THE CONTROL ROOMthe studio of our dreams · learn recording, mixing & mastering

← Jam Corner
ATTICRECORDINGVUVUVUVUVUVUVUVU24 TRACK · 2"15 / 30 IPSMASTER · ¼"Dub LabDUB LABATTIC ROOTSTAPE ECHOPROGRAM EQVUOPTO LEVELERVUFET LIMITERVUVARI-MU2.4 sDIGITAL REVERBSPRINGHIGH PASSVUVUVUVUVUVUVUVUTB…break?ECHO CHAMBER· · ·PLATESOUNDS FROM THE ATTICSIGNAL FLOW ↓PATCH IT ↘DAW ↓TAPE ↓GEAR →MICS ↓EAR ↓
← JAM CORNER

a rich friend’s studio · he’s on tour, we borrowed the keys

THE CONTROL ROOM

We make our records in an attic, with a tape machine and a lot of love. Rooms like this one we mostly know from record sleeves – which makes it the perfect place to learn: console, two-inch tape, a rack full of classic outboard, monitors in the wall – and a computer in the corner. Read the basics and the special knowledge, patch signals, mix a dub on the DAW and train your ears.

LESSONS

Thirteen chapters from the first decibel to dub. Start with the basics, then go deeper.

HOW TO USE THIS ROOM

1 · Read a chapter – start with the basics. 2 · Try it – every chapter links to a lab where you can hear and touch it. 3 · Check yourself in the quiz. Ten minutes a day beat one long evening.

BASICS
Sound, hearing & decibelsBASICS

In a nutshell Decibels compare, they don’t count – and our ears hear differently at different volumes.

Sound is a fast change in air pressure. Its frequency (Hz) is the pitch, its amplitude the level. Young ears hear roughly 20 Hz to 20 kHz; the top end drops with age.

  • Decibels are ratios, not absolute amounts. +6 dB = twice the voltage or pressure, +3 dB = twice the power. About +10 dB is usually felt as “twice as loud” – a rule of thumb, not physics.
  • dB SPL measures sound pressure: 0 dB SPL = 20 µPa, roughly the quietest sound young ears can hear.
  • Our ears are not flat: at low listening levels we hear less bass and less extreme treble (the equal-loudness contours). That is why a mix balanced very loud sounds thin when you turn it down – and the other way round.
  • Polarity vs phase: flipping polarity turns the waveform upside down. Phase is about time: two mics at different distances get the same sound slightly delayed, some frequencies cancel and you get comb filtering – hollow, “phasey”.
Go deeper – tricks & background
  • Wavelength = speed of sound ÷ frequency. At 20 °C sound travels about 343 m/s: 100 Hz is about 3.4 m long, 1 kHz about 34 cm, 10 kHz about 3.4 cm (calc.). That is why a few centimetres of foam do nothing for the bass – to a 141 Hz wave a few inches of foam are “essentially invisible” (Recording Magazine).
  • Room modes: between two parallel surfaces the first axial mode sits at f = speed of sound ÷ (2 × distance). An 8-ft (2.44 m) ceiling gives about 70.6 Hz, with harmonics at about 141, 212 and 282 Hz. That note booms or vanishes depending on where you sit. Avoid perfectly cubic rooms – their modes pile up on the same frequencies.
  • Comb-filter maths: if a copy arrives t seconds late, the first cancellation lies at 1 ÷ (2t), then every 1 ÷ t above it. A mic 34 cm further away is about 1 ms late: notches at about 500 Hz, 1.5 kHz, 2.5 kHz … (calc.). Move a mic by a few centimetres and the whole comb moves – that is why tiny position changes sound so different.
  • Adding levels: two equal but unrelated signals (two singers) add up to about +3 dB; two identical, in-phase signals (a copied track) to +6 dB; identical with opposite polarity to silence. A copied track does not sound like two players – it is just 6 dB louder.

(calc.) = calculated from the formula given, rounded.

Try it: frequency ear training · phase demo in the DAW lab

Levels & gain stagingBASICS

In a nutshell Every stage wants a healthy level: well above its noise, safely below distortion.

Every device in the chain has a sweet spot: loud enough to stay well above its noise, quiet enough not to distort. Setting each stage right is called gain staging.

LEVELWHAT
mic leveltiny voltages straight from a microphone – needs a preamp
instrument (Hi-Z)guitar or bass pickups – high impedance, best into a DI box or Hi-Z input
line +4 dBupro gear: 1.228 V
line −10 dBVconsumer/semi-pro gear: 0.316 V – about 11.8 dB lower than +4 dBu
speaker levelafter the power amp – never into a line input
  • dBu: 0 dBu = 0.775 V. dBV: 0 dBV = 1 V. dBFS: digital full scale – 0 dBFS is the ceiling, nothing gets louder, it clips.
  • A VU meter (1939, standardised 1942) is slow – 300 ms – and shows something like the average level, not the peaks. 0 VU is usually calibrated to +4 dBu.
  • Analog meets digital: EBU R68 aligns 0 dBu with −18 dBFS; SMPTE RP155 aligns +4 dBu with −20 dBFS. That leaves 18 dB (EBU) or 20 dB (SMPTE) of headroom above the alignment level.
  • Practice: when recording at 24 bit, aim for peaks somewhere around −10 dBFS. There is no prize for recording hot – the noise floor of 24 bit is far below anything you will hear.
Go deeper – tricks & background
  • Where the converter clips, in analog terms: with EBU R68 (0 dBu = −18 dBFS) 0 dBFS equals +18 dBu; with SMPTE RP155 (+4 dBu = −20 dBFS) it equals +24 dBu (calc.). So the same hot desk output can be fine on one interface and clip on another – check the input sensitivity.
  • Every 6 dB lower costs about one bit (≈ 6.02 dB per bit). At 24 bit you can easily afford recording 12–18 dB under full scale; at 16 bit it hurts more.
  • Calibrating monitors, K-system style (Bob Katz): pink noise at −20 dBFS RMS, one speaker at a time, set to 83 dB SPL on a meter set to C-weighting, slow. Check: 6 dB less noise should read 77 dB SPL. On a K-14 meter, a true K-14 mix has its loud (forte) passages around the meter’s zero.

