Compendium

Mastering Chain — Deep Reference

28 min read

Mastering Chain — Deep Reference

A working reference for the modern mastering signal path, the decisions that distinguish a competent master from a great one, the loudness specifications imposed by streaming platforms, the format-specific deliverables (vinyl, Atmos, CD, MFiT), and the engineers whose signature decisions have shaped the sound of recorded music. Companion to the Tier-1 audio-production-mixing-mastering note, which covers the broader mix-and-master chain; this note drills into the mastering stage specifically.

See also

1. What mastering is, and isn’t

Mastering is the last creative + technical step before distribution. It does five things:

  1. Tonal correction — fix problems the mix engineer couldn’t or didn’t catch.
  2. Cohesion — make the album sound like one record across tracks recorded in different sessions, rooms, mixes.
  3. Loudness + level — get the program to the right loudness for its destination, with appropriate dynamic range.
  4. Format adaptation — generate the deliverable specific to each destination (CD, streaming, vinyl, Atmos, DDP).
  5. Quality control — final listening pass for clicks, distortion, dropouts, sequencing, metadata.

What mastering is not: a way to fix a broken mix. The mastering engineer has a stereo file (or stems if delivered); the leverage is small compared to mix moves on individual tracks. A bad mix masters to a bad master. The Bob Katz adage: “Mastering can make a good mix great. It can make a great mix sound on every system. It cannot make a bad mix good.”

The mastering chain is short — usually 4 to 8 processors — because every additional stage erodes transients and the things that make a master sound “open.” A surgical engineer like Bob Ludwig will often run a song through nothing but EQ, gentle bus compression, a saturation stage, and a limiter.

2. The canonical signal path

Mix bounce (32-bit float or 24-bit, peak ~ -6 dBFS)
    ↓
1. High-pass / DC removal (subsonic clean-up)
    ↓
2. Subtractive corrective EQ (cut problem resonances, mud, harshness)
    ↓
3. Multiband dynamics (control unruly bands)
    ↓
4. Bus / glue compression (1–2 dB gain reduction)
    ↓
5. Saturation / tape (harmonic body, soft transient control)
    ↓
6. Tonal-shaping additive EQ (broad-stroke musical curve)
    ↓
7. Mid/Side or stereo image processing (width, mono compatibility)
    ↓
8. Limiter / maximizer (final ceiling + loudness)
    ↓
9. Dither (only if reducing bit depth on output)
    ↓
Master file (24-bit WAV typical, 16-bit for CD/DDP)

The order varies by engineer. Bob Katz famously puts mid-side EQ before compression so the compressor isn’t reacting to imbalanced sides. Mandy Parnell often runs analog saturation early in the chain to soften the signal before any digital processing. Emily Lazar runs parallel chains and blends.

3. Subtractive corrective EQ

The first sonic move is to find and fix problems. Standard targets:

  • Subsonic energy (below 20–30 Hz) — wasted headroom, especially on vinyl masters where it causes cutter-head failure. HPF at 20–30 Hz with a 12 or 18 dB/oct slope.
  • Mud (200–400 Hz) — broad cuts of 1–2 dB with a wide Q (0.5–1.0) clean up congested mixes.
  • Boxy mids (400–600 Hz) — cardboard tone often sits here.
  • Honk / nasal (700 Hz – 1.5 kHz) — common on close-mic’d vocals.
  • Harsh (2–4 kHz) — vocal sibilance, cymbal edge; surgical cuts.
  • Sizzle (6–9 kHz) — cymbal wash, sibilant esses; dynamic EQ or de-esser is often better than static cut.
  • Resonances — narrow Q peaks (Q 8–20) at specific problem frequencies; sweep with a narrow boost to find, then cut at the same frequency.

Tools at this stage:

  • FabFilter Pro-Q 4 (2024) — the dominant modern mastering EQ; linear-phase, natural-phase, and zero-latency modes; dynamic bands; M/S support; EQ Match (analyze a reference and match curve).
  • DMG Audio EQuilibrium — every classic EQ topology (Pultec, API, Neve, SSL, Massive Passive) emulated and freely combinable; mastering favorite for surgical work.
  • Maag EQ4 (and EQ2) — passive-style with a “Air Band” shelf at 40 kHz (yes, above hearing); the “open up” EQ used on countless modern pop masters. Hardware version by Cliff Maag is in many mastering rooms.
  • Manley Massive Passive (and Massive Passive Mastering) — passive parallel-band tube EQ; the bold, musical wide curves; expensive (400 plugin).
  • Sontec MES-432D / MEP-250EX — Burgess Macneal’s Sontec is the mastering EQ; Bernie Grundman’s room has had them since the 1980s. No modern plugin emulation does it justice.
  • GML 8200 — George Massenburg’s transparent surgical mastering EQ; clean Q response.
  • SPL PQ Mastering EQ — discrete + analog VU.
  • Weiss EQ1 / Weiss DS1-MK3 — Daniel Weiss’s Swiss-built digital mastering EQ; reference standard. UAD has a Weiss EQ1 plugin.
  • Bettermaker Mastering EQ — hardware/plugin hybrid (controlled-by-plugin analog hardware).

