Parallel Processing
Audio Effect Rack chains in parallel, phase and latency with Align Delay, parallel distortion and reverb, three ways to go multiband (racks + EQ, Multiband Dynamics, Roar Multiband), returns vs rack parallel, and wet/dry rules.
Parallel processing means splitting a signal, treating the copies differently, and summing them back. It is how you get the character of a heavy process at a fraction of its cost: a crushed drum chain under a clean one, folded sides under a clean centre, a distorted mid band next to an untouched sub. In Live the Audio Effect Rack is the parallel tool; returns are the other one, and they are not interchangeable. The thing that decides whether parallel sounds huge or hollow is phase: any latency or phase shift between chains turns the sum into a comb filter.
Concept
Summing two copies of the same signal is +6 dB if they're identical and in phase. If one is delayed by t seconds, the sum has notches at every odd multiple of 1/(2t): a 1 ms offset puts the first notch at 500 Hz and the next at 1.5 kHz, 2.5 kHz… That is a comb filter, and it sounds thin and phasey. Even without a time delay, a process that shifts phase (any IIR filter, most EQs, any multiband crossover) does the same thing frequency-dependently. So the rules are:
- Latency between chains must be zero or compensated. Live compensates latency between tracks, not between chains inside a rack. (Live's rack does align chain latency in most cases, but devices that report latency incorrectly, and anything hand-built with delays, won't be.) Check by summing a chain with its inverted-polarity twin: silence means aligned.
- Filters in one chain shift phase around their cutoff. EQ Eight in Linear Phase mode doesn't, at the cost of latency on both chains (fine, since the rack compensates for a device's reported latency).
- The crushed chain lives under the clean one. −12 to −6 dB is the normal range. If the parallel chain is louder than the dry, you've built a serial chain with extra steps.
Techniques
Audio Effect Rack: chains in parallel, done right
Every chain in an Audio Effect Rack receives the full input and the chains are summed at the rack output. Chain volume, pan, solo and mute are per chain; the chain selector lets you crossfade between chains with a macro. An empty chain is the dry signal.
- Cmd+G on an effect to put it in a rack. Show chains, right-click the chain list, Create Chain for the dry. Name them.
- Build each wet chain with a filter first so it operates in a defined band, the process second, a Utility last for trim.
- Set wet chain volumes to −inf, raise one at a time until it contributes, then back off 2 dB.
- Polarity check: temporarily set the dry chain's Utility to invert polarity (Phase L+R). With the wet chains muted you should hear silence only if there's a second dry copy; more usefully, invert the dry and listen to the wet-only remainder to understand what the chain adds.
- Map chain volumes to macros; use Macro Variations to store mix states.
Phase and latency: Align Delay and the polarity test
Align Delay is a sample-, millisecond- or metre-accurate delay with no filtering; it exists to line up parallel paths. Use it in the chain that has less latency, set to the difference. Inside racks Live compensates reported device latency, so you usually need Align Delay only for external hardware, for M4L devices that don't report latency, or when you deliberately delay a chain for a Haas effect.
- Record a single click (Impulse) through both chains onto a new track, solo each chain in turn.
- Zoom to sample level, measure the offset. Set Align Delay in the earlier chain to that offset (Samples mode for precision).
- Confirm: invert polarity on one chain with Utility. Summed output should null (silence) if the chains are otherwise identical. With different processing there's no full null, but the low end should drop dramatically when aligned.
- Remember hardware latency changes with buffer size; re-measure after changing audio settings.
| Offset | First comb notch | Audible as |
|---|---|---|
| 0.1 ms (5 samples at 48k) | 5 kHz | Slightly dull top end |
| 0.5 ms | 1 kHz | Hollow mids |
| 2 ms | 250 Hz | Weak low-mids, phasey |
| 10 ms | 50 Hz | Low end cancels; also starts to sound like a slapback |
Parallel distortion: harmonics without losing the fundamental
Distort a band-limited copy hard and tuck it under the clean signal. The dry keeps the sub and the transient; the wet chain contributes harmonics in the range you choose. Far more controllable than driving the whole signal.
