Why psychoacoustics matters
We mix for ears and brains, not for analyzers. Psychoacoustics studies how we perceive sound: some of its rules explain why a mix sounds different at different volumes, why a delay of a few milliseconds widens a sound, why a loud hit “hides” what comes right after it.
Equal-loudness contours (Fletcher-Munson)
Harvey Fletcher and Wilden Munson (1933) measured that the ear is not equally sensitive to all frequencies: we are most sensitive between roughly 2 and 5 kHz and much less so to the lows and the very highs, especially at low volume. The curves have been updated by the ISO 226 standard (2003, revised in 2023).
Consequences for mixing:
- At low volume the mix seems poor in lows and highs; at high volume it seems fuller and brighter. This is why “louder seems better” (see “Review and A/B comparison at equal loudness”).
- Mix at a moderate, constant level: many engineers work around 75–85 dB SPL; Bob Katz, with the K-System, proposed calibrating the reference at 83 dB SPL per channel with pink noise at −20 dBFS RMS (for small rooms people often go down to 73–79 dB).
- Also check the mix at very low volume (mid balance) and briefly at high volume (lows and fatigue).
The Haas effect (precedence effect)
When the same sound arrives from two directions with a delay between roughly 1 and 30–40 ms, the brain perceives it as a single sound coming from the direction of the one that arrived first (Helmut Haas, 1951). The second arrival adds width and volume, not an echo. The delayed sound can even be about 10 dB louder before it “shifts” the localization.
Uses:
- Widening a mono element: a copy on the right delayed by 10–25 ms (better with a slight EQ change).
- Problems: in mono the two copies are summed with a delay and produce comb filtering. A 10 ms delay creates notches at 50, 150, 250 Hz... Always check mono compatibility.
- Beyond 35–50 ms the delay is heard as a distinct echo: the slapback zone.
Temporal masking
A loud sound masks weak sounds close to it in time:
- Post-masking: after a loud hit, for several tens of ms (up to 100–200 ms in some measurements) weak sounds are less audible.
- Pre-masking: surprisingly, even sounds a few ms before (about 5–20 ms) are masked.
Consequences: ghost notes right after the snare are hard to hear; a hidden editing click next to a transient is inaudible; MP3/AAC codecs exploit this principle to discard data.
Other useful rules
- Missing fundamental: if a sound contains the harmonics (110, 165, 220 Hz...) but not the fundamental (55 Hz), the brain “reconstructs” it. That is why saturating an 808 makes it audible on smartphones (plugins like Waves MaxxBass exploit this effect).
- Adaptation: after a few minutes the ear gets used to a timbre (too much treble starts to seem normal): take breaks and use references.
- Ear fatigue: after hours at high volume the perception of highs changes; work in sessions with breaks.
- Critical bands and frequency masking (see the dedicated lesson).
- Loudness and duration: short sounds seem less loud than long sounds at the same peak: this is why loudness is measured in LUFS (with integration over time) and not in peaks.
Practical example
Mixing a pop track in A minor at 100 BPM: a mono synth pluck widened with Haas at 18 ms → check in mono: the pluck loses 300 Hz; solution: delay at 12 ms with an HP at 400 Hz on the delayed copy (comb filter shifted upward and less noticeable). The 808 bass is inaudible on the phone: saturation that generates harmonics at 110–220 Hz. The snare ghost notes vanish behind the backbeat: move them a sixteenth note farther away or raise them by 3 dB.
Exercise
- Listen to the same mix at 60, 75 and 90 dB SPL (use a sound level meter or an app) and note how the balance changes.
- Try the Haas effect with delays of 5, 15, 30, 50 ms and listen in stereo and in mono.
Related topics
“Frequency masking”, “Phase and polarity”, “Mono compatibility”, “Review and A/B comparison at equal loudness”, “Mixing for social media and smartphones”.