4. Mid-Side (M/S) processing

M/S decoding turns a stereo signal into a Mid channel (L+R, the mono part) and a Side channel (L−R, the difference, the stereo information). Process them independently and re-encode. The technique was originated by Alan Blumlein in 1933 for stereophonic recording; it became a mastering staple in the 1990s as digital tools made the decode/encode trivial.

4.1 Where M/S helps

  • Tighten low end without mono’ing the bass — HPF the Side below 120 Hz so sub-bass is mono (vinyl-safe, club-PA-safe) while leaving the wide pad untouched.
  • Widen the high air without phase issues — high-shelf boost the Side above 6–8 kHz adds perceived width without compromising mono compatibility.
  • De-mud the mids — cut 300–500 Hz on the Mid channel to clear vocal congestion without affecting the stereo guitars.
  • Tame an aggressive lead — compress the Mid channel only; the Side stays uncompressed for ambience.
  • Mono-compatibility check — phase-flip the Side and listen to what disappears in mono; reveals overly wide elements.

4.2 Tools

  • Brainworx bx_digital V3 — the canonical mastering M/S plugin; 11-band parametric per channel, M/S meter, mono-maker, side-channel level.
  • FabFilter Pro-Q 4 — M/S, L/R, or stereo mode per band.
  • Plugin Alliance bx_console N, bx_console SSL 9000J — channel-strip emulations with M/S processing.
  • iZotope Ozone 11 EQ, Imager — mastering-suite M/S.
  • Mäag Audio EQ4M — M/S version of EQ4.

4.3 Mono compatibility, correlation, vectorscope

The correlation meter shows the dot product of L and R: +1 = identical (mono), 0 = uncorrelated, -1 = phase-inverted (will cancel in mono). A healthy modern pop master sits around +0.3 to +0.7 average correlation. Anything that goes negative for extended periods will collapse on mono playback (AM radio, Bluetooth speakers, club PAs summed to mono, vinyl played back through a single speaker).

The vectorscope (Lissajous display) plots L vs R on a 45° rotated axis: vertical = mono, horizontal = anti-phase, blob shape = stereo image. Stock plugins: iZotope Insight 2, Waves PAZ, Pro Tools Phase Scope, Nugen Audio Visualizer 2.

5. Multiband dynamics

Multiband compression splits the program into 3–6 frequency bands, compresses each independently, then re-sums. Lets you tame a boomy 200 Hz region without compressing the top end, or de-ess at 6 kHz without affecting the 2 kHz body.

5.1 When to use it

  • Inconsistent low end across tracks — multiband compresses sub + bass band to glue them.
  • Excessive sibilance in the mix bus — narrow band at 5–9 kHz, fast attack/release.
  • Mid-range honk that varies with vocal phrase — dynamic EQ does this better, but multiband works too.
  • Album-cohesion balancing — gentle multiband moves homogenize across-track tonal differences.

5.2 The hazards

Over-multiband-compression is the most common mastering mistake. Symptoms: pumping in one band, weird “phase-y” coloration from band-split filters, loss of impact on transients. Modern dynamic EQ (FabFilter Pro-Q 4 in dynamic mode, Pro-MB) often does the same job more transparently because it’s just EQ that engages above a threshold — no band-split crossover artifacts.

5.3 Tools

  • FabFilter Pro-MB — modern multiband; up to 6 bands; clean dynamic mode; the workhorse.
  • FabFilter Pro-Q 4 (dynamic bands) — for narrow-band ducking moves; often replaces Pro-MB.
  • iZotope Ozone 11 Dynamics — bundled in Ozone mastering suite; module-based; intelligent assistant suggests crossovers.
  • Waves C6 / C4 — older but still solid; C6 has two floating bands you can place anywhere.
  • TDR Nova GE (free Nova, paid Nova GE) — dynamic EQ in the same family.
  • Brainworx bx_dynEQ V2 — dynamic EQ specifically.
  • Maselec MLA-4 — hardware Leif Mases multiband; rare and revered.
  • Weiss DS1-MK3 — Daniel Weiss’s hardware multiband/de-esser; reference unit in many mastering studios.

6. Bus compression — the glue

A single broadband compressor at 1–2 dB gain reduction adds cohesion. It’s not for loudness, it’s for “feel.” Slow attack (30–100 ms) preserves transients; auto-release or 100–250 ms release matches the program’s groove.

