Detecting high-frequency animal sound

This page has a information on building bat detectors with circuit diagrams and design discussion. Bertrik’s bat detector page

Condenser mics are the only sensors that have a broad frequency response and work up to 100 – 200 kHz.

The sensor from Polaroid Sonar One Step SX-70 camera is reported to be a SENSCOMP #604142, a condenser microphone with frequency response up to 100 kHz. This sonar paper has a diagram of the sensor (pdf).

Here is the SensComp’s 600 Series Smart Sensor datasheet–includes the circuit diagram (pdf). It uses TL851 and TL852 to send and receive a distance sensor pulse train. This page has more detail on the same (or a very similar) circuit (link). A ppt from the web showing the circuit (ppt).

A Linear Tech app note on using the sensor in a acoustic thermometry circuit. pdf, EDN article

The Polaroid Sonar cameras are available for $10 – $20.

Discussions about making condenser microphones (groupdiy link).

Source for new microphone capsules (
Aurycle microphone kit

How to process and store data at 1-2 Msps

At 1 Msps, one cycle of 100 kHz is 10 samples, 10 kHz is 100 samples, and 1 kHz is 1,000 samples

“A fast FFT, 128 points and 8 bits, will take about 20 – 30ms on a 8Mhz / 8 avr.”

Arduino FHT library
“FHT 256 samples, 3.18 ms on an Arduino”

Auduino FFT code

Use a DSP like the Microchip dsPIC33FJ16GS502. It can sample at 2 Msps, 10-bit, and perform FFT. Perform a FFT every 1000 samples, every 1 ms, and save only the FFT results. The low frequency end won’t be captured well in a ms of data, so also pull out a sparse 1:100 sample set and run a 2nd FFT. This is still a lot of data–a FFT with 256 bins x 2 bytes x 1000 / s = 1.8 Gb / hr, so discarding empty bins to compress it would be useful.

The output can be shown immediately by lighting up a row of LEDs. Sound can be reconstructed from the FFT record on a PC later.

Compressing audio data is a well-examined problem, this is what the MP3 format does.
Audio compression sound example
Speex voice compression codec
Microchip dsPIC Speex library

Sampling and frequency output:

1 Msps, 1024 samples (1 ms), top FFT bin is 500 kHz, bins are 500 Hz
At 20 Ksps, 1024 samples (50 ms), top FFT bin is 10 kHz, bins are 10 Hz

So combining low and high frequency FFT gives: 10 Hz – 10 kHz (low end +/- 10 Hz), 10.5 kHz – 200 kHz (high end +/- 500 Hz). Music notes are 15 Hz apart, so at the low end the artifacts would be noticeable.

Development and prototyping

The basic operations required in this project are quite similar to well established audio handling projects. Develop the basic functions as an audio project, then scale up. The microphone amp, ADC, and data handling can be done with a audio as the testing target.

For example, at 50Ksps, 1,000 samples blocks for FFT, this can be done for audio on a PIC. Save the data to an SD card. Any microphone can be used. When it works, upgrade the mic to a condenser, then the microprocessor.

A PIC running at 64MHz, sampling at 50Ksps, each 1,000 sample block is 20 ms. This gives 1+ million ticks per FFT.

PIC circuit design

SD card library

SD card PIC book chapter

MEMS ultrasonic microphone
SPH0641LU4H-1, $2.01, response to 100kHz
bat detector project using MEMS mic, on/off

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>