(calc.) = calculated from the formula given, rounded.

Try it: DAW lab

MicrophonesBASICS

In a nutshell Pick the pattern for the job – and placement often matters more than the model.

  • Dynamic (moving coil): robust, handles loud sources, needs no power. Classic: the Shure SM57 (1965) on snare and guitar amps.
  • Condenser: more detail and treble, needs power – phantom power (P48: 48 V, standard IEC 61938) or, for tube mics, its own power supply. Classics: Neumann U47, U67, U87.
  • Ribbon: a thin metal ribbon between magnets – warm, smooth, usually figure-8. Handle with care; never plug or patch with phantom power switched on, vintage ribbons can be damaged.

Polar patterns and where they are deaf (the null): cardioid at 180°, supercardioid at about 125°, hypercardioid at about 110°, figure-8 at 90° to the sides. Omni hears everything.

  • Proximity effect: directional mics get more bass the closer you get – crooners and radio voices use it on purpose.
  • 3:1 rule: the distance between two mics should be at least three times the distance from each mic to its source. Then the spill is at least 9 dB down and comb filtering stays small.

Stereo techniques: XY – two cardioids, capsules on top of each other, angled (often 90°); ORTF – two cardioids 17 cm apart at 110°; AB – two spaced omnis; Blumlein – two figure-8s at 90°; Mid-Side – a forward mic (M) plus a sideways figure-8 (S), decoded as L = M + S and R = M − S. In mono the S part cancels, so M/S is mono-safe.

Go deeper – tricks & background
  • Proximity effect in detail: only pressure-gradient (directional) mics have it. Omnis (pure pressure mics) show none; a figure-8 shows the most – a cardioid about half as much (Sound On Sound). Remedies: more distance, the mic’s high-pass switch, singing “across” the mic, or angling it on a guitar cab.
  • Ribbons & phantom – the real danger: according to Royer, damage comes from miswired cables and from patching at the bay with 48 V on – the brief short while a jack goes in can send a voltage spike that stretches the ribbon. Rule: unplug the mic or switch phantom off, then patch.
  • Why the stereo techniques sound different: XY (capsules together) works with level differences only, so it folds to mono without comb filtering. AB (spaced omnis) relies mainly on time differences – wide and airy, less mono-safe. ORTF (French radio, around 1960: 17 cm, 110°) combines both.
  • Time alignment: two mics on one source at different distances – 1 ms ≈ 34 cm. Shift the closer mic’s track later by the distance difference and the low end comes back (calc.; fine-tune by ear).

(calc.) = calculated from the formula given, rounded.

Try it: quiz

The console & signal flowBASICS

In a nutshell Inserts process the whole signal, sends share a copy – and pre or post decides whether the fader matters.

A classic channel strip, from top to bottom:

MIC PRE → HIGH-PASS → EQ → DYNAMICS → INSERT → FADER → PAN → BUSES / STEREO MIX

The order of EQ and dynamics is not carved in stone – some desks (the SSL 4000, for example) let you switch it.

  • Insert: the whole signal leaves the channel (insert send), goes through an outboard unit and comes back (insert return). Serial – for things that should change the entire signal: EQ, compressor, gate, de-esser.
  • Aux send: a copy of the signal, adjustable per channel, goes to an effect or a headphone mix and comes back on an effect return. Parallel – for effects many channels share: reverb, delay.
  • Pre or post fader? Pre-fader sends ignore the channel fader – right for headphone mixes, so the musician’s mix does not change when you mix. Post-fader sends follow the fader – right for reverb and delay, so the effect fades with the instrument.
  • Group / subgroup: several channels are summed to a bus, so you can process them together (e.g. one compressor on all drums). A VCA group carries no audio at all – it remote-controls the faders (and their post-fader sends).
  • Direct out: one channel straight to one tape track. Monitor path and talkback: what you hear in the control room is not always what goes to tape – and the TB button lets you talk to the live room.
Go deeper – tricks & background
  • “Effect only”: set an aux send to pre-fader and pull the channel fader down – the dry signal is gone, only the reverb or echo remains. A favourite for ghostly dub and ambient moments.
  • Parallel compression on the desk: send the drums to a spare bus or aux, compress that copy hard (typically 4:1 to 10:1, 6–10 dB of gain reduction) and blend it in under the untouched drums (Slate Digital).
  • Key / side-chain input: patch another signal into the compressor’s detector – e.g. the kick, so the bass ducks. On a drum-bus compressor, a high-pass in the side-chain makes it less sensitive to the kick, so the kick keeps its weight (Mike Senior, Sound On Sound).

