Engineering reference · Editorial desk, Chicago
Digital Audio Processing & Loudness Mastering Standards in 2026
A working reference for engineers who measure, repair and finish audio. This desk gathers practical notes on frequency-domain editing, spectral restoration, headroom behaviour, non-destructive signal graphs, loudness normalisation and the delivery targets that streaming outlets and broadcast outlets now expect from finished masters.
Open the referenceWhat this reference covers
The material is arranged as five channel groups, in the same order a session usually unfolds: understand the signal spectrum, clean up what is broken, choose a number format that will not betray you, build the mix in a reversible way, then finish to a measured loudness target. Every article is written as plain engineering prose with tables of figures, so a reader can move from the theory of a transform to a concrete parameter value without leaving the page.
The desk does not rank brands and does not host product files of any kind. Where a named audio environment appears, it appears only as an object of technical comparison, described by measurable characteristics such as bit-depth handling, edit-graph model, metering resolution and signal routing. No logos, no screenshots and no interface copies are reproduced anywhere in this reference.
Fundamentals
How a transform turns a waveform into bins, why windowing matters, and how to read a spectrogram without fooling yourself.
Restoration & architecture
Estimating a noise profile, subtracting it, gating the residual, and keeping the whole repair reversible.
Mastering & delivery
Loudness units, true peak, and three reference processing chains for voice, archive and music delivery.
A session in five moves
The diagram below is the shape most sessions follow on this desk. Each node is a decision point, and each decision point has its own article. Moving left to right keeps measurement ahead of processing: you first learn what is in the recording, then change only what you can later justify.
Reference values at a glance
These are the headline figures quoted throughout the desk. They are starting points for a delivery decision, not universal law, and every one of them has a dedicated article that explains when it applies and when it should be set aside.
| Quantity | Typical value | Where it is used |
|---|---|---|
| FFT size (music) | 4096 – 8192 | spectral editing, wide-band view |
| FFT size (speech) | 1024 – 2048 | noise profiling, mouth noise |
| Hop size | 25 % of FFT | overlap-add reconstruction |
| Internal format | 32-bit float | mix and effects stage |
| Loudness target, podcast | -16 LUFS | spoken-word delivery |
| Loudness target, music | -14 LUFS | streaming delivery |
| True peak ceiling | -1 dBTP | lossy encode guard band |
Index of the desk
Every article on this reference is reachable from the list below. The first two entries under each heading are the channel group named in the fader bank; the remaining entries are supporting legal and editorial pages.
Analysis and repair
Mixing and finishing
Desk
Editorial and legal
Corrections are welcome. If a figure here disagrees with a published standard, write to the editorial desk and the article will be revised with a dated change note.