Recording ultrasonic vocalisations of bats


Imagine a sonic environment untroubled by aircraft noise, almost immune to the effects of wind noise without using any particular wind protection. This environment is in fact already there all around us - and small mammals in particular vocalise in this space.

Humans are large mammals, and this benign audio environment lies above our hearing range. It is the spectrum known as ultrasound. Many mammals vocalise in the range from 20kHz to 100kHz, rodents in particlar, such as rats and squirrels. One type of mammal in particular is associated with ultrasonic vocalisations because it uses these signals to hunt its prey in the dark - bats. In the past recording this frequency range was the domain of specialist equipment, but with the advent of digital recorders sampling at 192kHz and more, wildlife sound recordists can start to venture into these regions.

Making bat sounds audible


There are three main approaches to recording ultrasonic vocalisations. One is to frequency shift them down into a range we can hear. This is typically done by heterodyning, which shifts the sound down in frequency to less that 10kHz. The disadvantage of this method is that we have to listen to a selected 10kHz  chunk of the 20-100kHz spectrum, for example 40kHz to 50kHz, and miss everything else outside that range. The advantages of this method are that it is cheap, reasonably sensitive and with skill can be used to determine bat species from their call. It also works in real time, ie audible sound is available at the same time as the ultrasonic vocalisation.

frequency division

The second approach is to use frequency division. Also used in bat detectors, this severely distorts the signal particularly in the presence of more than one sound source and can be discounted for the purposes of wildlife sound recording where the recording, rather than realtime detection is the aim.

time expansion

The third approach is time-expansion - the ultrasonic waveform is recorded and the playback speed is reduced by  about 10 times. Spectrograms can be made of the original recording which captures the entire signal in a similar way to normal spectrograms of audio recordings, and the slowed playback enables us to hear details that woould be beyond the relatively coarse temporal resolution of human hearing compared with smaller mammals.

Slowing the playback speed is easy on a coputer - simply record at 192kHz, adjust the sampling rate of the recording from 192000 Hz to 19200 Hz without resampling and play the sound!


Richard Margoschis interviewed Mike Houston about recording Curlews (from WSRS Circulating Tape CT21)

articles index

Valid XHTML 1.0 TransitionalValid CSS!