded when a common pipistrelle flew past, leaving me with this single 10 second recording.

common pipistrelle bat echolocation social call spectrogram
Common Pipistrelle recorded 25th Jan 2019 at 5.32pm

On this particular night, the temperature was approximately 9'C according to the local weather forecast (from the BBC) at the time of the recording. So it was not particularly cold. You can see both echolocation calls, and the lower frequency social calls. It seems strange to me that this apparently solitary bat should waste energy producing social calls, as these are always considerably louder and of longer duration than the echolocation calls, and therefore must consume a lot more energy.

The other noticeable characteristic of this call sequence is the horrendous echo.

common pipistrelle bat echolocation social call spectrogram
Illustration shows 4 echolocation calls, the rest are believed to be just echos.

This is an unfortunate problem created by my choice of microphone location.

This is the view from my microphone's position on the 2nd floor.

There are a lot of reflective surfaces on the surrounding buildings, and the widows directly behind the microphone may be a major contributor.

I plan to operate this system out in the open during the warmer, dryer summer months to see if I can get some 'clean' recordings.

In the meantime I also want to make a revised system using an electret mic insert which (unlike my old system) does not need batteries. On the old system, even with batteries providing some degree of power supply isolation, I still get a kind on switching noise which is not present on my new 'DodoPi'.

Here are the system details



  1. What are your thoughts about this:
    example python application:

    1. I suppose the first question is what is your objective?

      Unfortunately I still don't know why I'm doing this, but I'd estimate its 2 thirds about designing a system and getting it working, and 1 third about recording bats and trying to work out what they are and what they are doing.

      I don't know much about AudioMoth but it looks pretty good and has low power requirements. Naturally when used as a bat recorder you will need to set rate to 192kHz or above.

      You are probably already aware of their forum:
      ...and I'd also suggest you join the Facebook group: Bat Call Sound Analysis Workshop
      ...and post your question there, as there are certainly a few AudioMoth users in this group.

      In my opinion, no system is capable of 100% accuracy when identifying bat species. Call variation and overlap between species appear to make this impossible at the present time. And manual identification is down to judgement (e.g. I'm 90% sure its bat A but it could be C or D). However, I suspect your 'invisible' interest in bats is like your 'visible' interest in garden birds.

      If I understand correctly, these are priced at $80, which is much cheaper than my Dodotronic mic. One disadvantage of both AudioMoth & Dodo is the lack of a high pass filter. Being so close to a main road, the odd boy-racer can saturate the a-d converter. I still haven't had a chance to test my cheap mic + preamp + Wolfson + Pi, which should filter out the traffic noise.

      Take a look at Arnold Andreasson's comments here: (you have to download the PDF to read it). He also mentions the lack of a high pass filter and some distortion as negative points for the AudioMoth.

      Also look at his 'Clouded Bats' project (lots of Python code):

      Arnold likes to receive email on this subject, but he is a busy-boy, so may take a while to get back to you.

      I hope some of this is helpful, but do come back if you have more questions.

    2. Oh, and this may also be of interest:

  2. Someone on the radio 4 "in our time" echolocation podcast pointed out that bats lungs are worked by their wing muscles so they get their calls energy free if they are flying.

  3. Thanks, that's interesting.

    I think each pulse/call happens when the wings come down. The call length and duration is probably influenced by the bat size; hence small bats tend to have shorter call lengths and often higher repetition rates.