# What are the units of a .wav file?

I am working on calibrating my recording system. I have a signal generator playing a 500mVpp sine wave signal into my utrack24 recorder. I am trying to determine what the units of the recorded utrack .wav file is so I can compare that to the input signal. I have read the .wav file into matlab with audioread. I can use either the normalized version of the file or the 'native' version; I'm not sure which is better.

• the numbers Tendero alludes to are proportional to the input voltage at the sampling instances, but there are at least one gain knob in between that scales this and makes your units undetermined. if you are certain about the 500 mVpp, then you should use that information viewing the scaled numbers in your digital recording (use the 'native' version) as you examine the waveform from peak to peak. Feb 28 '18 at 21:36

Once an audio signal has been digitalized, it has no units and it is treated as a number.

You could use dBFS (decibels relative to full scale) as unit, though. The value of 0 dBFS represents the maximum digital signal level.

• So would I just convert to dBFS using 20log10(x)? But then how do I compare that to the input signal of 500mVpp? I can't convert that to dBFS...or can I? Mar 1 '18 at 19:38
• As far as I understand, your input signal is analog while your output signal is digital. I wouldn't say you can just 'compare' them. One of them represents voltage, the other one represents dimensionless numbers. Mar 1 '18 at 20:49

You must consider the microphone sensitivity while calibrating the recording system if you want to have an idea of what your digital samples actually meant because once they are digitized this information is lost. They are just numbers and you can't be sure what they meant or how strong was the audio source that generated them.

The sensitivity is the electrical signal at the microphone output to a given acoustic pressure as input. The reference of acoustic pressure is 1 Pa or 94 dBSPL at 1 kHz (this is the effective pressure a source transmitting a determined power level at a certain distance delivers at the microphone) compared to the threshold of hearing (20 μPa). This means that unless otherwise stated the microphone sensitivity will be tested by putting it in front of a speaker that delivers this amount of sound pressure.

The level of 0 dBFS is assigned to the maximum possible digital level (for instance). 'FS' stands for 'full scale'. 0 dBFS is the level of a signal that is at the maximum level your system can cope with without clipping the tips of the waveform. For example, a signal that reaches only 50% of the maximum level has a level of −6 dBFS, which is 6 dB below full scale. Conventions differ for RMS measurements, but all peak measurements smaller than the maximum are negative levels.

The sensitivity of digital microphones, with units dBFS (decibels with respect to digital full scale), is not straightforward. For an analog microphone with a voltage output, the only limit to the size of the output signal is the practical limit of the system’s voltage supplies. Although it may not be practical for most designs, there is no physical reason why an analog microphone couldn’t have 20 dBV sensitivity, with a 10 V output signal for a reference-level input signal. This sensitivity could be accomplished as long as the amplifiers, converters, and other circuits could support the required signal levels.

Sensitivity of a digital microphone is less flexible; it depends on a single design parameter, maximum acoustic input. As long as the full-scale digital word is mapped to the microphone’s maximum acoustic input (the only sensible mapping, really), the sensitivity must be simply the difference between this maximum acoustic signal and the 94 dB SPL reference. So, if a digital microphone’s maximum SPL is 120 dBSPL (let’s assume that this will produce an output of 16 bits), then its sensitivity will be –26 dBFS (94 dBSPL – 120 dBSPL).

If the system isn't calibrated or has any arbitrary gain stage(s), then the units are relative to whoever or whatever (AGC, limiter, etc.) twiddled the gain knob(s). Could be anything, especially if you see either clipping, or all values are well below the max sample magnitude.