6.1 Canonical mastering compressors

  • Shadow Hills Mastering Compressor — Peter Reardon’s stepped-control 2-stage compressor (optical + discrete VCA), with switchable Iron output transformer (Nickel for transparent, Iron for color, Steel for big). The current must-have hardware in elite mastering rooms; $7,000+. Plugin: Plugin Alliance Shadow Hills Mastering Compressor.
  • Manley Variable Mu (stereo, mastering version) — EveAnna Manley’s vari-mu tube compressor; smooth, slow, the “glue” sound on the Daft Punk Random Access Memories master. Plugin: UAD Manley Variable Mu.
  • SSL G-Series Bus Compressor (XLogic G-Bus, Fusion’s compressor section) — the “SSL glue”; VCA, 4:1 ratio, fast attack, often used in mastering for a slight punch increase. Plugin: SSL Native Bus Compressor 2, Waves SSL G-Master Buss, Cytomic The Glue, UAD SSL 4000 G Buss Comp.
  • Maselec MLA-3 / MLA-2 — Leif Mases’s mastering compressor; transparent VCA.
  • Crane Song STC-8 / Trakker — Dave Hill’s hybrid stereo compressor.
  • Chandler Limited TG12413 / Zener Limiter — EMI-style vari-mu.
  • Tube-Tech LCA 2B / SMC 2BM — Danish vari-mu and mastering multiband.
  • API 2500 — VCA with “Thrust” character control; can be used on the mix bus or as a mastering glue.
  • Tokyo Dawn Labs Kotelnikov GE (TDR) — mastering-grade plugin VCA with program-dependent release. Free version Kotelnikov is excellent.
  • FabFilter Pro-C 2 (Mastering or Bus mode) — clean transparent compression.
  • DMG Audio Compassion — extreme flexibility, mastering-oriented.

6.2 Parallel compression in mastering

Less common than in mixing, but used: bus a copy of the master through aggressive compression (4:1, fast attack, ~6–8 dB GR) and blend at -15 to -20 dB underneath. Adds density without losing transients. The “Andrew Scheps trick” used in some rock mastering chains.

7. Saturation — harmonic body

A mastering saturation stage adds 2nd-order (tubes), 3rd-order (transistors / FETs), or mixed harmonics that the ear reads as warmth, weight, and presence. Gentle saturation also softly limits transients before the final limiter, allowing more loudness without harshness.

7.1 Saturation models

  • Tape (Studer A800, Ampex ATR-102, MCI JH-110) — adds 3rd-order on heavy saturation, soft transient compression, gentle high-frequency roll-off. UAD Studer A800 and Ampex ATR-102 are reference emulations.
  • Tube (Pultec EQP-1A, Manley VOXBOX, Fairchild 670 tube path) — 2nd-order dominant; warm, “big.” UAD Pultec Pro Legacy, UAD Manley Massive Passive, Pulsar Audio Massive.
  • Transformer (Neve 1073, API 2520+325 transformer, Cinemag, Carnhill) — saturates at high signal levels; adds low-mid weight.
  • FET / transistor (SSL VHD, Empirical Labs Distressor, Black Box HG-2) — 3rd-order harmonics; aggressive at high drive.
  • Sampled / Acustica — convolution-based Acustica Audio plugins capture full circuit response from real hardware via sampled impulse responses.

7.2 Specific go-to processors

  • Soundtoys Decapitator — fast, simple, five saturation models; the everyone-uses-it modern saturator. “E” mode (EMI TG12345) for British console crunch, “N” (Neve 1057) for vintage warmth, “A” (Ampex 350) for tape.
  • FabFilter Saturn 2 — multiband + multistage drive; precise control.
  • Plugin Alliance Black Box Analog Design HG-2 — Mac Marshall’s hardware HG-2 emulated; tubes + transformer; mix and mastering saturation flagship.
  • Plugin Alliance Black Box Analog Design HG-2MS — M/S version, with stereo-specific control.
  • UAD Studer A800, Ampex ATR-102, Oxide Tape — the mastering tape emulations.
  • Acustica Audio Taupe — sampled tape behavior, multiple machine models.
  • Slate Digital Virtual Tape Machines (VTM) — Studer A827 + MCI JH-24 emulation.
  • Softube Tape — three machine models; lower CPU than UAD.
  • Klanghelm SDRR — tape + tube + console hybrid; cheap and fast.
  • iZotope Ozone 11 Exciter — multiband saturation with band-specific harmonic generation.
  • Pulsar Audio Mu, Pulsar Massive, Pulsar Echorec — French studio emulating French + British hardware.

7.3 Saturation chain depth

Common mastering arrangement: one analog/tape stage early in the chain at low drive (for character), one digital saturator late (for “loudness” — pushing average level without limiter pumping). Stacking saturators rather than driving one hard distributes harmonic distortion across processors and avoids any single stage sounding distorted.

8. De-essing in mastering

A mix-bus de-esser is sometimes needed when the mix engineer left sibilance unaddressed. Modern de-essers are spectrally aware (only react to actual sibilant content) and often M/S-aware (de-ess the Mid only, preserving Side air).

