dB loudness reference

A decibel is a ratio, not a unit. It is 20 times the log of a measurement divided by a reference, and the suffix names the reference. dB SPL compares air pressure against 20 micropascals, roughly the quietest sound a healthy ear detects, so the scale counts up from 0 and louder sounds are bigger numbers. dBFS compares a digital sample against full scale, the largest value the format can hold, so 0 sits at the top and every real signal is a negative number. Same math, different worlds.

What connects them is your signal chain. A sound at some SPL reaches the mic, the capsule turns pressure into voltage, the preamp gain scales that voltage, and only then does it become a dBFS value. Turn the gain knob up 10 dB and the same 60 dB SPL conversation lands 10 dB higher on the digital ladder. That is why this chart draws a divider between the columns instead of arrows across them: any chart that lines up a specific SPL with a specific dBFS number is quietly assuming one particular mic at one particular gain setting.

On the digital side, the gap between your average level and your peaks is your working dynamic range, and controlling it, so quiet words stay audible while loud moments stay under the ceiling, is exactly the job of a compressor; see what a compressor does. And you can measure where your own room actually sits on the dBFS ladder with the mic noise floor test, which reports an honest dBFS reading of your floor in five seconds.

Six dB is a doubling. Adding 6 dB doubles the amplitude, whether that is air pressure or a sample value, and subtracting 6 halves it. So -6 dBFS is half of full scale, -12 is a quarter, and the drop from healthy speech peaks at -8 down to a quiet noise floor at -60 spans a factor of several hundred. Compressing enormous ranges into small readable numbers is the whole reason audio runs on logarithms.

The dBFS targets are conventions, not laws. 0 dBFS is defined, because a sample cannot exceed full scale, but -18 for average speech and -8 for peaks are working habits from voice production: they keep a laugh from clipping while holding the voice well above the floor. The SPL column is looser still. Whisper 30 and conversation 60 are textbook approximations that shift with distance and metering weighting, so treat every SPL row as a landmark, not a measurement.

If your levels sit far below the landmarks, say peaks near -25 and an average near -35, nothing is broken, your gain staging is just low. The fix is usually the input gain on your interface, or a clean gain boost placed before everything else in the chain. The step-by-step walkthrough is in how to fix a mic that is too quiet.