Raise the low end of a recording, and generate low end that is not in it. A shelf can only multiply what the file already contains, which is why a thin recording stays thin no matter how far the bass control goes.
The three stages
Protect below. A high-pass that runs before anything else, so the shelf cannot amplify what sits under it. Rumble, air-conditioning, footfall, and any DC offset all live in the bottom two octaves, and raising a shelf over them turns a small problem into a large one. Set to 0 the stage is bypassed.
Sub harmonics. A frequency divider watches the band below the frequency you set and produces a square wave at half its rate, which is then filtered down to a sine and mixed back in at the level of the band it came from. A 120 Hz bass note produces content at 60 Hz. At 0 the generator contributes nothing and the tool is a shelf.
Boost. A low shelf at the same frequency, up to +18 dB, covering the original low end and the generated octave together.
Frequency sets all three
One control moves the shelf corner, the top of the band the divider listens to, and the low-pass on the generated sub. That coupling is deliberate: a sub generated above the shelf corner would be raised unevenly, and a divider listening above the shelf would generate content the shelf never touches.
Around 60 to 90 Hz suits material that already has a low end and needs more of it. Between 120 and 200 Hz suits thin sources, where the divider has to work on the lowest content that actually exists in the file.
Division, not distortion
Rectifying or clipping a signal adds harmonics above the fundamental. Neither produces anything below it, which is why a saturator makes bass sound louder without making it deeper.
Halving the rate is the operation that goes down. The divider tracks the band’s zero crossings with hysteresis set at a tenth of its envelope, flips its output once per input cycle, and so completes one of its own cycles every two of the source’s. The result is filtered twice at the frequency corner so what reaches the shelf is a sine an octave down rather than the square the divider made.
The divider runs on a 4x oversampled grid so its switching edges land closer to the true crossing. Switching on the sample grid alone would fold high-frequency energy back down into the band the tool is meant to be cleaning up.
Peak level rises faster than loudness
Low frequencies carry more energy per unit of perceived loudness than mid frequencies, so a +6 dB shelf at 60 Hz can add far more than 6 dB to the peak while sounding only slightly heavier. A file that measured -3 dBFS before can clip after.
The output meter latches Clip when a sample reaches full scale. When it does, either lower the boost or follow this step with a limiter set below 0 dB.
Monophonic sources
The divider runs per channel but shares one envelope across them, so the sub stays centred with the source rather than drifting between the speakers. On a mono file both channels produce the same sub, as expected.
Very low frequencies are close to omnidirectional and carry almost no stereo information at the distances speakers are heard from, so a centred sub is what a mix would have had anyway.