  • FabFilter Pro-DS — modern transparent de-esser with adjustable spectrum.
  • oeksound Soothe 2 (and Soothe Live) — dynamic resonance suppressor; engages on any sharp peaks above threshold, not just sibilants; effective for cymbal harshness, guitar resonance, and vocal sibilance.
  • Waves DeEsser, Sibilance — older but solid.
  • Eiosis E2 — adjustable detector frequency band.
  • iZotope Ozone 11 De-esser module — bundled in mastering suite.

9. Stereo imaging and width

Beyond M/S EQ, dedicated imagers manipulate the L/R relationship:

  • iZotope Ozone 11 Imager 2 — 4-band stereo widener; the free standalone Ozone Imager 2 plugin is also widely used.
  • Brainworx bx_stereomaker / bx_stereo / bx_meter — pseudo-stereo, widening, and metering.
  • Waves S1 Imager — classic widening + rotation + asymmetric Side processing.
  • Plugin Alliance bx_solo (free) — solo M, Side, L, R.
  • Polyverse Wider (free) — minimum-phase widening with strong mono compatibility.

9.1 Don’t over-widen

Aggressive widening (Side channel boosted +6 dB or more) creates phase issues and mono collapse. Modern algorithms (Polyverse Wider, Brainworx Pseudostereo) maintain phase coherence; older approaches (Haas delay tricks, simple Mid attenuation) trade mono compatibility for perceived width. Mastering should err on the conservative side.

10. The limiter

The final-stage processor sets the loudness ceiling and brings the master to its target level. Modern limiters look ahead 1–10 ms, detect peaks, and reduce gain transparently — the modern reference is FabFilter Pro-L 2.

10.1 Limiter algorithms (FabFilter Pro-L 2 styles)

  • Transparent — minimum coloration; clean at moderate gain reduction; reveals problems at high GR.
  • Punchy — preserves transient impact; slight harmonic generation.
  • Aggressive — more density; usable distortion at higher GR; pop/EDM mastering.
  • Modern — balanced; the new default in many chains.
  • Bus — designed for sub-master limiting before mastering.
  • Safe — more conservative attack; protects against artifacts.
  • All-Round — generic.
  • Magnetic — tape-like saturation in the limiter path.

10.2 True-peak limiting

After the digital limiter ceiling, D/A reconstruction or lossy encoding (MP3, AAC, Opus) can produce inter-sample peaks up to ~3 dB higher than the sample-domain peak. True-peak limiters oversample (4–16×) to catch these. True-peak ceiling of -1 dBTP is the modern recommendation, -2 dBTP for safety with aggressive lossy encoding (Spotify Ogg Vorbis q5, YouTube Opus).

10.3 Reference limiters

  • FabFilter Pro-L 2 — 8 styles, oversampling up to 32×, true peak, lookahead 0–20 ms; the transparent reference.
  • iZotope Ozone 11 Maximizer (IRC IV) — Intelligent Release Control; Modern, Transient, Soft modes; bundled in Ozone mastering suite.
  • Waves L2 / L3 Ultramaximizer / L3-LL / L3-16 — Michael Gerzon’s L1 (1994) lineage; the original look-ahead limiter; the L2 is the classic.
  • Sonnox Oxford Limiter v3 — Sony Oxford lineage; “Enhance” knob adds harmonic interest.
  • Newfangled Audio Elevate — Eventide-distributed multiband limiter with intelligent transient retention via 26-band spectral analysis.
  • DMG Limitless — extreme tweakability and oversampling.
  • DMG Compassion — compressor + limiter with mastering features.
  • Sonible smart:limit (2023) — AI-assisted limiter; suggests settings for a target genre/platform.
  • Maselec MPL-2 / MPL-1 — hardware mastering limiter (rare).
  • Weiss DS1-MK3, MM-1 — Daniel Weiss’s mastering limiter / multifunction.

10.4 Limiter tuning

Standard process: set ceiling (-1 dBTP), pull threshold down until target LUFS is reached, then back off 0.5–1 dB and listen for transient damage. Modern targets are far more conservative than the 2005–2015 “loudness wars” era when masters often hit -6 to -8 LUFS — streaming normalization removed the incentive.

11. Loudness targets — LUFS and the streaming era

LUFS (Loudness Units relative to Full Scale) was defined in ITU-R BS.1770 (2006, revised through BS.1770-5 in 2023) to model perceived loudness — K-weighted (a tilted frequency curve favoring 2–4 kHz where ears are sensitive), gated (silent passages excluded), and integrated over the whole program. LUFS replaced peak metering as the loudness measurement.

