What is it with all these different ways of measuring and reporting dB’s?
We know that dB(A), or A-weighted measurements, is used in noise dosimetry, even though it is an ancient standard, and represents the sensitivity frequency profile (approximately) of our hearing at levels around 40 dBSPL.
And we know that dB(C), or C-weighting, is more representative of our frequency response at loud levels, actually up around 100 dBSPL. …but we generally listen around 70 dBSPL which has dB(B) more suited…
And then comes along the EBU and BS.1770-3 for reporting R128 compliance. And that supposedly better represents our loudness perception.
Which one should be used? (ignoring legal requirements for the moment).
We are really talking about multiple different things here, and unless you understand this fact, it will seem hopelessly confusing.
dB(A) actually is a crude approximation of our threshold hearing frequency profile, as seen here in the Fletcher-Munson equal-loudness contours:
What we’re interested is the threshold level, or bottom contour. But in the old days, they took the 40 dB contour as an approximation (it’s far easier to measure!). If you take that contour and invert it, it becomes close to a dB(A) measuring filter.
And the reason that is significant is that dB(A) allows us to measure the physical impact of sound on our hearing. Its shape is relatively unaffected by things relating to loudness at louder levels. That represents the efficiency of sound mechanically coupling to our hearing mechanism. Mechanical efficiency is affected by everything from the length of our ear canals, to the tapering of the basilar membrane, and the viscosity of the cochlear fluid.
dB(A) represents the degree to which sound at any frequency contributes to sound exposure. It has nothing to do with our perceptual abilities.
Now, dB(C) is specified in the EBU R128 documents for use in establishing sound system calibration for listening levels. The C-weighting, while supposedly representing our listening sensitivity versus frequency at 100 dBSPL, or one of the contours near the top of the bunch in the above diagram, also has nothing to do with our hearing sensitivity. It is chosen because of its relatively flatter response from 500 Hz to 2 kHz.
And when you play the EBU Reference Noise track for calibrating your system for R128 at -23 LUFS, you care about the flat response across its frequency range. Sound meters generally offer only two choices: A-weighting or C-weighting. Choose C, because it has flatter frequency response over its mid-range.
Finally, we come to LUFS measure, which is intended to reflect how we perceive loudness. It also isn’t particularly concerned about frequency response either. It has simply been found that by using a shaping filter ahead of RMS power measurement, with the lowest bass below about 37 Hz eliminated, and sounds above 1.5 kHz boosted by 4 dB, gives a gated RMS measurement about the same ability to predict human loudness perception as real listeners engaged in comparative studies.
So, dB(A) for gauging sound physical impact, dB(C) for calibration flatness, and LUFS for perceptual estimation.