Try it: routing lab

EQBASICS

In a nutshell Cut what is in the way before you boost what is missing – and judge EQ in the mix.

  • Shelf lifts or lowers everything above (high shelf) or below (low shelf) a frequency. Bell / peak works around a centre frequency; Q sets the width – high Q = narrow. High-pass / low-pass filters remove everything below / above, with slopes of 6, 12, 18 or 24 dB per octave.
  • Rough map (boundaries differ from source to source): sub 20–60 Hz · bass 60–250 Hz · low mids 250–500 Hz (where “mud” builds up) · mids 500 Hz–2 kHz · upper mids 2–4 kHz · presence around 5–6 kHz · brilliance and “air” above 6 kHz.
  • Habits that work: high-pass everything that has no business in the low end; cut before you boost; cut narrow, boost wide; and judge EQ in the mix, not in solo.
  • The Pultec trick: on the passive Pultec EQP-1A you can boost and cut the same low frequency at once. Because the two curves differ, they do not cancel – you get a fat low boost with a dip just above it.
  • Minimum vs linear phase: every analog EQ (and most digital ones) shifts phase around the frequencies it touches. Linear-phase EQs avoid that, but cost latency and can cause pre-ringing before transients.
Go deeper – tricks & background
  • Q in octaves: Q 1.41 ≈ one octave wide, Q 2.87 ≈ half an octave, Q 4.32 ≈ a third of an octave (calc.: Q = √2ᴺ ÷ (2ᴺ − 1) for N octaves). Handy when a manual talks octaves and the plug-in shows Q – or the other way round.
  • Why the Pultec trick works: “the Boost control has slightly higher gain than the Attenuation control has cut, and the frequencies they affect are slightly different” (Universal Audio). So they never cancel – you tune their interaction: more bottom without mud on the mix bus, punch on a kick while the cut keeps the very low end in check.
  • When linear phase: where phase shifts would hurt – parallel paths, processing one side of a stereo or M/S pair, mastering. Its price: latency and pre-ringing before transients, so on drums a normal (minimum-phase) EQ is often the better choice (FabFilter).
  • EQ before or after the compressor? Before: you change what the compressor reacts to (boost the lows and it clamps down harder on them). After: you shape the already compressed sound.

(calc.) = calculated from the formula given, rounded.

Try it: DAW lab · frequency ear training

INTERMEDIATE
The patchbayINTERMEDIATE

In a nutshell Usually outputs on top, inputs below – a normalled connection is broken as soon as you plug into the input.

The patchbay is the studio’s switchboard: every input and output of the console, the tape machine and the outboard ends up on small jacks (TT/bantam or 1/4-inch), usually outputs on top, inputs below. With short cables you re-route anything without crawling behind the desk.

  • Normalled: top and bottom are connected inside while nothing is plugged in (e.g. a channel’s insert send to its own insert return). Plugging into either jack breaks the connection.
  • Half-normalled: plugging into the top jack only takes a copy – the normal connection stays. Plugging into the bottom jack breaks it. The most common setup for outputs you want to “listen in on”.
  • Thru / open: no internal connection – nothing happens until you patch.
  • Mult: a row of jacks wired together, to split one signal into several.
  • Tie lines: lines from the wall boxes in the live room to the patchbay – this is how a mic reaches the right preamp.

The routing lab below lets you try it.

Go deeper – tricks & background
  • Parallel processing without a plug-in: if the insert send is half-normalled, a cable from that top jack only takes a copy – the channel keeps running. Patch the copy through a compressor into a spare channel and you have parallel compression.
  • Mults for double processing: split a snare into two or three channels and treat each differently – one bright, one squashed, one into the spring.
  • Troubleshooting: a channel suddenly silent? Look for a forgotten cable in a normalled bottom jack – it breaks the connection even when the other end hangs loose.
  • Mic lines and 48 V: never cross-patch mic tie lines with phantom power on – the brief short as the plug goes in can hurt ribbons (Royer).

Try it: routing lab

Tape & the multitrackINTERMEDIATE

In a nutshell Few tracks forced decisions – and every bounce added a little noise.

  • Speeds: 7½, 15 and 30 inches per second (ips). Faster tape gives a better high end and usually more dynamic range; the low-frequency “head bump” shifts with speed – some engineers even prefer the lows at 15 ips over 30.
  • Tracks: Abbey Road’s Studer J37 (1964) had four tracks on 1-inch tape. Sgt. Pepper was made on two of them, with lots of bouncing from one machine to the other – several tracks mixed down onto one free track to make room. Every bounce adds a generation of noise. By the end of the sixties 8- and 16-track machines arrived; in the seventies 24 tracks on 2-inch tape became the standard (Ampex MM1200 1976, Studer A800 1978).
  • Noise reduction: Dolby A (1965) splits the signal into four bands, compresses on recording and expands on playback – roughly 10 dB less hiss.
  • Tape saturation: driven harder, tape squashes peaks softly and adds harmonics – one reason analog recordings feel “glued”. Too hot and it smears transients and loses treble.
  • Small is beautiful: Lee “Scratch” Perry made the Black Ark records on a TEAC 3340 – a ¼-inch four-track.
Go deeper – tricks & background
  • Wavelength on tape = tape speed ÷ frequency. At 15 ips (38.1 cm/s) a 10 kHz tone is about 38 µm long on the tape, at 7½ ips only about 19 µm (calc.). The shorter the wavelength, the harder it is to record – slow tape loses treble first.
  • Head bump is a resonant peak in the low end of the tape playback; tape emulations let you set its gain and shape (iZotope Vintage Tape: up to +10 dB).
  • Reference level: machines are aligned with a test tape to a reference fluxivity in nanowebers per metre (nWb/m). Quarter-track decks like about 200 nWb/m; the old German “320 nWb/m” reference measured about 290 nWb/m with the American method – about 1 dB lower (reeltoreel.nl).
  • NAB vs IEC: America and Europe used different tape equalisation curves – a tape recorded with one sounds tonally off on a machine set to the other. Check before you transfer old reels.

