Serial and parallel are different tools
Serial processing replaces what is there: the output of each stage becomes the input of the next. Parallel processing adds to it: a copy is sent elsewhere, treated, and summed back with the original left intact.
The distinction matters more than it sounds. Serial processing shapes a signal. Parallel processing layers a treatment over one, which opens up a different set of options that a single chain cannot offer.
What a threshold really does
A threshold sets the level at which a processor begins working. Below it, the signal passes through untouched; above it, the ratio takes over. That makes the threshold a gate on the effect rather than a global setting.
On a parallel bus this is the control that makes the technique useful. Set the threshold so the effect only engages on the loudest moments, and you get density in the peaks without the effect colouring the quiet parts.
What parallel buses are for
Four uses cover most of what the technique gets asked to do:
- Adding compression density without the level loss of serial compression
- Applying saturation or distortion as a layer while keeping the clean signal underneath
- Creating an aggressive version of a vocal for a specific section, gated by threshold
- Adding reverb or width to a chosen range rather than to the whole source
The common thread is selectivity. A parallel bus is how you apply something to part of a signal without committing to it everywhere.
Retuning a summed bus
A parallel bus produces a summed result, and that changes the retuning question. If you retune only the dry signal, the wet return arrives at the old pitch and the two no longer agree.
The practical answer is to retune the sum, after the parallel path is mixed in. 432Hz MASTER placed on the mix bus after the parallel returns covers the entire summed signal, so the dry and wet stay locked together. RASTA 528Hz does the same for the 528Hz reference.
When not to use it
Parallel processing is a precise instrument, and precision costs clarity. If the goal is simply more compression across a whole mix, a well-set serial compressor is easier to reason about and easier to undo.
Reach for parallel when the requirement is selective. Reach for serial when it is global.
Setting up a parallel bus
The mechanics are simple enough that the difficulty is entirely in restraint. Send a copy of the source to a return, process the return, and sum it with the dry signal still at full level.
| Control | What it decides | Typical starting point | ||
|---|---|---|---|---|
| Send level,How much of the parallel path is mixed in,Low, then raise until the effect is obvious | Threshold,Where the effect begins on the source,Just under the level of the loudest moments | Ratio,How hard the threshold is enforced,Moderate; high ratios sound abrupt | Attack,How quickly the effect engages,Slow enough to let transients through | Release,How quickly the effect lets go,Matched to the tempo of the material |
Attack deserves particular attention because it is where parallel processing most often goes wrong. A fast attack catches the transient, which is the part of the signal that already sounds correct, and the result is an effect that squashes exactly what you wanted to keep. A slower attack lets the transient pass untouched and applies the effect to the body of the sound, which is usually the intent.
Set the send low and raise it until the effect is clearly present rather than vaguely felt. If you cannot hear the parallel path, it is either too quiet to matter or doing nothing useful.
The bottom line
Serial processing shapes a signal; parallel processing adds a second treatment of it. That difference is the whole idea. When you need an effect to be obvious in one part of the arrangement and absent in another, a threshold-controlled parallel bus gives you a control that no single processor can. Use it when the problem is selective, and keep it simple when it is not.
Retune a parallel bus cleanly
Parallel buses add signals in, so a retune needs to cover the summed result. 432Hz MASTER and RASTA 528Hz handle the whole bus.
Frequently asked questions
What is parallel processing in mixing?
Sending a copy of a signal to a separate processing chain and summing it with the original. The original stays untouched, so you add an effect rather than replacing one.
Why use parallel compression instead of normal compression?
Parallel compression lets you add density and control without the pumping and level reduction that serial compression introduces, because the dry signal is never attenuated.
What does a compression threshold control?
The level above which the processor begins acting. Above it, the signal is compressed; below it, it passes through at the ratio you set, or at unity if you have set a threshold gate.
Does parallel processing use more CPU?
Yes, generally, since you are running a second chain. It is a normal cost, but worth knowing if a project already runs many instances.