Plugin order changes what each processor hears

Same plugins, different record

You have an EQ and a compressor on a bass. Out of curiosity you drag the EQ below the compressor. Nothing else changes, same settings, same plugins, and the bass now sounds different: the low notes stop pulling the whole sound down, and the movement feels looser. Then you swap two EQs on the vocal and hear no difference at all.

In a serial chain, every processor works on what the one before it hands over. That matters when a processor reacts to level, and a clean EQ does not.

Why it matters

Order is part of the sound

An insert chain is a series of stages, and each stage hears only the output of the stage above it. A compressor after an EQ hears the equalized signal. A saturator after a compressor hears a signal whose level has already been evened out. A limiter at the end hears everything every earlier plugin did.

If you never think about order, you end up fixing its side effects with more plugins: a compressor that pumps on the low end because a boost sits in front of it, distortion that comes and goes with the performance, an EQ after the limiter that pushes peaks back over the ceiling.

The science

Which stages commute

A clean digital EQ is linear and time-invariant. Two such filters in series multiply their frequency responses, and multiplication does not care about order, so any ordering of filters in series gives the same overall response (Smith, 2007). Two EQs, a high-pass and a shelf, a delay and an EQ: swap them freely. The same is not true of an EQ that models analog saturation, because that is no longer linear.

A compressor is not linear. Its gain depends on the level it receives (Giannoulis, Massberg and Reiss, 2012; see how compression shapes motion), so whatever changes that level changes the gain. Take a bass note whose low end drives the detector, 6 dB over the threshold of a 4:1 compressor. The compressor removes $6 \times 3/4 = 4.5$ dB. Put a 6 dB low boost in front of it and the detector is 12 dB over, so the compressor removes $12 \times 3/4 = 9$ dB.

On a steady note and at matched level, the tonal balance at the output ends up the same both ways: the lows sit 6 dB above the rest because the compressor turns every frequency down together. What changes is the gain movement. With the boost in front, the compressor works twice as hard and follows the low end, so the mids and highs dip twice as deep with every bass note. With the boost after it, the compressor reacts to the original signal and the boost shapes only the tone. The same logic is why a high-pass in front of a compressor is common: it stops rumble from steering the gain.

Saturation is not linear either. A saturator bends loud input more than quiet input, so how much it distorts depends on the level that reaches it (Dutilleux, Dempwolf, Holters and Zölzer, 2011). Put a compressor first and every note reaches the saturator at a similar level, so the distortion is even. Put the saturator first and the distortion follows the performance, loud notes grittier than quiet ones, and the compressor then hears peaks that were already rounded off.

The demo runs bass and chords through a waveshaper at matched level. Move the drive and listen to how much the sound changes with the level going in. That level is exactly what the stage before a saturator decides.

A limiter at the end of a master works the same way. A bright EQ boost in front of it gives the limiter more high-frequency peaks to catch, so it turns down harder on cymbals and consonants. The same boost after the limiter changes the tone of the limited signal and can push peaks back over the ceiling the limiter just set.