(calc.) = calculated from the formula given, rounded.

Try it: quiz

Dynamics: compressor, limiter, gateINTERMEDIATE

In a nutshell A compressor turns loud parts down; attack and release decide how it feels.

  • Threshold: above this level the compressor starts working. Ratio: 4:1 means a signal 8 dB over the threshold comes out only 2 dB over. Attack: how fast it grabs. Release: how fast it lets go. Knee: hard or gentle onset. Make-up gain: brings the level back up.
  • Limiter: a compressor with a very high ratio (by convention about 10:1 and up) – a ceiling.
  • Gate / expander: the opposite – they turn down what falls below the threshold (spill between drum hits, noise).
  • De-esser: compresses only the sibilance – roughly 5–8 kHz on many female voices, 3–6 kHz on many male voices.
  • Sidechain: the compressor listens to a different signal than the one it processes – e.g. the bass ducks every time the kick hits.
  • Parallel (“New York”) compression: blend a heavily compressed copy under the dry signal. Quiet details come up, the transients stay intact.

Four legends, four principles: the LA-2A (1962) uses light and a photocell (T4) – slow, smooth, two-stage release. The 1176 (1967) uses a FET – very fast: attack 20–800 µs, release 50 ms–1.1 s, and both are fastest fully clockwise; there is no threshold knob – you drive the input. The Fairchild 660/670 is a vari-mu tube limiter (the 670 has 20 tubes). The SSL bus compressor “glues” mixes with ratios of 2:1, 4:1 and 10:1.

Go deeper – tricks & background
  • Gain-reduction maths: a signal X dB over the threshold comes out X ÷ ratio over it, so the gain reduction is X − X ÷ ratio. 12 dB over at 4:1 → 3 dB over → 9 dB of gain reduction (calc.).
  • 1176 secrets (Universal Audio): in “all buttons” mode (also called “British” mode) “compression distortion increases radically”. With no ratio button pressed there is no compression at all – but the signal still passes through the circuit, so you get its colour. The classic starting point (often called the “Dr. Pepper” setting): attack at 10 o’clock, release at 2 o’clock, ratio 4:1.
  • Attack decides the punch: a slower attack lets the first transient through before the compressor grabs – more snap; a very fast attack flattens it.
  • Drum bus losing weight? Put a high-pass in the compressor’s side-chain: it reacts less to the kick and stops pulling it down (Sound On Sound).

(calc.) = calculated from the formula given, rounded.

Try it: DAW lab

Reverb, delay & modulationINTERMEDIATE

In a nutshell Reverb puts a sound into a room, delay repeats it – time both to the song.

  • Reverb parameters: pre-delay = the gap between the direct sound and the first reflection; decay / RT60 = the time it takes to fall by 60 dB; early reflections arrive within roughly the first 50–100 ms; diffusion = how smooth the tail is; damping = how fast the highs die away.
  • Where reverb came from: echo chambers – a speaker and a mic in a hard, reflective room (Capitol Studios built theirs in 1956). Plates – the EMT 140 (1957), a large steel plate in a wooden box with an adjustable damper. Springs – in guitar amps and dub mixers. Digital – the EMT 250 (1976) was the first commercial digital reverb, the Lexicon 224 (1978) became the most widely used.
  • Delay in time: a quarter note lasts 60 000 ÷ BPM milliseconds. At 120 BPM that is 500 ms; a dotted eighth is ¾ of it (375 ms), an eighth half (250 ms).
  • Tape echo: Echoplex (from 1959, a moving record head sets the delay), Binson Echorec (a magnetic drum instead of tape), Roland RE-201 Space Echo (1974, three playback heads plus a spring). Each repeat gets darker and dirtier – the sound of dub.
  • Modulation: a flanger mixes in a copy delayed by about 1–5 ms, with feedback, and sweeps it; a chorus uses longer delays of about 20–30 ms; a phaser uses all-pass stages, so its notches are not evenly spaced. ADT (automatic double tracking) was invented at Abbey Road in 1966 by Ken Townsend – a second tape machine with a wobbling speed.
  • Haas / precedence effect: a copy arriving within a few up to roughly 30–40 ms is heard as part of the original, and the sound seems to come from the side that arrives first.
Go deeper – tricks & background
  • Delay table at 90 BPM (calc.: 60 000 ÷ BPM; dotted = × 1.5, triplet = ÷ 1.5): quarter 667 ms · dotted eighth 500 ms · eighth 333 ms · eighth triplet 222 ms · sixteenth 167 ms.
  • Haas widening – and its trap: Helmut Haas described the effect in 1949. Delay one side by 5–35 ms and a mono part sounds wide. But in mono it can “disappear or, at best, change in tone and level” through phase cancellation (Sound On Sound). With 10 ms the notches lie at about 50, 150, 250 Hz … (calc.). Mono-safe alternatives: reverb, chorus, real double-tracking, M/S widening.
  • The EMT 140 up close: a steel plate of about 2 × 3 m, the whole unit around 600 lb (about 270 kg); a driver in the middle sets it vibrating, pickups take the sound off, a damper sets the decay time (Universal Audio).
  • Why big rooms ring longer: Sabine’s formula RT60 ≈ 0.161 × V ÷ A (V = volume in m³, A = absorption in m²). Double the volume with the same absorption and the decay doubles.