11.1 Platform targets (2026)

DestinationIntegrated LUFS targetTrue peak ceilingNotes
Spotify (normalization on, default)-14 LUFS-1 dBTPLoud setting -11, Quiet -19; user-selectable
Apple Music (Sound Check)-16 LUFS-1 dBTPMastered for iTunes (MFiT) badge requires LUFS-aware master
YouTube-14 LUFS-1 dBTPAuto-normalization since 2015
Tidal-14 LUFS-1 dBTPTidal Connect maintains; MQA being phased out
Amazon Music-14 LUFS-2 dBTPHD and Ultra HD streams
Deezer-15 LUFS-1 dBTP
SoundCloud-14 LUFS-1 dBTP
Pandora-14 LUFS-1 dBTP
EBU R128 (EU broadcast TV/radio)-23 LUFS ±1-1 dBTPHard regulatory floor in EU
ATSC A/85 (US broadcast)-24 LKFS-2 dBTPLKFS = LUFS in different ITU-R revision
Netflix (stereo content)-27 LKFS dialog-anchored-2 dBTPNEW: -18 LKFS for ATMOS/IMAX Enhanced
Cinema (Dolby)-23 to -27 LUFS-2 dBTPReference monitoring at 85 dB SPL
Audiobook (ACX, Audible)-23 to -18 LUFS, peak ≤ -3 dBFS, noise floor ≤ -60 dBFS—Strict spec
Vinyl (LP)No standard; engineer judgment—See §13

11.2 The end of the loudness wars

From the late 1990s to mid-2010s, mastering engineers pushed masters louder and louder, knowing louder masters sounded “bigger” on radio comparison. Records like Metallica’s Death Magnetic (2008, -4 LUFS, clipping audible) and the Red Hot Chili Peppers’ Californication (1999, -8 LUFS) became the canonical case studies in over-limiting damage.

When Spotify (2014), YouTube (2015), Apple Music (2015), and Tidal (2014) all introduced loudness normalization, the loudness arms race ended overnight. A master hitting -8 LUFS gets attenuated 6 dB by Spotify; a master at -14 LUFS plays back at the same loudness with all its transients intact. Modern best practice: master to -14 to -10 LUFS-I for most pop, R&B, hip-hop; -16 to -12 LUFS for jazz, classical, acoustic; -18 to -14 LUFS for film score.

11.3 LRA — Loudness Range

LRA measures the dynamic range of a program in LU (Loudness Units). High LRA = wide dynamic range; low LRA = consistent loudness. Targets:

  • Pop / EDM: 4–7 LU
  • Rock / indie: 6–10 LU
  • Jazz / acoustic: 8–14 LU
  • Classical / film score: 10–25 LU
  • Audiobook: 5–9 LU

11.4 Metering tools

  • iZotope Insight 2 — multifunction meter: LUFS-S, LUFS-I, LUFS-M, LRA, True Peak, intersample, correlation, spectrum, vectorscope, surround/Atmos meters.
  • Nugen Audio VisLM-H 2 / MasterCheck Pro — broadcast-spec LUFS metering; MasterCheck previews streaming-codec’d playback.
  • Waves WLM Plus — LUFS metering with simple display.
  • Klanghelm VUMT Deluxe — VU meter + LUFS.
  • TC Electronic Clarity M / Loudness Radar — TC’s loudness-radar metering paradigm.
  • YouLean Loudness Meter 2 (free + Pro) — popular freemium LUFS meter.
  • DAW-internal: Pro Tools Loudness Meter, Logic Pro Loudness Meter (built-in since 10.7).

12. Dolby Atmos mastering

Dolby Atmos arrived in cinema in 2012 (with Pixar’s Brave); Apple Music Spatial Audio launched in 2021 making Atmos a major music format. By 2026 most major-label releases have an Atmos master alongside the stereo.

12.1 Atmos format

  • Up to 128 audio objects plus a 7.1.2 bed (7 ear-level channels + 1 LFE + 2 height channels).
  • Objects carry metadata (XYZ position over time); the renderer places them in the playback configuration (5.1, 7.1, 7.1.4, 9.1.6, headphone binaural).
  • Deliverable: ADM BWF (Audio Definition Model Broadcast WAV) — single 24-bit / 48 kHz multichannel WAV file with object metadata in the bext + axml chunks. Standardized in ITU-R BS.2076.

12.2 Atmos workflow

  1. Mix in DAW with Atmos panner (Pro Tools panner since v2021.6, Logic since 10.7 in 2021, Nuendo with NuendoLink, Studio One 6.5+).
  2. Route to Dolby Atmos Renderer (standalone Dolby app, or built into Logic 11 since 2024, Pro Tools Hybrid Engine).
  3. Monitor in 7.1.4 speaker array if available; binaural monitor on headphones via Dolby Renderer’s binaural mode.
  4. Render to ADM BWF for delivery.

12.3 Renderer settings

  • Loudness target: -18 LUFS dialog-anchored (Apple Music spec); Netflix targets -27 LKFS for Atmos.
  • Bed channel routing: how the bed gets placed when downmixing to stereo.
  • Trim controls per object: how an object behaves in 5.1 vs 7.1.4 vs stereo downmix.
  • Headphone binaural mode: Off (downmix to stereo), Near (head-locked elements pulled toward the listener), Mid, Far (elements pushed out).
  • Re-render: bake the Atmos session to specific output configurations (7.1, 5.1, stereo) for QC.

