Mix & Master Your Tracks: The Engineering Behind a Finished Record

Studio Guide 05 · Cloud Atelier · Updated April 2026 · ~16 min read

After the take is recorded, the mix is where physics meets taste. EQ shapes the frequency domain. Compression shapes the time domain. Saturation adds harmonic content the source did not have. Mastering translates the result across every playback system on Earth at a loudness that does not embarrass the music. This guide treats mixing the way an audio engineering textbook treats it — as four interlocking domains, not a list of plugin presets.

1. The post-recording reality — what mixing is and is not

Recording captures sources. Mixing makes those sources relate to each other in a single coherent space. Mastering takes that mix and makes it survive every playback environment without falling apart. These are three different jobs with three different success criteria.

A common amateur mistake is treating mixing as “making each track sound good in solo.” That is not the goal. A vocal that sounds beautiful in solo often disappears in the mix. A snare that sounds huge in solo often muddies the verse. Mixing decisions are always relative: this loud relative to that, this bright relative to that.

Mastering is also frequently misunderstood. It is not “make it louder.” It is “make it translate.” A master that sounds good on the studio monitors but harsh on a phone speaker is a failed master. A good mastering engineer is essentially a translator who works in the mid-range, true peak, and frequency balance to ensure the song lands the same emotional weight whether it plays on a $50,000 monitoring chain or a $20 Bluetooth speaker.

2. Gain staging and headroom — the most overlooked mistake

Modern DAWs work in floating-point arithmetic, which means clipping inside the DAW between plugins is mathematically impossible to lose information — the signal can exceed 0 dBFS internally and still be recovered. So why do experienced engineers still mix at −18 dBFS RMS? Two reasons:

• Plugin behaviour. Many analog-modeled compressors and saturators have internal calibration based on classic studio levels. A 1176 emulation expects roughly −18 dBFS to behave like the hardware sitting at +4 dBu / 0 dBVU. Slam it with −6 dBFS and the saturation curve compresses and the model breaks down.
• Mastering headroom. A mix bouncing at −1 dBFS true peak gives a mastering engineer no room to push. A mix bouncing at −6 dBFS true peak with comfortable internal levels gives them options.

The K-system (Bob Katz) formalises this. K-20 metering pegs 0 VU to −20 dBFS, leaving 20 dB of digital headroom for transients. It is the calibration most narrative film mixes and orchestral music use. Pop and electronic mixes more often work to K-12 or K-14, but the principle is the same: do not mix “to the top of the meter.”

The mechanical version: get a clean rough mix at −18 dBFS RMS on the master bus. Push plugins from there. If at the end of the mix the master bus is still around −6 dBFS peak, you have given the mastering chain real room to work.

3. EQ — subtractive vs additive, parametric vs shelving, masking

EQ shapes the frequency-domain balance of a sound. Three principles cover most mixing EQ decisions:

Subtractive over additive, by default

If two instruments fight in the 200 Hz region, you have two options: cut 200 Hz on one of them, or boost 200 Hz on the other. Cutting is almost always better. Boosting raises the absolute level (eats headroom), often shifts the phase response of analog-modeled EQs (causes a smeared transient), and tends to mask other problems instead of solving them.

Parametric for surgery, shelving for tone

Parametric (peak/notch) bands cut or boost a narrow Q around a centre frequency — the scalpel. Shelving bands lift or drop everything above (high shelf) or below (low shelf) a corner frequency. Use parametric to remove a specific resonance or boxiness. Use shelves to shape overall tonal character (“a touch more air,” “less low end”).

Masking is the real enemy

Two sounds in the same frequency region create masking: the louder one perceptually hides the quieter one. The classic case: kick drum and bass guitar both occupying 60–120 Hz. A 2 dB cut on the bass at the kick’s fundamental, or sidechain compression of bass triggered by kick, separates them more cleanly than EQ alone. The principle: find what each element uniquely does and protect that frequency range on that track.

4. Compressor topology — VCA, FET, Opto, Tube

A compressor reduces dynamic range above a threshold by some ratio. That much is universal. The differences in character come from the gain-reduction circuit topology.

Two compressors in series with mild settings (one Opto for smooth levelling, one VCA for peak control) usually outperform one compressor with aggressive settings. The Opto handles average level, the VCA catches transients. Each does about 2–3 dB of gain reduction and the result sounds untreated to the listener while doing real work.

