Four stages, in this order: a bus compressor working at 2:1 with one to two dB of gain reduction, an equalizer making moves of a dB or less, a light saturation stage, and empty space where a limiter would sit. The limiter goes on a separate reference bounce for the artist and never on the file that leaves for mastering. I mixed and mastered Stefania for Kalush Orchestra, and my bus chain has stayed this short for years, because every insert on the two-bus processes every decision made upstream of it. The order matters as much as the settings.
Why does the compressor go on early instead of at the end?
Because a bus compressor changes what every fader does. Push a vocal 1 dB into an active 2:1 bus compressor and the mix gives back less than 1 dB of vocal, since the louder vocal drives the compressor a little harder. The Sound On Sound guide to mix compression makes the same point from the other direction: identical fader settings produce a different balance depending on whether the bus compressor is in and how hard it works. So I set a clean static balance first, engage the compressor while the mix is still simple, and make every later decision through it. Switching glue on during the last hour quietly rebalances a mix that everyone already approved.
What is the glue compressor actually set to?
A ratio of 2:1. Attack between 10 and 30 milliseconds, slow enough to let the front of the kick and snare through before the compressor grabs. Release on auto, or around 300 milliseconds when auto breathes strangely against the tempo. Threshold set so the loudest chorus shows one to two dB of gain reduction and the needle returns to zero between phrases. The drums do most of the triggering, which is the point: the whole mix ducks a fraction of a dB with each hit, and everything moves together.
The sound this describes comes from one specific place. The stereo compressor in the centre section of SSL consoles has been doing this job since the 4000 B console in 1976, and the version most people mean by glue arrived with the G Series in 1987. It is a VCA design, which is why it reacts quickly and stays clean; the mechanics of the detector, ratio and makeup stage behind any of these boxes are laid out under dynamic range compression. An honest VCA-style bus compressor, hardware or plugin, gets there with the settings above.
Where does the EQ sit, before or after the glue?
After it, in my chain, with one exception. Post-compressor EQ shapes tone without changing how hard the mix drives the detector, so a half dB of top end stays a half dB of top end. The exception is low end: when a dense chorus makes the compressor pump on every kick hit, a wide shelf pulling 1 dB below 100 Hz goes in front of the compressor to calm the detector, and the weight comes back after the compressor if the mix misses it. Bus EQ moves stay broad, a dB or less, wide Q. Anything narrower is a channel problem, and the channel is where I go fix it.
Does saturation belong on the bus?
In small amounts, gain-matched before it gets judged. A console-style or tape-style stage adds harmonic density that reads as size long before it reads as distortion, and the failure mode is level: drive adds output, louder wins an A/B, and the saturation takes credit the volume earned. So the stage gets trimmed until bypassing it moves the loudness reading by less than about 0.2 LU, and only then does it stay or go. Two more checks: mono, because some saturation stages smear the sides when the mix folds down, and the quietest verse, because a drive setting chosen on the chorus is usually too much for the verse.
Why does the limiter stay off the print?
Because loudness is a mastering decision made against playback normalization, and I keep those two jobs in separate sessions even when both are mine. The print that leaves the mix session peaks around minus 6 dBFS at 24-bit, with the glue, EQ and saturation printed and nothing across the peaks. The artist hears a different file: a reference bounce through a limiter at a competitive level, because a client comparing an unlimited print against a commercial release will hear quiet as worse, every time. How the final master then lands against each platform's playback target is its own subject, one I walked through in how streaming normalization decides a master's level.
What gets checked before printing?
Bypass tests, all level-matched: the compressor alone, then the whole bus chain, then mono. If the full chain bypassed sounds just as good at matched loudness, the chain comes off, and some sparse arrangements are better served by an empty bus. Then one pass at conversation volume, around 60 dB SPL, listening only for whether the vocal and the kick still lead. The bus chain is the last item on the same checklist that starts before the first fader moves, the one I laid out in how a session gets set up before mixing.
Frequently asked
What are typical glue compressor settings on a mix bus? A ratio of 2:1, attack between 10 and 30 milliseconds, release on auto or around 300 milliseconds, and a threshold set so the loudest chorus shows one to two dB of gain reduction.
Should the mix bus have a limiter on it? Only on the reference bounce the artist hears. The file that goes to mastering leaves with its peaks intact, around minus 6 dBFS, at 24-bit.
When should bus compression go on, at the start of a mix or the end? Early. A bus compressor changes how every fader move reads, so switching it on at the end quietly changes every balance already approved.
If you are about to print a mix
Bounce it twice: the mastering print with headroom and no limiter, and the loud reference for everyone else. If you want a second pair of ears on the bus chain itself, send me both bounces and the session tempo. Credits are at /credits, background at /about.