Plugin workflows
Multiband compression crossovers: control a region without dividing the mix blindly
Choose useful band boundaries, compare crossover processing modes, and check whether frequency-dependent compression solves a changing tonal problem.
Multiband compression is useful when one frequency region changes too much with level while the rest should remain comparatively untouched. If that region is excessive throughout the song, compare ordinary EQ first. If the offending bass track is still editable, changing that track avoids also treating the kick, piano and voice that share its frequencies.
By Spectraforge Audio3 min read
Published
Guide review: Captum Records and AS Mastering. AI-assisted writing and source checks. How these guides are made.
Use a band for a changing region, not a permanently wrong balance
This guide focuses on band boundaries and their consequences. The separate dynamic-EQ guide covers a different choice: shaping a selected EQ band dynamically rather than assuming every changing tonal problem needs a divided full mix.
Move the boundary between two competing costs
Suggested exercise: choose a passage where one bass note blooms and its neighbor does not. Create one low band and audition it. Move the upper boundary while listening to both notes, then return to the full mix. A boundary too low can miss part of the bloom; one too high can pull kick attack, low vocal body or piano into the correction.
Illustrative comparison: try an upper boundary near 100 Hz and another near 200 Hz, keeping other settings unchanged at first. These are deliberately different audition positions, not prescribed bass splits. Ask what extra musical material enters the controlled region at the higher boundary. A crossover has a transition slope; it does not assign entire instruments to one side of a vertical line.
Bound the reduction and check the following note
Set threshold so the bloom receives useful control, then use a reduction-range limit if available. In Pro-MB’s downward-compression mode, a negative Range limits attenuation; a positive Range changes the processing direction. Confirm both mode and sign before treating Range as a simple amount knob.
An illustrative maximum reduction of 2 dB can make a first comparison easier to judge than an unconstrained setting. It is not a target. Tune recovery against the next ordinary note and the kick: if either remains reduced unnecessarily, revise timing, threshold or the affected region. Do not compress ordinary notes equally just to make the graph uniform.
Compare the processing method on an exposed attack
Pro-MB documents minimum-phase crossover phase changes, linear-phase latency and possible pre-ringing, and a Dynamic Phase method whose phase changes accompany active gain processing. These are product-specific implementations. They do not make linear phase automatically preferable, nor do all multiband plugins provide the same choices.
Save the settings and compare modes on an exposed kick and a dense passage at similar loudness. If a mode changes the front before the intended reduction becomes useful, compare a gentler boundary or another mode. Also check monitoring latency before putting that choice in a recording path. Read timing units carefully: Pro-MB presents material-dependent attack and release as percentages, not interchangeable millisecond values.
Test the same boundary on another instrument
Play a tom fill or piano section that shares the band. Does the correction still solve the bass bloom without thinning that unrelated material? Compare the complete processor against bypass and against a source-level edit or dynamic EQ treatment. Keep the solution whose improvement survives those passages. If every new section requires another compensating band, the original target may be better handled on its source track.
Technical references
- FabFilter Pro-MB: band display and boundary workflow
- FabFilter Pro-MB: processing mode tradeoffs
- FabFilter Pro-MB: threshold, range and band timing
Listening exercises are proposed practice routines. Worked examples illustrate the stated setup; they are not reports of our listening tests. Editorial process and corrections.