CH-04 · Mastering
Processing chains
A chain is an order of work, and the order matters as much as the settings. The three reference chains below cover the jobs this desk meets most often: a spoken-word podcast, an archive restoration, and a loudness normalisation pass to a streaming target. Each is presented as a table so it can be copied to a session sheet and followed step by step.
Podcast voice chain
The aim is intelligibility on a small speaker without losing the natural shape of speech. Work in this order, and bypass the whole chain to check that each step earns its place.
| # | Stage | Setting | Purpose |
|---|---|---|---|
| 1 | High-pass filter | 80 Hz, 12 dB/oct | remove rumble and desk thumps |
| 2 | De-hum | 50/60 Hz comb | remove mains tone and harmonics |
| 3 | Downward expansion | threshold -42 dB, ratio 2:1 | settle breath and room tone |
| 4 | De-esser | 5.5 kHz, 4 dB reduction | tame sibilance |
| 5 | Presence equaliser | +2 dB at 3.5 kHz, wide | lift consonant clarity |
| 6 | Compressor | 3:1, 8 dB reduction | even out the take |
| 7 | Gain to target | -16 LUFS integrated | delivery loudness |
| 8 | True-peak ceiling | -1 dBTP | encode guard band |
Steps one to three repair the recording; steps four to six shape it; steps seven and eight deliver it. If a step cannot be justified in one sentence, it should be removed before delivery.
Archive restoration chain
Restoration runs narrow repairs first and broad tone work last, so that each stage can be auditioned against an untouched reference. Keep the source region read-only and record every setting.
| # | Stage | Setting | Purpose |
|---|---|---|---|
| 1 | Reference capture | noise profile, 3 s | estimate the unwanted signal |
| 2 | De-hum comb | mains plus 8 harmonics | remove steady tone |
| 3 | Impulse repair | threshold +8 dB over floor | rebuild clicks and snaps |
| 4 | Spectral subtraction | factor 1.5, floor -6 dB | reduce broadband hiss |
| 5 | Downward expansion | 1.6:1, slow release | settle residual wash |
| 6 | Low-band tilt | -1.5 dB below 120 Hz | compensate shelf loss |
| 7 | Gain to target | -18 LUFS integrated | access-level listening |
Stop as soon as the recording sounds clean. Every extra decibel of reduction takes something from the wanted signal, and a restored archive that has lost its room is no longer an archive.
Loudness normalisation to -14 and -16 LUFS
Normalisation is a measurement pass followed by a single gain decision. It is not a reason to limit harder; on streaming delivery the outlet will normalise the loudness anyway, so the only lasting effects of extra limiting are added density and softened transients.
| # | Step | Setting | Check |
|---|---|---|---|
| 1 | Measure integrated | whole piece | read the loudness figure |
| 2 | Compare to target | -14 or -16 LUFS | note the difference |
| 3 | Apply gain trim | exact difference | no other change |
| 4 | Re-measure loudness | whole piece | target reached |
| 5 | Check true peak | -1 dBTP | below ceiling |
| 6 | Encode guard check | lossy pass | no overs |
Three chains, one discipline: measure first, change one thing, listen, and keep the source untouched so the next revision takes no extra work.