12.4 Atmos mastering chain

  • Mastered for Atmos is generally lighter touch than stereo mastering — the renderer’s binaural processing and the platform’s playback are part of the perceived sound, so the master shouldn’t be over-processed.
  • Per-object EQ + compression is rare; most engineers process at the bed/object output stage or use the renderer’s own EQ.
  • Pro Tools’ Hybrid Engine and Nuendo’s ADM Renderer are the production-side tools. Mastering studios offer Atmos mastering as a service; rooms typically have 7.1.4 or 9.1.6 monitoring.
  • Notable Atmos engineers: George Massenburg, Bob Ludwig, Alan Meyerson (Hans Zimmer’s mixer), Jonathan Wyner, Mandy Parnell.

13. Vinyl pre-mastering

Vinyl is a physical medium with hard constraints. A vinyl pre-master is a separate version of the master prepared for the cutting lathe; it differs from the digital master in several ways:

13.1 The constraints

  • Low-end mono: stereo low frequencies make the cutter head jump out of the groove. Bass below 100–150 Hz should be summed to mono (M/S, with HPF on the Side below the threshold).
  • Sibilance: high-frequency sibilance heats the cutter head and distorts; aggressive de-essing is required.
  • High-frequency energy above 16–18 kHz wastes groove width and adds to cutter heat without audible benefit; LPF or shelf cut.
  • Out-of-phase content in the high end can cause cutter damage; check Side channel above 5 kHz.
  • RIAA pre-emphasis is applied by the lathe automatically (bass cut, treble boost) and de-emphasized on playback; nothing for the engineer to do, but be aware of the curve.
  • Side length limits: longer sides = quieter cut (groove pitch must be wider to fit longer programs). Typical maxima:
    • 12” LP at 33⅓ RPM: 22 min/side for moderate loudness, up to 26 min with quality loss.
    • 12” single at 45 RPM: 12 min/side for loud club cut.
    • 7” single at 45 RPM: 4–5 min/side.
  • Spacing of loud passages: a loud cut near the end of a side (inner groove) has the worst tracking (smaller circumference = lower linear velocity = higher distortion). Sequence quiet songs last.

13.2 Vinyl-specific tools

  • Elysia Mpressor / Karacter — German hardware popular for vinyl mastering.
  • Cranesong HEDD (Harmonically Enhanced Digital Device) — adds tube/tape harmonics; used to prep masters that translate well to lacquer.
  • Sphere DLY-9000 — analog delay used in cutting.
  • Sonnox Oxford Inflator — quasi-tape transient effect.
  • iZotope RX 11 De-click + Mouth De-click — cleans up any artifacts that would print as audible pops.
  • HOFA IQ-Series MS-Pan + MS-Comp — M/S processors.
  • Reference manuals: Larry Boden, Vinyl: A History of the Analogue Record (2017); Eric T. Rocco’s mastering-for-vinyl tutorials.

13.3 Cutting houses

The lacquer/cutting engineers are distinct from the digital mastering engineers, though some (Bernie Grundman, Chris Bellman) do both:

  • Bernie Grundman Mastering (Hollywood) — Bernie Grundman, Chris Bellman; both do digital + lacquer.
  • Sterling Sound (NYC + Nashville) — Ted Jensen, Tom Coyne (until 2017), Greg Calbi, Joe LaPorta.
  • Gateway Mastering (Portland, ME) — Bob Ludwig (semi-retired 2023).
  • The Lodge (NYC) — Emily Lazar; cuts at various lacquer houses.
  • Black Saloon (Berlin) — Mandy Parnell.
  • Abbey Road Mastering (London) — Sean Magee, Frank Arkwright, Christian Wright.
  • Metropolis Mastering (London) — Stuart Hawkes, Tony Cousins, John Davis.
  • Battery Studios (London) — Rachel Stone.
  • GZ Media (Czech Republic) — largest vinyl pressing plant in the world; in-house mastering.
  • Optimal Media (Germany) — major EU pressing plant.

13.4 Half-speed mastering and direct metal mastering

  • Half-speed mastering: the master is played back at half speed and cut at half speed; the cutter head has more time to track high-frequency detail. Used by Abbey Road’s Miles Showell for audiophile reissues (Beatles, Pink Floyd).
  • Direct Metal Mastering (DMM): cuts directly into a copper-coated metal master instead of lacquer; sharper top end, more accurate transients; some find it cold. Used by Optimal Media and Pallas.

14. CD vs streaming masters

There’s a school that argues for separate masters per format; another that argues a single 24-bit master suffices and platforms downsample/down-convert as needed.

14.1 The case for separate masters

  • CD master: 16-bit / 44.1 kHz, dithered, can run hotter (no streaming normalization).
  • Streaming master: 24-bit / 44.1 kHz, mastered to -14 LUFS-I, ceiling -1 dBTP.
  • MFiT / Apple Digital Masters master: 24-bit, headroom ≥1 dB, no clipping including inter-sample, includes album-cohesion checking.
  • Vinyl pre-master: separate file with constraints in §13.