(calc.) = calculated from the formula given, rounded.

Try it: DAW lab

MixingINTERMEDIATE

In a nutshell Balance first, effects later – and check in mono, on small speakers, at a moderate level.

  • Static mix first: faders and pan only, no plug-ins – if the balance works, half the job is done.
  • Panning & pan law: when you pan a mono signal to the centre, the console lowers it so it does not get louder than at the sides – by −3 dB (constant level for stereo listening), −6 dB (constant level in mono) or the compromise −4.5 dB.
  • Check in mono, and on small speakers – the little cubes on the meter bridge are there for a reason. Phase problems and a bass that only lives on big speakers show up immediately.
  • Listening level: moderate and consistent. Bob Katz calibrates each monitor to 83 dB SPL with pink noise in his K-system. Loud for a moment to check the low end is fine – loud all day ruins your ears and your judgement.
  • Reference tracks: compare with a record you love, level-matched – louder always seems better.
  • Automation: rides on vocals, mutes, effect throws – the mix becomes a performance (in dub literally).
  • For mastering: leave some headroom and no brickwall limiter on the mix bus unless the mastering engineer asks for it.
Go deeper – tricks & background
  • Moving sessions between DAWs? Different default pan laws (−3, −4.5 or −6 dB) shift the balance between centred and panned parts by 1.5–3 dB (calc.). Check the setting before you blame your ears.
  • Kick vs bass: they fight over the same frequencies. A compressor on the bass keyed from the kick ducks the bass for a moment on each hit (side-chain).
  • Your room lies in the bass: with an 8-ft ceiling around 70 Hz can boom or vanish (see “Sound”). Check the low end on headphones and at a second listening position before you EQ it.
  • Width that survives mono: prefer reverb, chorus, real double-tracking or M/S over Haas tricks – and press the mono button often.

(calc.) = calculated from the formula given, rounded.

Try it: DAW missions

PRO
Mastering & loudnessPRO

In a nutshell Streaming services turn loud masters down – the goal is balance, not maximum loudness.

Mastering is the last creative and technical check: tonal balance and level across a whole record, plus the delivery formats. Typical tools: EQ, compression, sometimes stereo work, and a limiter at the end.

  • LUFS (also LKFS, standard ITU-R BS.1770) measures loudness the way we hear it; 1 LU = 1 dB. Momentary = 400 ms window, short-term = 3 s, integrated = the whole track. LRA = loudness range.
  • Normalisation: streaming services turn loud tracks down. Spotify normalises to −14 LUFS (Premium users can choose “loud” −11 or “quiet” −19) and asks for true peaks below −1 dBTP – below −2 dBTP if your master is louder than −14 LUFS. Apple’s Sound Check works out at about −16 LUFS; YouTube is commonly measured at about −14 LUFS – neither company publishes an official figure. Broadcast: EBU R128 −23 LUFS, US ATSC A/85 −24 LKFS.
  • True peak: between the digital samples the real waveform can go higher than 0 dBFS (inter-sample peaks) – a true-peak limiter oversamples to catch them.
  • Dither: add it once, as the very last step, when you reduce bit depth (e.g. 24 → 16 bit). A 32-bit float export needs none.
  • Vinyl: keep the bass centred (below about 150 Hz), tame sibilance, avoid boosting above 10 kHz. Longer sides mean lower cutting level, and the inner grooves lose treble.
Go deeper – tricks & background
  • How “integrated” is measured: EBU R128 (based on ITU-R BS.1770) gates the measurement – everything below −70 LUFS (absolute gate) and everything more than 10 LU below the measured level (relative gate) is ignored. Silence and very quiet passages don’t drag the number down. Target −23 LUFS ±0.5 LU (±1 LU where exact normalisation isn’t practical), max −1 dBTP. First published August 2010, version 4.0 August 2020.
  • What normalisation does to a loud master: a track at −8 LUFS integrated is turned down by about 6 dB on Spotify’s normal setting (calc.) – just as loud as a −14 master, but with less dynamics. When people listen to a whole album, Spotify normalises per album, so the level differences between your songs stay as mastered.
  • Why dither at all: shortening the word length (24 → 16 bit) without dither – truncation – causes quantisation distortion that follows the signal. Dither replaces it with a very low, steady noise. Exactly once, as the last step (Sound On Sound).
  • Vinyl sequencing: towards the centre of the record the response is down about 3 dB at 15 kHz, so put quiet songs and ballads towards the inside; avoid ending a side with the loudest song. Bass below 150 Hz centred – phase problems in the bass can collapse the groove and make the needle skip (Kindercore vinyl guide).

(calc.) = calculated from the formula given, rounded.

Try it: DAW lab

Digital audio & the DAWPRO

In a nutshell The highest frequency is half the sample rate, bit depth sets the dynamic range – and latency per buffer = buffer ÷ sample rate.

