I'm not familiar with the topic of sound. I have ultrasonic audio recordings sampled at 250kHz (16bit) in WAV format. The equipment used to record them is linked here: http://www.avisoft.com/usg/usg116Hn.htm

I've use the matlab programming language to plot the audio signal. While the X axis is just going to be the sample number (divide this by the sampling rate to get time), I don't quite know what the units for the amplitude are.

Typical values (which can be positive or negative) have very high precision and are of the order of 10^-4 or 10^-5. An example amplitude value is:

-3.051757812500000e-05 (at least according to Matlab)

Anyone have any idea of what units I am dealing with? I did not use the equipment myself, my supervisor did. When I asked him what units the amplitude were, he advised I look at the link, but I am not too certain I understand.

  • 1
    $\begingroup$ is there a gain knob between the transducer (what we call a "microphone" in less-than-ultra sound) and the A/D converter? do you have specs on the transducer (like look at the "Calibration Chart" in this B&K mic spec sheet). if you're in SI units, your ultrasound is in pascals and the transducer will convert to volts and the A/D converter will convert volts into dimensionless numbers which is what you will see in MATLAB. that is the only way i know to get a handle on the actual units of the actual data. $\endgroup$ Commented May 13, 2015 at 19:09
  • $\begingroup$ @robert bristow-johnson I know of the recording equipment but I've not actually used the equipment. I've only been provided with the audio recordings. The only thing I know about the equipment is from the link posted. I've read some research papers about some rats ultrasound calls producing 65-80dB, and I wanted to see what the decibel of my data would be. But I don't quite see how I can obtain this. $\endgroup$ Commented May 13, 2015 at 19:21
  • $\begingroup$ yeah, i found one page that relates Pa to mV and the page you mention relates sensitivity to dBu (about 775 mV if i recall). but it doesn't say anything about what is "full scale" (which likely is +1 or -1 in your MATLAB file if you're reading in a .wav). $\endgroup$ Commented May 13, 2015 at 19:28
  • $\begingroup$ another thing, even assuming a reference level, like 0 dB is 20 micropascals (common in acoustics), saying your rat produces 65 dB (relative to 20 micropascals) means nothing without knowing the distance between rat and transducer. there's that inverse-square law to think about. $\endgroup$ Commented May 13, 2015 at 19:37

1 Answer 1


Given that the gain of the system and/or ADC seems to be unknown, the scale of the sample values can only be determined from samples of a calibrated source of known value at a known distance from the transducer. With a variable enough gain knob (or AGC) the scale of the samples could be anything.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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