14.2 The case for a single master

Most modern engineers deliver:

  • Master 24-bit / 44.1 kHz WAV, -14 LUFS-I, -1 dBTP for digital distribution.
  • DDP image for CD (16-bit, dithered, with PQ codes, ISRCs, CD-Text).
  • Separate vinyl pre-master if vinyl is in the release plan.
  • Atmos ADM BWF if Atmos is in the plan.

Streaming services accept the 24-bit master and downsample / re-encode as needed; for most program material the single-master approach is fine. Niche audiophile labels (Reference Recordings, ECM, Channel Classics) deliver separate hi-res (88.2 / 96 / 192 kHz) masters.

14.3 MFiT / Apple Digital Masters

Apple’s Mastered for iTunes (MFiT, since 2012) was renamed Apple Digital Masters in 2018. The spec:

  • Source: 24-bit at native session rate (44.1, 48, 88.2, 96).
  • Headroom: ≥ 1 dB below 0 dBFS (so AAC encoding has room).
  • No clipping (sample or inter-sample) when encoded to AAC.
  • The Apple-provided AURoundTripAAC tool simulates the final AAC encoding so the engineer can audit.

Most major labels deliver MFiT/ADM-compliant masters as default.

15. DDP — Disc Description Protocol

DDP is the standard delivery format for CD replication. A DDP image is a folder containing:

  • The audio data (16-bit / 44.1 kHz, dithered from 24-bit).
  • PQ subcode (track marks, indices, pauses).
  • ISRC codes per track (International Standard Recording Codes).
  • CD-Text (artist + track names).
  • A descriptor file referencing the audio and metadata.

Tools: Sonoris DDP Creator, HOFA CD-Burn & DDP, Sequoia (Magix), MAGIX Sound Forge Pro with CD project. The DDP image is uploaded to the replication plant.

16. Mastering engineers — signature sounds

The engineer’s name on a record influences sound as much as the gear. The canonical list:

16.1 Bob Ludwig

Gateway Mastering, Portland, ME. Started at Sterling Sound 1976, then Masterdisk, then founded Gateway 1992. 13 Grammy wins; mastered Led Zeppelin’s In Through the Out Door, Bruce Springsteen’s Born in the U.S.A., Daft Punk Random Access Memories (which won Album of the Year 2014), Beyoncé’s catalog, Tool’s Lateralus, 10,000 Days, Fear Inoculum, Radiohead. Style: open, dimensional, never over-limited. Semi-retired 2023; passed his Gateway clients to colleagues including Adam Ayan and Andrew Mendelson.

16.2 Bernie Grundman

Bernie Grundman Mastering, Hollywood + Tokyo. Started at A&M Mastering in 1968; founded BGM in 1984. Mastered Thriller (Michael Jackson, 1982), Steely Dan Aja (1977), Carole King Tapestry (1971), Dr. Dre The Chronic (1992), Kendrick Lamar good kid, m.A.A.d city (2012), Frank Ocean Blonde (2016). Cuts lacquers as well as digital. Style: warm, full, slightly mid-forward. Still actively mastering at 80+ in 2026.

16.3 Bob Katz

Digital Domain, Orlando, FL. Author of Mastering Audio: The Art and the Science (3rd ed. 2014), the standard textbook. Coined K-system metering (K-20, K-14, K-12 — three monitoring/loudness targets). Less commercial pop than Ludwig or Grundman; more jazz, classical, audiophile. Style: dynamics-preserving, transparent, anti-loudness-wars from the 1990s onward.

16.4 Emily Lazar

The Lodge, NYC. First woman to win the Best Engineered Album (Non-Classical) Grammy, 2020 (for Beck’s Colors, 2017). Mastered Foo Fighters Wasting Light, Vampire Weekend Modern Vampires of the City, Father of the Bride, Sia, Coldplay Music of the Spheres. Style: modern pop polish, careful dynamics.

16.5 Mandy Parnell

Black Saloon, Berlin (relocated from London ~2017). Mastered Björk’s catalog from Vespertine (2001) onward, The xx, Sigur Rós Valtari, Aphex Twin Syro. Style: detailed, ambient-aware, careful with stereo image; one of the leading electronic and experimental mastering engineers.

16.6 Greg Calbi

Sterling Sound, NJ. Started 1976. Mastered John Lennon Imagine, David Bowie Heroes, Bruce Springsteen Born to Run, The River, U2 catalog, The National, Bon Iver For Emma, Forever Ago. Style: rock-oriented, dynamic, “open.”

16.7 Ted Jensen

Sterling Sound. Mastered Norah Jones Come Away with Me (2002), Green Day American Idiot (2004), Lady Gaga The Fame (2008), Bruno Mars catalog, Eagles Hotel California (1976 original; remastered 2013), 2018–2024 Taylor Swift re-recordings (Fearless (Taylor’s Version), Red (Taylor’s Version), Speak Now (Taylor’s Version)).