  • Sample rate: how often per second the signal is measured. The highest frequency you can capture is half of it – the Nyquist frequency (22.05 kHz at 44.1 kHz). Anything above folds back down as aliasing.
  • Bit depth: each bit gives about 6 dB of dynamic range – 16 bit ≈ 96 dB, 24 bit ≈ 144 dB (in theory; real converters manage less).
  • 32-bit float: inside a DAW, tracks and buses can go over 0 dBFS without clipping – but the converters and fixed-point files cannot. Keep the master below 0.
  • Latency = buffer ÷ sample rate: 256 samples at 48 kHz ≈ 5.3 ms – and the round trip (in and out) is about double. Small buffers for recording, big buffers for mixing.
  • Plug-in delay compensation (PDC): plug-ins that need time (look-ahead limiters, linear-phase EQs) delay their track; the DAW delays all others to match, so everything stays in time.
  • Oversampling: distortion, saturation, clipping and limiting create new harmonics, some above Nyquist – they would fold back as aliasing. Running the plug-in at a higher internal rate avoids that, at the cost of CPU and a little latency. EQs and delays rarely need it.
  • Convolution vs algorithmic reverb: convolution plays back a recorded impulse response of a real room or a real plate; algorithmic reverb builds the space from delay lines and all-pass filters.
  • Freeze / bounce: render a track with its plug-ins to audio – saves CPU and commits decisions, like printing to tape.
Go deeper – tricks & background
  • Aliasing, worked example: saturate a 15 kHz tone and you get a 3rd harmonic at 45 kHz. At 44.1 kHz it folds back to 45 − 44.1 = 0.9 kHz – an unmusical tone far down in the mids. With 2× oversampling (88.2 kHz) it folds to 88.2 − 45 = 43.2 kHz instead – inaudible, and the filter on the way back down removes it (calc.).
  • Latency in practice: 256 samples at 44.1 kHz ≈ 5.8 ms per buffer; input plus output roughly doubles it. Record with a small buffer and few plug-ins, raise it for mixing (Ableton).
  • Linear-phase EQs and look-ahead limiters add latency – while tracking, bypass them or the DAW’s delay compensation will delay what the musician hears.

(calc.) = calculated from the formula given, rounded.

Try it: DAW lab

Dub: the mixing desk as an instrumentPRO

In a nutshell In dub the mixing desk is played like an instrument – mutes, sends and filters are the notes.

In Kingston in the late sixties and seventies, engineers began to take finished reggae tracks apart live on the desk – that became dub.

  • King Tubby bought a used MCI console from Byron Lee’s Dynamic Sounds in 1972 (some sources say 1971): 12 inputs, 4 outputs, two reverb returns – and a large stepped high-pass filter knob he called “the Big Knob”, used to sweep whole instruments.
  • Lee “Scratch” Perry’s Black Ark (1973–79) ran on a TEAC 3340 four-track, a Roland Space Echo, a Mu-Tron Bi-Phase and a Grampian spring reverb.
  • Channel One went 16-track in 1975 – the “rockers” sound with the Revolutionaries and Sly Dunbar on drums. Studio One’s riddims were versioned again and again.

Techniques you can try in the DAW lab: drops – mute instruments in and out with the band still playing; throws – send a single snare hit or a vocal phrase into the tape echo for one moment; feedback – turn up the echo feedback until the repeats build up (careful with your speakers); filter sweeps on the high-pass; and the spring – a kick to the spring tank crashes like thunder.

Go deeper – tricks & background
  • Not a one-man show: Tubby was always assisted by a team of engineers – Pat Kelly, Philip Smart, Prince Jammy and Scientist (Ableton). His MCI desk from Dynamic Sounds stood in the front room at 18 Dromilly Avenue.
  • The Black Ark started small: a TEAC four-track, an Alice mixing desk, a Grampian spring reverb and an Echoplex; the Roland Space Echo and a Mu-Tron phaser came later (Ableton).
  • Build a dub echo in the DAW: echo on an aux, time from the tempo (dotted eighths at 90 BPM = 500 ms, see “Reverb, delay”), a high-pass and some saturation on the echo return so each repeat gets thinner and dirtier, then automate the send for single throws.
  • Ghost parts: send the guitar skank pre-fader to the echo and pull its fader down – only the echo is left.

Try it: DAW missions

THE GEAR

The classics in this room and their stories. Where sources disagree, there is a little flag.

CONSOLES

1970 · Rupert Neve

Neve 1073

Mic preamp and EQ module, first built for the Wessex A88 console. High-pass at 50/80/160/300 Hz, low shelf at 35/60/110/220 Hz, six mid frequencies from 360 Hz to 7.2 kHz, and a fixed high shelf at 12 kHz. Its transformers are a big part of the famous thick sound.

1969 · Dick Swettenham

Helios

Designed by Olympic’s technical director. The first desk went into Olympic Studio Two in 1969; Island’s Basing Street studio got a 20-input Helios the same year – Bob Marley & the Wailers, Led Zeppelin and Black Sabbath recorded there. The Rolling Stones Mobile had one too.

1968 · Abbey Road

EMI TG12345

Abbey Road’s first transistor desk, with a limiter/compressor on every channel – new at the time. The valve REDD desks before it had reached their limits with 8-track recording. Abbey Road (1969) was largely recorded and mixed on it.

late 1960s · Saul Walker

API 550A & 2500

API was founded in 1969. The 550A is a three-band EQ with seven frequencies per band and up to 12 dB of boost or cut; its “proportional Q” gets narrower the more you boost. The 2500 bus compressor switches between “old” feedback and “new” feed-forward detection.

