Phase 2 · The filament · L14
Cooling
Not one fan speed — a small control system with two thresholds, a speed brake and an interpolation between them. Worth understanding rather than copying.
Prusa frame the problem in two halves. Printing must always happen on an already solidified layer, or the new pass drags the old one around. And when printing overhangs or bridges you need to freeze the plastic in place before it sags.1 A fan solves both — and with some materials makes things "actually worse and cause warping or layer separation".1 That is the whole tension of this lesson: the same air that buys you crisp overhangs costs you layer strength.
Everything here lives at Filament Settings → Cooling. And it is not one number: it is a
little control system that decides fan speed from how long each layer takes, and which will also
slow the printer down if a layer is too quick to cool.
With Enable auto cooling on, PrusaSlicer shows a text description above the fan settings that updates with every change of any cooling setting and explains in plain language what the resulting behaviour will be.1 Change a value, read the sentence, change it back. Five minutes of that teaches the interaction better than this lesson can.
1How the fan speed is actually chosen
You set a Min and a Max fan speed, as a percentage of the fan's maximum RPM. Which one gets used — or what blend of them — depends on two time thresholds, not on geometry:1
- Layer time below "Slow down if layer print time is below" Max fan speed is used. The layer is finishing so fast that it needs everything you have — and the printer will also slow the print down to stretch the layer out to that duration.
- Layer time above "Enable fan if layer print time is below" Min fan speed is used. The layer takes long enough that it has time to cool on its own.
- Layer time between the two thresholds The fan runs proportionally, interpolated between Min and Max. This is the case most of your layers are actually in.
Two more overrides sit on top of that. Bridges fan speed replaces the calculated value whenever the printer is bridging or printing overhangs — typically set higher than normal printing.1 And Disable fan for the first X layers suppresses cooling at the start of the print entirely; Prusa note the first layer usually doesn't need cooling since it's on a flat surface, and that skipping a few more prevents warping and detachment.1 Typical values are 1–5.
Keep fan always on guarantees at least the Min speed at all times. Prusa say it is typically on for most materials "with the exception of ABS, ASA, PC, and some flexibles".1 Of the three materials in this course, that exception clause touches only TPU — and only sometimes. Turn it off deliberately and log the reason, never as a default.
With auto cooling disabled entirely, the fan simply runs at the Min speed for the whole print, with bridges still governed by Bridges fan speed.1 That is a useful diagnostic mode — a fixed, known fan speed — but not how you should ship a profile.
2Min print speed, and the heat-creep trap
Slow down if layer print time is below is a brake: if a layer would finish too fast, PrusaSlicer scales print speeds down to stretch it out.1 Min print speed is the floor it will not go below.
Prusa's warning is explicit: "Printing at extremely low speeds for extended periods of time could lead to problems such as heat creep."1 A nozzle crawling over a tiny layer is barely pulling filament through, so heat climbs the filament path instead of being carried away by it — and you get a jam that looks like a retraction problem (Lesson 13).
The counter-warning is on the same page: raising Min print speed too high "may partially invalidate the minimum layer time",1 because even at that floor the layer can still finish faster than you asked for. Both settings are real constraints and they genuinely fight. If a tall thin part comes out droopy and the extruder jams, you are caught between them — the answer is usually to print two copies at once so each layer has more work in it.
3Dynamic fan speeds, and a documentation trap
Dynamic fan speeds vary the fan as a function of overlap percentage — how much of the current extrusion sits on top of the previous layer. Prusa define it clearly: 100 % overlap is a full overlap (no overhang), while 0 % overlap represents a full overhang — a floating extrusion, a bridge. Speeds between the control points are linearly interpolated.1
The relationship you want is less overlap → more fan. A bridge (0 % overlap) has nothing beneath it and needs maximum cooling; a vertical wall (100 % overlap) is fully supported and needs minimal.
The example image in Prusa's own article shows fan speed increasing with overlap, which is
backwards, and several readers have said so in the comments — one of them verified the correct behaviour by
slicing test geometry and reading the resulting M106 values out of the
G-code.1 The text is right; the picture isn't.
Do what that reader did. If you use this feature, slice a test part and read the M106
commands. It takes two minutes and it settles the question for your version of the slicer rather than for
someone else's.
