Minus Nine LUFS and a Hum That Won't Leave
Tomas is mixing a 38-minute two-person interview for web release. Host A's lav went to a 48 kHz recorder; Host B's lav to a 44.1 kHz portable recorder. There is constant HVAC hum around 60 Hz plus a room-tone bed. Tomas normalized every dialogue clip to -3 dB peak, added DeNoise and a Hard Limiter at -0.1 dB on the Master, and pushed the Master fader to +3 dB. The result: dialogue pumps whenever either host laughs, the hum is still clearly audible, and integrated loudness measures -9 LUFS. The platform target is -14 LUFS integrated.
1. Diagnose the pumping and the loudness overshoot, then describe the correct gain-staging order.
Read the worked answer
The pumping is gain staging, not a limiter fault. Three compounding errors:
- Normalizing to -3 dB peak raises each clip by a different amount, so quiet and loud lines now hit the limiter at different moments. Peak normalization says nothing about perceived loudness.
- A Hard Limiter at -0.1 dB on the Master while the master fader sits at +3 dB means the limiter is working continuously. When a host laughs, it clamps and the whole mix ducks - that is the pump.
- The master fader is the wrong place to add level. It just moves the ceiling the limiter is already fighting.
Correct order: clip gain or Essential Sound > Loudness Auto-Match so dialogue averages around -18 dBFS; balance on the tracks and a dialogue submix; limiter on the Master as a safety catching 1–2 dB on peaks only; then read integrated loudness on a meter and reach -14 LUFS with a gain stage, never by driving the limiter harder.
2. Why did DeNoise leave the hum, and what does the -9 LUFS reading tell you about how the mix will behave on the platform?
Read the worked answer
DeNoise is broadband; hum is tonal. It lowers the noise floor across the spectrum but does not notch a 60 Hz sine and its harmonics at 120/180/240 Hz, which is exactly why the hum survives. Use DeHum at 60 Hz with 4–6 harmonics, or a narrow-Q parametric notch at 60/120/180. Do it before DeNoise - a strong tonal component confuses the noise-print learning and leaves artefacts.
The -9 LUFS reading is the arithmetic of the gain staging above. It also predicts the outcome: a platform that normalizes to -14 will simply turn this mix down by 5 dB, and because the limiter has already flattened the peaks, it will sound quieter and flatter than a mix built at -14 with intact dynamics. Loudness is not a number to maximise - it is a target to hit while keeping the limiter idle.
3. What are the real risks with the 44.1 kHz clip over 38 minutes, and how would you verify before committing?
Read the worked answer
Normally none - but verify, because the failure mode is severe.
- Premiere resamples 44.1 → 48 kHz on playback and export automatically, so the clip plays at correct pitch and speed and does not drift.
- Drift appears only if the clip was misinterpreted. Right-click > Modify > Interpret Footage must read 44.1 kHz. If it reads 48 kHz, the file plays about 8.8% fast and slides badly out of sync over 38 minutes.
- Conversion quality: for a music bed under dialogue it is inaudible. For a critical solo, render it to 48 kHz in Audition so you choose the converter rather than leaving it to the sequence.
Verification: drop a sync reference at the head and tail of the clip and confirm the waveform still lines up with the picture, then read the Interpret Footage value. If both check out, leave the clip alone - "fixing" a correctly interpreted 44.1 kHz file is how editors create the drift they were trying to prevent.