⚑ sources differ Year of the 2500 not verified.

1979 · Solid State Logic

SSL 4000 E

The E series added Total Recall (settings stored on floppy disk) and a master bus compressor – ratios 2:1, 4:1 and 10:1 – that became the “glue” of countless mixes. The G series followed in 1987.

⚑ sources differ The bus compressor’s release values are documented from an emulation, not from SSL itself.

TAPE

1964 · 4 tracks on 1-inch tape

Studer J37

Abbey Road eventually owned eight of them. Sgt. Pepper was made on two J37s, with tracks bounced from one machine to the other.

A80 c. 1970 · A800 1978

Studer A80 · A800

The A800 (1978) was Studer’s first microprocessor-controlled multitrack: up to 24 tracks on 2-inch tape at 15 or 30 ips. Its predecessor, the A80, is the classic workhorse of the seventies.

⚑ sources differ A80 introduction: 1970 or 1973 depending on the source.

350 1953 · MM1200 1976 · M79 1973

Ampex & 3M

The Ampex 350 was a two-speed ¼-inch recorder; the MM1200 came in 8-, 16- and 24-track versions. The 3M M79 used a closed-loop “IsoLoop” capstan drive and variable speed from 6 to 36 ips.

1965 · noise reduction

Dolby A

Splits the audio into four bands, boosts quiet signals in each band by up to 10 dB while recording and turns them back down on playback – loud signals pass untouched. Result: about 10 dB less tape hiss, up to 15 dB at 15 kHz.

EQ & DYNAMICS

1950s · Pulse Techniques

Pultec EQP-1A

A passive EQ with a tube make-up stage. Low band at 20/30/60/100 Hz, high boost at 3/4/5/8/10/12/16 kHz, high cut at 5/10/20 kHz. The famous trick: boost and cut the same low frequency at once – the curves differ, so you get a big low boost with a dip just above it.

⚑ sources differ Year of the original EQP-1 disputed: 1951 or 1956.

1962 · James F. Lawrence II

Teletronix LA-2A

An optical leveler: a light panel shines on a photoresistor (the T4 cell). Attack about 10 ms; release about 60 ms for the first half, then 1–15 seconds for the rest – that is why it sounds so smooth. Controls: Gain, Peak Reduction and a Compress/Limit switch.

1967 · Bill Putnam

UREI 1176

A FET limiter. Attack 20–800 µs, release 50 ms–1.1 s – both fastest fully clockwise. Ratios 4, 8, 12 and 20:1; no threshold: the input knob decides how hard it works. Press all four ratio buttons and it turns into a distorted, pumping beast – the “all buttons in” mode.

660 from 1959 · Rein Narma

Fairchild 660 / 670

A variable-mu tube limiter; the stereo 670 has 20 tubes, 11 transformers and weighs about 30 kg. Abbey Road bought a dozen 660s and used them on Beatles vocals and, from 1966, on the drums.

ECHO & REVERB

Echoplex 1959 · Echorec 1960s · Space Echo 1974

Tape echoes

Echoplex: tape in a cartridge, a moving record head sets the delay (EP-3 from 1970). Binson Echorec: a magnetic drum with four playback heads – Pink Floyd and the Shadows. Roland RE-201 Space Echo: a tape loop, one record and three playback heads, a 12-position mode selector and a spring reverb – a dub essential.

⚑ sources differ Exact Echorec year not verified.

1957 · EMT

EMT 140 plate

A thin steel plate of roughly 1 × 2 m in a frame about 2.4 m long, weighing around 270 kg. A driver shakes the plate, pickups take the reverb off it; a damper of fibre board moved closer to the plate shortens the decay. Stereo version from 1961.

Capitol Studios 1956

Echo chamber

A speaker plays into a hard, reflective room and a microphone picks it up again – fed from a send, back on a return. Capitol’s chambers (about 2000 cubic feet each, lacquered concrete, no parallel walls) gave up to five seconds of reverb. CBS engineers added pre-delay by putting a 15-ips tape machine in front of the chamber.

Hammond patent 1939/41 · Type 4 tank around 1960

Spring reverb

An input transducer sets metal springs vibrating; a transducer at the other end turns the vibrations back into a signal, with that typical metallic twang. Hammond built it into its organs, and from 1961 Fender sold a stand-alone Reverb unit based on Hammond's design. In dub it's a core sound: Lee "Scratch" Perry had a Grampian spring reverb at the Black Ark, and King Tubby physically hit his spring reverb to get a thunderclap crash. Classic studio units include the AKG BX 15, the Fisher Space Expander and the Fostex 3180.

⚑ sources differ Which spring reverb King Tubby used is poorly documented: dub engineer Daniel Boyle says Tubby used Fisher SpaceXpander units "for a little while" for big snare splashes, but the exact model isn't confirmed.

1976 · 1978

EMT 250 · Lexicon 224

The EMT 250 (1976) was the first commercial digital reverb – only about 250 were built. The Lexicon 224 (AES 1978, David Griesinger) cost about half as much and became the most widely used reverb of its era. Also digital and a studio legend: the Eventide H910 Harmonizer (1975), a pitch shifter – Tony Visconti used it on Bowie’s Low.