- Decide what the dry chain keeps: for bass, sometimes low-pass it at 100–150 Hz so the sub is only clean and everything above comes from the distorted chain. For drums, leave the dry full-range.
- High-pass the wet chain at or above the dry's low-pass so the two don't overlap at the crossover (phase shift there is worst).
- Drive the wet chain much harder than you'd dare on a serial chain. Level it at −12 dB, then raise until the harmonics are noticeable, then back off.
- Post low-pass the wet chain. Cabinet is a good low-pass with character.
Parallel reverb inside a rack vs on a return
Reverb is always parallel in effect (dry/wet), but where it is parallel matters. In a rack chain, the reverb sits after the track's own processing and before the track's fader: the reverb is part of the sound and follows the fader. On a return it's shared, post-fader by default, and processed once for several sources. Rack reverb for a sound-design element (a pad that is its own space); return reverb for a mix space.
| Aspect | Rack chain | Return |
|---|---|---|
| Follows track fader | Yes (pre-fader relative to mixer, but it is in the device chain) | Send is post-fader by default; can be pre |
| Sidechain the reverb | Easy: Compressor in the chain, key = the track | Key from the source track, Post FX |
| Shared space | No | Yes, same room for several sources |
| CPU | One reverb per track | One reverb for all |
| Freeze / Bounce in place | Reverb is baked | Reverb stays live |
Multiband via racks + EQ: split, process, sum
Three chains, each with an EQ Eight that isolates a band, then whatever process you want per band. It's the most flexible multiband and the one most likely to cause phase problems, because crossover filters shift phase around their cutoff. Two solutions: use 12 dB/oct slopes with matching cutoffs on adjacent chains (Linkwitz-Riley-ish, sums close to flat), or use EQ Eight in Linear Phase mode so the sum is phase-exact at the cost of latency.
- Build the three chains with EQ Eight only, all processes bypassed. Set the same mode (Linear Phase, or all minimum phase with 12 dB slopes) on all three.
- Compare rack on/off with a polarity-inverted dry copy on a separate track: it should null or nearly so. If not, the crossover isn't summing flat; adjust cutoffs so adjacent bands cross at −6 dB.
- Add processing per band. Utility Mono on the low chain is free mono-bass.
Racks + EQ vs Multiband Dynamics vs Roar Multiband
Three devices, three sets of trade-offs. Multiband Dynamics gives you dynamics per band with proper crossovers and a sidechain, and its bands can be soloed. Roar Multiband gives you a saturation stage per band with a filter and modulation. The rack gives you anything per band but makes you manage phase.
| Rack + EQ Eight | Multiband Dynamics | Roar Multiband | |
|---|---|---|---|
| Bands | Any number | 3 | 3 |
| Crossover quality | Yours to get right | Clean, designed to sum flat | Clean |
| Per band | Any device chain | Compression / expansion up & down | One saturation stage + level + tone |
| Sidechain | Per-device, anything | External, applies to all bands | Env Follower via sidechain (12.2) |
| Latency | Linear Phase: yes; min-phase: none | Small, compensated | None noticeable |
| Per-band width / mono | Yes (Utility per chain) | No | No (use Mid-Side mode instead) |
| Best for | Sound design, mono low + wide high, odd processes per band | Mixing: dynamic EQ, de-essing, multiband sidechain, upward compression | Colour: drum bus, bass, mix-bus harmonics |
Return vs rack parallel: which one and when
Returns are parallel paths shared across tracks, post-fader by default, with their own sends to other returns (enable in the return's send section) and their own automation lane. Racks are per-track and travel with the sound (Freeze, Bounce in place, presets, Push macro control). Rule of thumb: effects that define a space (reverb, dub delay) go on returns; effects that define a sound (parallel compression, distortion, width) go in racks.
