Loudness vs. gain: why louder isn't just more level
Raising gain makes a mix louder by making everything bigger, peaks included. Shaping the peaks first lets you reach the same loudness with far less peak level. We measured both on the same full mix: same loudness, 3.3 dB apart at the peak.
By Dennis RosnerGründer, Entwickler & Audio Engineer
In short: Turning up gain makes a mix louder by making everything bigger, the peaks included. Shaping the peaks first lets you reach the same loudness with far less peak level. On the full mix we measured below, both versions end up at −9.3 LUFS, but one peaks at +3.0 dBTP and the other at −0.3 dBTP.
Loud has two numbers
When we say a track is loud, we usually mean two different things at once.
Loudness is what you hear: how much energy the track carries over time, weighted the way the ear hears it. We measure it in LUFS (integrated over the whole clip). Streaming services, broadcast and most loudness meters use it.
Peak level is how far the waveform swings at its highest moment. We measure it in dBTP, true peak, which also catches the peaks that happen between two samples.
The gap between the two is the PLR, the peak-to-loudness ratio: true peak minus integrated loudness, in dB. For the original below that is −1.2 dBTP − (−13.5 LUFS) = 12.3 dB. A dynamic mix has a large PLR: tall peaks, lower average. A dense mix has a small one.
Don't confuse it with the crest factor: peak level minus the RMS level of the signal. Both describe the distance between peak and average, but against a different average. RMS adds up the raw energy. LUFS weights frequencies the way we hear them and leaves out very quiet passages. For the same audio the two numbers differ. In this article we only use PLR, because that is what we measured.
The experiment
We took 15 seconds from the hook of a finished full mix, starting right at the first impact, and made two louder versions of it:
- Gain: the same audio, turned up by 4.2 dB.
- PUSH: the same audio through PUSH in Clean at 100 %, Shape 0, Mix 100 %, TP SAFE on, Band Split off, oversampling 4x, no other processing.
We measured the original and the PUSH version the same way: EBU R128, integrated loudness and true peak. The gain version needs no separate measurement. Gain moves loudness and peak by exactly the same amount, so we calculated it from the original.
- Original: −13.5 LUFS, −1.2 dBTP, PLR 12.3 dB
- Gain +4.2 dB: −9.3 LUFS, +3.0 dBTP, PLR 12.3 dB
- PUSH, Clean: −9.3 LUFS, −0.3 dBTP, PLR 9.0 dB
Both louder versions are equally loud. The difference is entirely in the peaks: +3.0 dBTP against −0.3 dBTP is 3.3 dB. The PLR shows the same thing from the other side. 12.3 dB means the peak sits 12.3 dB above the loudness, 9.0 dB means 3.3 dB less.
Why gain can't do this
Gain multiplies every sample by the same number. The quiet parts get 4.2 dB bigger, and so do the peaks. The shape of the waveform does not change at all, which is why the PLR stays at exactly 12.3 dB.
That means the true peak of the gain version sits 3.0 dB above full scale. Played back in floating point, nothing happens yet. But the moment this file is exported to a fixed-point format such as 16 or 24 bit WAV, everything above 0 dBFS is cut off. That is clipping you did not choose. Lossy formats like MP3 and AAC can overshoot full scale as well, so peaks this high are a problem there too.
To keep this loudness and stay below full scale, you have to put something after the gain that deals with those peaks, usually a limiter.
What shaping the peaks changes
PUSH doesn't scale the waveform, it changes its shape. Quiet parts pass through largely untouched in Clean. The closer the signal gets to the top, the more the curve bends it down. The tallest peaks give up the most, the body of the mix keeps its level.
The result is a waveform with less distance between peak and loudness: a PLR of 9.0 dB instead of 12.3 dB. The same loudness now needs 3.3 dB less peak level.
What this means for your mix
- Compare at equal loudness. Louder always sounds better for a moment. Match the loudness of any two versions before you judge them. PUSH has a switch for this (MATCH), any loudness meter and a gain plugin do the same.
- Watch the PLR, not just the peak. Two mixes with the same peak can be several dB apart in loudness. Subtract the integrated loudness from the true peak, most loudness meters show both. The result tells you how much room there is.
- Decide where the peaks go. Every louder master handles its peaks somewhere: a clipper, a limiter, saturation, or all three. Shaping them earlier means the last stage has less work to do.
Where it doesn't help
Shaping peaks needs peaks to shape. On quiet material, far below full scale, the signal barely reaches the part of the curve that bends, and PUSH mostly just adds level. The loudness gain there is real, but it behaves like gain.
It also costs something. A smaller PLR means less dynamic range. That is the mechanism, not a side effect. On very dynamic material, such as drums with a lot of attack, the tallest transients lose some of their edge at high settings. How much depends on the material: the denser the mix already is, the less there is to lose.
With PUSH
PUSH is a loudness and density processor. One control sets how much happens, three characters decide how it gets there: Clean, Color and Edge. The experiment above is Clean at 100 % on a full mix.