Why Gain Staging Matters So Much to Me - and Why 32-Bit Float Doesn’t Save Me

I’ve gone through the same process countless times: channel by channel, the levels gradually creep up until suddenly everything is in the red. At that point, it becomes almost impossible to maintain a meaningful balance. Every element is already close to its limit, and the mix begins to lose its internal relationships.

Some of my friends and colleagues remain calm, following the widespread assumption that this is not really a problem in a digital system. With 32-bit floating-point processing, signals can exceed 0 dBFS internally without clipping, and summing does not necessarily introduce distortion. To me, the issue is this:

32-bit float doesn’t fix poor gain staging. It merely hides some of the consequences.

Even if nothing clips, the behaviour of my processing can change: Compressors react differently, saturation increases, some EQs and analogue-modelled processors behave differently depending on the input level and my prepared processing chains drift away from their intended spots.

The result may still be technically clean, but it is no longer the emotion I originally built.

In analogue systems, this was often more obvious to me. I could see the meters reacting, and VCAs allowed me to control the overall level of groups without destroying their internal balance. When things became too loud, I could simply pull down the VCA and keep the rest of the signal chain operating within its intended range.

In digital environments, that overview is often missing, especially in complex sessions. Small level changes can occur deep inside a signal chain without being immediately visible. Suddenly, the entire setup begins to behave differently.

That is why proper gain staging is still essential to me.

My Workflow

In both my live and studio setups, I have built a system in Ableton Live that mirrors this analogue logic. I use Utility devices at the beginning and end of my channels. The first Utility controls how hard the signal hits the processing chain: the pre-gain. The second acts as a VCA-style output control, allowing me to adjust the final level without changing how the internal processing behaves. On top of that, I organise my signals into musical categories: Drums and rhythm, Bass, Synths, sometimes divided into frequency ranges, Lead vocals and backing vocals and Effects.

These groups are mapped to a hardware controller, in my case an Allen & Heath Xone:K2, giving me direct access to their overall levels.

So instead of adjusting the kick, snare, hi-hats, and percussion individually I can move entire sections with a single gesture.

The internal relationships remain intact, and the processing continues to behave as intended - without risking the relationships between their individual elements.

My Mixing Levels

I use −23 LUFS and −9 dB peak as internal reference levels.

This results in an approximate peak-to-loudness ratio of 14 dB. It is a practical choice that gives me enough headroom for summing, processing, and real-time level changes during a performance.

These values are not intended as final mastering targets. They define a stable internal working environment in which my processing remains predictable and the musical relationships remain intact.

Structured and Repeatable Processing

Combined with a mindset focused on preparation rather than constant reinvention, this creates a system that is flexible and reliable.

With only a few movements, I can push an entire mix forward in a specific direction - or pull it back. This creates room without compromising technical consistency - and I can concentrate on musical expression.