| Need | Use | Why |
|---|---|---|
| NY compression on the drum bus | Rack | Timing-critical; must freeze with the group |
| Dub delay on chords, snare, vocal | Return | One Echo for all, sends automated per clip |
| Parallel distortion on bass | Rack | Part of the patch; macro on Push |
| Shared room reverb | Return | Glues sources into one space |
| Reverb that is the sound (shimmer pad) | Rack | Resamples with the track |
| Sidechain-ducked reverb/delay to kick | Return + Compressor keyed from kick | One compressor cleans every send |
| Performance effects (Beat Repeat, filter) | Rack on the group, or return with send automation | Rack for per-stem, return for a global "FX send" macro |
Wet/dry rules and the Dry/Wet knob inside devices
Most Live effects have a Dry/Wet control (Compressor, Saturator, Drum Buss, Roar, Redux, Pedal, Hybrid Reverb, Echo, Delay, Chorus-Ensemble, Phaser-Flanger, Auto Filter, Spectral devices…). It's an internal parallel path that is guaranteed latency-aligned. Use it before building a rack: a Compressor at 40% wet is NY compression with no phase question.
| Effect type | Typical Dry/Wet | Note |
|---|---|---|
| Compression (parallel) | 30–60% | Compressor's own Dry/Wet; Glue has none, rack it |
| Saturation on bass/drums | 50–80% | Keeps transient and fundamental clean |
| Saturation on mix bus | 100% at very low drive, or 30% at moderate | Both work; the second is more controllable |
| Chorus / Phaser on bass | 20–35% | Fully wet modulation on a bass goes mono-incompatible fast |
| Reverb on a track (not return) | 10–30% | On a return: 100% |
| Delay on a track | 15–35% | Echo's Dry/Wet; on returns 100% |
| Redux / Erosion | 20–50% | Full wet is an effect, half wet is a texture |
| Beat Repeat | Mix mode: Mix / Ins / Gate | Not a percentage; Ins replaces dry while repeating |
Parallel chains as a performance structure
Instead of summing chains, use the chain selector to crossfade between parallel versions of the same processing with a macro: chain 1 clean, chain 2 compressed, chain 3 destroyed, with overlapping zones so the macro morphs through them. Each chain is a complete alternative, latency-aligned by the rack.
Settings per source / genre
| Source | Parallel chain | Band | Level under dry |
|---|---|---|---|
| Drum bus (techno, DnB) | Compressor 10:1 0.1 ms / 50 ms + Saturator | HP 80 Hz | −10 to −6 dB |
| Drum bus (house) | Glue 4, 1 ms, 0.1 s, 8 dB GR | HP 100 Hz, LP 10 kHz | −12 dB |
| Sub + mid bass | Roar Serial Diode, Cabinet | HP 150 Hz, LP 6 kHz | −8 dB; dry LP 150 Hz |
| Reese | Pedal Distortion + Chorus-Ensemble | HP 200 Hz | −6 dB, sides only via Utility Mid/Side in rack |
| Snare | Compressor smash + Hybrid Reverb 0.6 s gated | HP 200 Hz | −8 dB |
| Vocal | Compressor 8:1 fast + Saturator Analog Clip 6 dB | HP 150 Hz, LP 8 kHz | −10 dB |
| Pad | Hybrid Reverb Shimmer 100% + Roar sides Fold | HP 300 Hz | −9 dB |
| Dub chord | Echo 3/16 + Dynamic Tube, keyed Compressor from dry | HP 250 Hz, LP 5 kHz | −6 dB (it is the sound) |
| Mix bus | Compressor 4:1 30 ms RMS, 6 dB | HP 120 Hz | −14 to −10 dB, 20–30% wet equivalent |
Common mistakes
- Unaligned chains. Hardware returns and some M4L devices don't report latency; the result is a comb filter. Measure and use Align Delay.
- Minimum-phase 24/48 dB crossovers in a DIY multiband. The sum has dips at the crossovers. Use 12 dB slopes or Linear Phase.
- Wet chain louder than dry. Parallel is a seasoning; past −6 dB you've built a serial chain and lost the point.
- Dry signal on a return. Return devices at less than 100% wet double the dry signal with a latency offset.
- Overlapping bands with no filters. A full-range crushed chain under a full-range dry chain pumps the sub and smears transients; band-limit the wet chain.
- Stereo-widened wet chain under a mono dry on bass. The sum is mono-incompatible; keep parallel bass chains mono or Mid-only.
- Freezing a track whose parallel reverb is on a return. Freeze doesn't include returns; Bounce in place with sends or resample.