You swap to a new colour and the first layers come out contaminated — a white part with a grey tint that fades over twenty layers, or a yellow one that starts out green.
A melt zone is not a pipe with a valve
The volume of molten plastic inside a hotend is small, but flow through it is nowhere near a clean plug.
Material at the centre of the channel moves fastest. Material against the wall barely moves at all, held there by viscous drag, and it is released only gradually as the bulk flow scrubs past it. There is also a wider cavity above the nozzle's cone where material can sit in a slow eddy.
So a colour change is not one melt-zone volume of purge. It is several times that, and the tail is long: the last few percent of the old colour clears very slowly indeed.
Direction matters enormously
Pigment loading is not symmetrical in its effect. A small residue of black in white is instantly visible; the same residue of white in black is not.
That means the purge volume needed depends on which way you are going. Changing from a light colour to a dark one is nearly free. Changing from dark to light — and especially from black or a saturated red to white or a natural translucent — needs several times the purge, and it is where every complaint about this comes from.
Where the design allows it, sequencing a multi-colour print light to dark removes most of the problem without changing any setting.
Getting the purge right
- Increase purge volume for dark-to-light transitions specifically. Slicers that support a purge volume matrix let you set it per colour pair rather than globally, which is the efficient way to do this.
- Use a purge tower or a wipe object. It gives the nozzle somewhere to unload that is not the part, and it also keeps the nozzle primed and at pressure between changes.
- Purge into infill or into supports if your slicer supports it. This is the least wasteful option available, because the material is going into the part anyway where nobody will see it.
- Purge more when changing material type, not just colour. Different polymers have different melt viscosities and different temperatures, and a residue of one in the other prints badly regardless of colour — a stubborn slug of polycarbonate left in a hotend now running PLA will not melt at all.
The cost, which is not small
Purge is filament you bought and did not print. On a four-colour print with frequent changes, the purge tower can genuinely weigh more than the model.
That belongs in any estimate for work you are pricing: the slicer's headline material figure often reports the model alone, with the tower listed separately or not at all. If you are quoting multi-colour work, weigh the tower along with the part and put the total into the cost calculator — see the print cost more than expected for the other estimate gaps that stack on top of this one.
Hardware makes a real difference
A high-flow hotend with a large melt zone holds more material and takes longer to clear it, so the same purge setting that works on one machine will not on another. Nozzle geometry matters too: a wide internal cone retains more than a narrow one.
If you print multi-colour work regularly and the purge is the dominant cost, this is one of the few cases where hardware choice genuinely changes the economics rather than just the convenience.
When it is not purge at all
A colour that never fully clears, however much you purge, is not bleeding — it is residue stuck to the wall of the melt zone or lodged in a nozzle that needs cleaning. Run a cold pull and inspect what comes out; see clogged nozzle.
Streaking that appears without any filament change is a different fault entirely, usually a degraded fragment breaking loose from somewhere in the hotend.
Pigments and additives that cling longest
Anything heavily pigmented, and translucent grades in the receiving position — a natural or clear filament shows contamination that would be invisible in an opaque one. Silk and metallic grades carry additives that cling to the melt zone and take longer to clear than plain filament of the same colour. Glow-in-the-dark grades are the extreme case: the phosphor particles are abrasive and slow to flush, and they will haunt the next few prints if the purge is skimped.