5. Stereo width and panning — pan law, the Haas effect, M/S

Pan law is the centre attenuation when a mono signal is panned. Three common values:

• −3 dB pan law. Standard in most DAWs. Centre signal is 3 dB quieter than full-left or full-right. Mono compatibility good.
• −4.5 dB pan law. A compromise widely used in broadcast.
• −6 dB pan law. Centre is 6 dB quieter. Stereo image more dramatic, mono summing weaker. Common in some classical recording schools.

The Haas effect: when two copies of a signal are separated by 1–30 ms and panned hard left/right, the brain perceives a single wide source rather than two sources. This is the basis of many vocal doubling and stereo widening techniques. Beyond ~30 ms the brain hears a discrete echo.

Mono compatibility still matters. Many speakers (single Bluetooth speakers, phone earpiece, restaurant ceiling mono summing) collapse stereo to mono. Sources that rely on phase tricks (stereo wideners that pan duplicate signals out of phase) cancel and disappear in mono. The fix: check your mix in mono periodically. Anything that loses level when summed is a phase problem you should solve before mastering.

Mid/Side processing. M/S decoding splits a stereo signal into the sum (mono content, “mid”) and the difference (stereo-only content, “side”). EQ-ing the side channel only with a high shelf at 8 kHz adds “air” to a mix without affecting the mono centre. This is a mastering-stage trick that costs nothing and translates beautifully.

6. Saturation and harmonic distortion — loudness without level

Saturation adds harmonic content the original signal did not have. Even-order harmonics (2nd, 4th, 6th) add “warmth” and are perceived as musical — they are octaves and double-octaves of the fundamental. Odd-order harmonics (3rd, 5th, 7th) add “edge” or “grit” and are perceived as more aggressive — they are fifths and major ninths.

Different saturator topologies emphasise different harmonics:

• Tube saturation tends toward 2nd and 4th harmonics — smooth.
• Transformer saturation rolls off the very low end and adds 3rd harmonic on transients — the “Neve sound.”
• Tape saturation compresses transients (head bump at low frequencies, high-frequency loss) and adds soft odd harmonics.

The practical use: perceived loudness without raw level increase. A snare with 2 dB of mild tube saturation feels louder than the same snare with 2 dB of gain, because the harmonics add presence in the mid-range where the ear is most sensitive (Fletcher-Munson curves). This is the secret behind most modern pop drum sounds.

7. Loudness in 2026 — LUFS, true peak, platform truth

Streaming platforms apply loudness normalisation. Tracks above the platform’s target are turned down on playback; tracks below are not turned up. The current targets engineers should design for:

The implication of normalisation: mastering louder than the target (e.g. −7 LUFS) does not make the track louder on Spotify — it gets turned down to −14 by the platform. What it does do is reduce dynamic range, often permanently, and limit the headroom available to transients.

True peak versus sample peak. Sample peak meters look at digital sample values. True peak meters reconstruct the analog waveform between samples (oversampling) and catch inter-sample peaks that exceed 0 dBFS even when sample peaks are clean. Codecs (AAC, Opus) expose these inter-sample peaks as audible distortion on consumer playback systems. Master at −1 dBTP, not −1 dB sample peak.

8. Multi-system referencing — your room cannot be the only judge

Every mix sounds great on the speakers it was mixed on. The test is whether it survives anywhere else. The standard professional workflow:

1. Main near-field monitors — primary mixing surface, treated room.
2. A second pair of monitors with different character — verify frequency balance is not a flaw of the main pair.
3. Headphones (open-back) — check stereo image, vocal sibilance, the things headphones reveal that monitors mask.
4. Mid-fi check — phone speaker, single Bluetooth speaker, car stereo. The places people actually listen.

Calibration helps. Software like Sonarworks SoundID Reference (formerly Reference 4) measures your room with a calibrated microphone, generates a correction profile, and applies it to your monitor output. The room’s 200 Hz boost or 6 kHz suckout is actively flattened. This will not fix a bad room (treatment still matters), but it makes a treated room behave like a more accurate one.

9. Five mistakes home mixers actually make

1. Mixing too loud. Listening fatigue at high SPL distorts judgement. Mix at conversation level (~75 dB SPL).
2. Soloing too much. Mixing decisions are made in context. Solo is a diagnostic tool, not a workflow.
3. Boosting where you should cut. Subtractive EQ first, additive only after.
4. Compressing every track. Some sources do not need it. The piano performance with intentional dynamics did not ask to be compressed.
5. Mastering yourself. Two days into a mix, your ears are calibrated to its flaws. A second engineer with fresh ears catches what you can no longer hear.