I am an amateur entomologist (and complete Noob to the audio world) and would like to use an array of several microphones to triangulate and map out populations of calling insects. The idea is that the microphone array would be attached to a weather balloon. The time differential between the sound wave impacting the different microphones would allow the insect's location to be mathematically determined in 3-Dimensional space.

The limiting factor I seem to be running up against is the sampling rate of the recorder. Sound travels so fast that I need a very high Hertz rate to keep the mathematical error low. To boil this down into two questions:

  1. What commercially available recorder has the highest sampling rate? All the ones I have found online max out at 192 KHZ. I need closer to 2.8 MHZ to accomplish my goal.

  2. Why is the sampling rate of recorders so slow when computer clock speeds are up well into the GHZ range now? You would think a computer with a clock running that fast could manage a higher sampling frequency.

  • 1. What research have you done so far & why wasn't what you found suitable? 2. This is a logical fallacy from incredulity, again showing lack of research.
    – Tetsujin
    Nov 8, 2021 at 9:29
  • Also, how will you stop the balloon and/or the gondola holding the equipment from spinning? if you're using a mic array for triangulation then any kind of spin or rotation will distort the map. Nov 8, 2021 at 10:40
  • @7HzResearch I plan to attach a 9 DOF motion sensor and a GPS chip to my pi setup that will go on the balloon. That should counteract the movement of the vehicle.
    – Alex Heebs
    Nov 8, 2021 at 14:30
  • @AlexHeebs It's standard procedure to ask for clarity here. Telling us what you've tried/researched before and why it doesn't work for you helps us arrive at a better solution. Also, looking up the basics of how a thing works works before you talk about it to a room full of professionals is probably a sensible precaution. Please avoid making personal attacks. From what you describe, you may be better off modifying an off-the-shelf ultrasound ranging system - these sensors are also available in some Pi dev kits. Nov 8, 2021 at 20:35
  • As far as I can see there have been no smart alec answers or dogpiles. Check out our code of conduct if you're unsure about how to respond: sound.stackexchange.com/conduct Nov 9, 2021 at 11:22

2 Answers 2


2/ about the audio recorder limitations

Audio recording devices are made to record signals which can be heard. Since our ears are limited to 20kHz, we just need a 40kHz sampling (the 20kHz is the half of 40kHz, named Nyquist frequency).

If we just sample a signal at 40kHz, we would need sharp filters to avoid recording frequencies above 20kHz aliased to audible frequencies. This explains the need for higher sampling frequencies (88kHz, 192kHz), which avoids aliasing and sharp filter artefacts but the need stays at a same magnitude.

1/ if you want better

If you really want an higher sampling frequency, you need measurement devices like some of National Instruments. Your sensor should be adequate too. (A microphone may filter your sound in the 20-20,000Hz range).

EDIT : and with NI and analog measures, you are limited to 1MHz. Analog and accurate measure is harder than just recording a stream of bits. Some systems are very very slow to improve the accuracy.

The sound speed is 340m/s. Then a single sample at 192000Hz is worth 1.8mm… if your insect flies, this can be quite accurate. More than the location of your balloons.

  • Thank you for your answer Frédéric. I'm curious where you come up with 1.8mm? The calculations I did showed much more than that. For example, let's assume that the calling insect is 100m away from the microphone array. For simplicity, let's also assume that the insect's position forms a right triangle with two microphones that are on the same plane, and those mics are 3 meters apart. The math for the time differential between the sound wave striking the two mics... Where "sqrt(..." is the Pythagorean Theorem: sqrt(100^2 + 3^2)m / 340m/s - 100m / 340m/s = 0.0001323s
    – Alex Heebs
    Nov 8, 2021 at 15:47
  • Continued... A sampling rate of 192000Hz is 25.4 time faster than the 0.0001323s time differential. Which allows for you to triangulate within 1/25.4 of a meter. Or a resolution of approximately 39.4mm Of course this changes depending on how far apart the microphones are placed.
    – Alex Heebs
    Nov 8, 2021 at 15:47
  • I have just computed 340/192,000. This can be near the precision in a very adequate situation (three sensors which makes an angle of 120° with the insect, a fourth above to have the 3D). It is like the GPS : satellites have to be present on each quadrant if we want a precise measure. If all sensors are roughly in the same direction, surely, the precision will be worse and you will need to compensate with an higher timing frequency. Nov 8, 2021 at 15:59
  • Your dephasing of 25.4 is for a move of your insect of 1.5m, then a single sample gives you a precision of 1.5/25.4=59mm… however, this doesn’t measure the distance : you need a third sensor, and quite far away from the 2 others. Note : you can do some interpolation, especially if the microphones filter at 20kHz. A calibration may be needed (DC-offset…) Nov 8, 2021 at 17:33

You are basically imagining doing a photograph (or video), just with acoustical waves rather than light waves. Atmospheric distortions are a real problem for photography at the kind of distance and scale you are thinking about, and there air is not the actual medium propagating the waves but a slight disturbance. There just is no way you are going to get a sharp enough "image" for the sake of identification and tracking.

Bats mainly trace their prey with echolocation (and one by one) rather than passive listening, even though of course echolocation is important on its own for avoiding obstacles. And they are biologically evolved specialists in insect location by sound and sound imaging.

  • 1
    We can add that our ear (and brain) have a more accurate sound image because the ear form alter the sound in a way which depends of the location and is recognised by the brain… then we can even situate a sound up/down + front/rear even with a binaural sound. A stereo sound (unaltered by an ear form) doesn’t indicates the up/down location Nov 8, 2021 at 21:28

Not the answer you're looking for? Browse other questions tagged or ask your own question.