13 June 2013

Elephant Speak. Infra-sound Detector

In a recent discussion with my brother-in-law he mentioned that it would be neat to be able to detect elephants communicating with one another. Apparently elephants use so called infra sound frequencies to achieve this interesting form of communication. This is a band of frequencies in the range of 4Hz - 30Hz. A band of frequencies lower than can be detected by the human ear. A search on the web found one particular paper written on the subject by some members of the University of Colombo School of Computing. This design uses a woofer speaker as the detecting device since a woofer will be more closely matched to the long wavelength infra-sound frequencies than a microphone which is of course optimized for the higher audio frequencies.

As a prototype I have built a DC coupled inverting amplifier and 2 pole butterworth low pass filter. See the details in the attached diagram below. I have set the amplifier gain at 270 and the filter cutoff frequency at  100Hz. The circuit drives a milliamp meter.

Since I don't have a woofer to hand I simply used a pair of coil movement headphones as the sensor. Initially I tried to couple the sensor into the amplifier using a power transformer. This was interesting. The transformer is meant for transforming 240VAC to 12VAC. This provides a 20:1 turns ratio. Thus I could achieve a 400 times impedance transformation. This higher input impedance to the amp meant that the amp was extremely quiet. No noise floor could be detected in some earplugs which I connected to the output of the amp. I was easily able to hear loud audio in the earplugs from the earphone 'microphone'. The circuit exhibits excellent sensitivity. The problem is that the high input impedance and the transformer picks up significant hum. So while the amp is 'noiseless' the hum is problematic. I tried putting the transformer in a metal container and screening the cable, however I was unable to eliminate the hum.

In the end I settled for connecting the mic directly to the amp via a 1000ohm input resistor. This circuit is definitely noisier than the transformer coupled circuit. However there is no hum and the circuit is well behaved. I set the gain to 270.

Then I built a 2 pole filter as per the diagram below. This appears to be working. There is no detectable noise and  the system appears quite sensitive to low frequencies. The lack of noise certainly points to the filter working. I need to look at the performance on an oscilloscope (which I don't have) to really understand how it is performing. I am able to drive a 500mA meter to FSD by scratching the earphone diaphram.

Next Steps:

  1. Build a 4Hz - 300Hz variable oscillator and amp so that I can generate a test signal and thus get some idea of selectivity, sensitivity and linearity using two woofers (one as the transmitter and one as the receiver) (my sister has two medium range speakers which I will try next).
  2. Find some elephants.
The Future
If the prototype works then there are many applications and advancements that can be implemented. The most value-add would be to interface the detector to an A/D converter so that digital signal processing can be applied. The Colombo University project utilized a 44KHz samples/second fourier transform technique which would yield the elephant language patterns. Now that would be exciting to research!

Prototype 1 _ Infrasound detector
Prototype 1 Infrasound detector

No comments:

Post a Comment