16.8 Joe LaPorta

Sterling Sound NYC and Nashville. Mastered David Bowie Blackstar (2016), Vampire Weekend, Imagine Dragons, Death Cab for Cutie. Modern rock + indie standard.

16.9 Tom Coyne (1955–2017)

Sterling Sound. Mastered Adele 21 + 25, Taylor Swift 1989, Beyoncé Lemonade. Modern pop reference. Died of cancer 2017; his protégé Randy Merrill took over many Sterling clients.

16.10 Randy Merrill

Sterling Sound. Took over Tom Coyne’s client list. Mastered Lady Gaga Joanne, Chromatica, Harry Styles Fine Line, Harry’s House, Taylor Swift Lover, folklore, evermore, Midnights, The Tortured Poets Department.

16.11 Chris Gehringer

Sterling Sound. Mastered hip-hop and pop: Drake, The Weeknd, Beyoncé, Rihanna, Kanye, Travis Scott, Bad Bunny. Style: loud-but-tasteful modern hip-hop.

16.12 Howie Weinberg

Howie Weinberg Mastering, LA. Founded Masterdisk’s heyday with Bob Ludwig + Bob Katz. Mastered Nirvana Nevermind (1991), Beastie Boys Licensed to Ill + Paul’s Boutique + Check Your Head, Soundgarden Superunknown, R.E.M. Automatic for the People, Smashing Pumpkins Siamese Dream, Mellon Collie, Jeff Buckley Grace. Style: 90s alternative rock canon.

16.13 Stephen Marcussen

Marcussen Mastering, Hollywood. Mastered Tom Petty catalog, Stevie Nicks, Roy Orbison, John Mayer.

16.14 Doug Sax (1936–2015)

The Mastering Lab, Hollywood. Mastered Pink Floyd Dark Side of the Moon + Wish You Were Here, The Eagles, Earth Wind & Fire, Ray Charles. Defined classic American mastering aesthetic.

17. AI mastering — the 2024–2026 landscape

AI mastering services have moved from novelty to commodity. Most are competent for a baseline master; the leading services pair AI defaults with human review.

  • LANDR (since 2014) — original AI mastering service; cloud-based; tiered subscription. Trains on millions of mastered tracks. Now offers stem mastering and reference-track matching.
  • eMastered — co-founded by Grammy engineer Brian Lee; pairs AI with reference-track matching; targets specific genres.
  • CloudBounce — competitor; bought by BandLab 2020; now BandLab Mastering, integrated into BandLab DAW.
  • iZotope Ozone 11 Mastering Assistant (2023; 11.x ongoing) — analyzes the master and suggests EQ + dynamics + maximizer settings; not cloud — runs locally in Ozone 11.
  • Sonible smart:limit / smart:EQ 4 / smart:comp 2 — AI-assist for individual mastering processors.
  • Mastering.Studio (formerly Cryo Mix) — AI mastering with custom presets.
  • AI Mastering by aimastering.com — free tier.

17.1 What AI mastering does well

  • Genre-typical defaults: loudness, basic EQ curve, limiter settings appropriate for the genre.
  • Reference-matching: feed it a track and have your master matched to that target curve.
  • Bulk catalog work: mastering 200 podcast episodes or YouTube videos where human review per-piece isn’t economical.

17.2 What it still doesn’t do

  • Genre-bending or experimental material: AI defaults are interpolations; novel material gets pulled toward the mean.
  • Album cohesion: cross-track decisions about sequencing, key changes, dynamic flow.
  • Mastering for vinyl, Atmos: format-specific judgments.
  • Listening fatigue check: humans listen 8 hours and notice when something is wrong.
  • The “ear” question: AI doesn’t know what the artist intended.

By 2026, the standard workflow at small studios is: AI mastering for demos / drafts; human mastering engineer for releases.

18. The QC and delivery stage

The final mastering step is quality control:

  • Critical listening pass at multiple loudnesses on multiple systems (mains, near-fields, headphones, laptop, phone speaker, car).
  • Click / pop detection: visual scan of waveform + spectral display for any artifacts; iZotope RX 11 De-click and Mouth De-click are go-to tools.
  • Reference comparison: A/B against three commercial tracks in the same genre.
  • Loudness compliance check: verify LUFS-I, True Peak, LRA targets per platform.
  • Format-specific QC: AAC encode preview for MFiT, downsample preview for streaming, Atmos binaural preview, vinyl pre-master cutter-head simulation.
  • Metadata: ISRC codes, embedded BWF metadata, ID3 tags for MP3 delivery, album artwork attachment.

Deliverables typically:

  • WAV 24-bit / 44.1 kHz master (digital distribution baseline).
  • WAV 24-bit / 48 kHz master (video sync).
  • DDP image for CD replication.
  • Vinyl pre-master WAV (if vinyl release planned).
  • ADM BWF Atmos master (if Atmos release).
  • MP3 320 kbps and AAC 256 kbps reference encodes for client approval.
  • Stems (vocal + music, or full track stems) for sync licensing.

Adjacent

Graph View