MICROPHONES

late 1940s · 1960 · 1967

Neumann U47 · U67 · U87

The U47 (VF14 tube, cardioid/omni switchable) defined the big studio vocal. The U67 (1960, EF86 tube) added figure-8, a bass roll-off and a pad. The U87 (1967) is its solid-state (FET) successor – omni, cardioid, figure-8 – still in studios everywhere.

⚑ sources differ U47 launch given as 1947 or 1949.

1953 · 1971

AKG C12 · C414

The C12 used the CK12 dual-diaphragm capsule and offered nine polar patterns, selected remotely from its power supply. The C414 (1971) carried the idea on as a solid-state mic.

1965 · 1973

Shure SM57 · SM7

The SM57 (cardioid dynamic, Unidyne III capsule) is on snares and guitar amps all over the world – and on every US presidential lectern since Lyndon B. Johnson. The SM7 (1973) built on it, with bass roll-off and presence switches; Bruce Swedien used one on Michael Jackson’s Thriller.

1931 · 1953

RCA 44 · Coles 4038

Ribbon mics with a figure-8 pattern. The RCA 44 (Harry Olson, from 1931) is the classic broadcast and big-band mic. The Coles 4038, a BBC design from 1953, became a favourite drum overhead – the Beatles and Led Zeppelin used them.

DUB STUDIOS

Stepped high-pass on Tubby's MCI desk · "the Big Knob" · Waterhouse, Kingston, 1970s

King Tubby’s high-pass filter

Tubby's signature filter sat on the second-hand MCI console he bought from Byron Lee's Dynamic Sounds in the early 1970s: a large knob, top right, that he called "the Big Knob". It switched a high-pass filter through ten notched frequency steps. Tubby swept horns and other parts through it until they thinned into a squeal, and the discrete steps added clicks and phasing effects. Product coverage identifies it as an Altec 9069B; Audio Merge's KTBK hardware and AudioThing's Dub Filter plugin recreate it.

⚑ sources differ The MoPOP museum record and Wikipedia call the Big Knob a "parametric equalizer", while product sources describe a passive Altec 9069B high-pass, and the step count is given as ten (Quietus, Audio Merge) or eleven (AudioThing).

1972 · Waterhouse, Kingston

King Tubby’s MCI

A used MCI console bought from Byron Lee’s Dynamic Sounds: 12 inputs, 4 outputs, 2 reverb returns and a big red stepped high-pass filter – “the Big Knob” – that Tubby swept across whole instruments. Together with two four-track machines it shaped the sound of dub. The desk is now in the collection of the Museum of Pop Culture in Seattle.

⚑ sources differ Purchase year: 1972 (MoPOP) or 1971 (Wikipedia).

1973–1979 · Lee “Scratch” Perry

Black Ark

Perry’s studio behind his house in Kingston: a TEAC 3340 ¼-inch four-track, first an Alice desk and later a Soundcraft, a Roland Space Echo, a Mu-Tron Bi-Phase and a Grampian spring reverb. King Tubby did the wiring. Few tracks, endless bouncing – and a sound nobody has copied since.

⚑ sources differ How and when it burned (1979 or 1983) is told differently.

ROUTING LAB

A real session, nine jobs. Click an output jack (top row), then an input jack (bottom row) to patch a cable. Click a cable end again to pull it. Dashed lines are normalled connections – they work until you plug into them.

In the live room: MIC 1 vocal, MIC 2 kick drum, DI bass guitar.

AUX 1
AUX 2

← swipe the patchbay →

BUILD THE CHANNEL STRIP

Tap the blocks in the order the signal travels through a classic console channel.

DAW LAB

Four stems of “Ratty Dub” by the Black Oak Roots Allstars – drums, bass, skank, lead. Every track has inserts (EQ, compressor, saturation) and two sends (plate-style reverb, tape-style delay). Play with it like a dub engineer – and read what each tool does.

momentary–LUFS approx.
short-term–LUFS
peak–dBFS

SPECTRUM

MISSIONS

Little jobs for your ears – they tick themselves off while you mix.

  1. Make room for the bass: high-pass the SKANK somewhere between 150 and 400 Hz.
  2. Tame the drums: compressor on DRUMS with 3–8 dB of gain reduction while it plays.
  3. Dub throw: hold THROW on the LEAD for a moment.
  4. Drop: mute SKANK and LEAD together while it plays – then bring them back.
  5. Space: give the SKANK a reverb send of at least 0.3 and a pre-delay of 30–80 ms.
  6. Mono check: switch MONO on while it plays.
  7. Hear a comb filter: turn on the PHASE demo with 1–5 ms delay.
  8. Loudness lesson: push the limiter until short-term is louder than −12 LUFS (e.g. −10). Then switch the limiter off and turn the volume up until both sound equally loud – hear what the limiter took away.

The DAW lab needs a current browser. On phones turn the volume up – the little speaker swallows the bass; headphones are better.

EAR TRAINING: FREQUENCIES

Engineers learn to name frequencies by ear. A bell EQ boosts one octave band – which one? Toggle EQ IN/OUT to compare, then answer.

Typical character: 125 Hz boom · 250 Hz warmth/mud · 500 Hz boxy · 1 kHz honk · 2 kHz nasal · 4 kHz presence/harsh · 8 kHz sizzle · 16 kHz air. (Rules of thumb – every source is different.)

QUIZ

Eight questions per round. Pick a level.