Dynamic fan speeds are optional. Plain Min/Max plus Bridges fan speed handles most of what you need, and it is easier to reason about. Reach for the dynamic curve when you have a specific part with gradual overhangs that the binary bridge/not-bridge distinction handles badly.
4Per material, per machine
The material determines the shape of the answer; the machine determines how far you have to move from it. Open your machine.
MK3S · open frame, and the fan is the whole story
Nothing between the part and the room. Ambient is whatever the room is, and in Mauritius that is warm and humid year-round — expect to need slightly more fan than a temperate-climate profile suggests, and log the room temperature alongside the result.
Starting bands4 — these are entry points to test, not answers:
- PLA — Min/Max 100 / 100 %, bridges 100 %, disable first 1 layer
- PETG — Min/Max 30 / 50 %, bridges 100 %, disable first 2–3 layers
- TPU 95A — Min/Max 50 / 80 %, disable first 1 layer
Why PETG's band is so much lower: PETG's strength comes from layers staying warm enough to weld. Too much fan gives you delamination and poor overhangs; too little gives you drooping and stringing.4 It is the one material in this course where the correct answer is genuinely in the middle rather than at an extreme.
This machine's slower speeds mean longer layer times, which means the auto-cooling thresholds fire less often than on the MK4S. That is a real difference: a threshold pair copied from a fast machine will behave differently here even with identical fan percentages.
MK4S · open frame, but fast enough that the thresholds matter more
Starting bands4:
- PLA — Min/Max 100 / 100 %, bridges 100 %, disable first 1 layer
- PETG — Min/Max 40 / 60 %, bridges 100 %, disable first 2–3 layers
- TPU 95A — Min/Max 50 / 80 %, disable first 1 layer
The MK4S-specific point is the time thresholds, not the percentages. This machine prints the same layer in a fraction of the MK3S's time, so far more of your layers fall below "Slow down if layer print time is below" — which means Max fan more often, and the speed brake engaging more often. A part that prints at 40 mm/s despite a 120 mm/s profile is usually not a max volumetric speed problem (Lesson 15) — it is the cooling brake.
How to tell them apart: slice and look at the preview. If the slow regions are small layers near the top of a tapering part, it's the cooling brake. If they're large flat infill regions, it's MVS.
Core One · enclosed, and cooling now has two knobs
Starting bands4:
- PLA — Min/Max 100 / 100 %, bridges 100 %, disable first 1 layer, vent open
- PETG — Min/Max 30 / 50 %, bridges 100 %, disable first 2–3 layers, closed
- TPU 95A — Min/Max 40 / 70 %, disable first 1 layer, closed
Note the inversion. PETG's fan band here is lower than the MK4S's (30–50 % against 40–60 %) even though the two machines share a hotend — because the enclosure is already doing part of the job of keeping layers warm. And PLA needs the opposite: the same enclosure that helps PETG is actively working against you, so it wants the vent open and full fan.
The working targets — vent open with the chamber at or below 30 °C for PLA, closed at 35–40 °C for PETG — come from community practice and this course's own guide, not from a Prusa KB article.4 Prusa's cooling documentation says nothing about chamber temperature at all.1
So treat them as a hypothesis you are testing, log what your chamber actually reached, and be willing to find that your machine in your room wants something different. Remember from Lesson 2 that PLA's glass transition is low enough that a hot chamber softens the part itself — that part is physics, not opinion.
Record the door and vent state with every cooling number you log. On this machine a fan percentage without a chamber state is not a measurement.
5Work out what the fan will actually do
Answer each before opening it. If you can predict these, you understand the pane. Then use the calculator to sweep layer time across the whole range and watch where the two thresholds bite — the interpolation is the part nobody has an intuition for until they've seen it move.
Four settings from Filament Settings → Cooling, plus how long one
layer of your part actually takes. Drag the layer time; the answer is what PrusaSlicer will write into
the M106 for that layer.
Drill 1 · Min 40 %, Max 60 %, enable-fan-below 60 s, slow-down-below 15 s. Layer takes 8 s.
Max — 60 %. And the print also slows down.
8 s is below the 15 s "slow down" threshold, so two things fire at once: the fan goes to Max,1 and PrusaSlicer scales print speeds to stretch the layer out toward 15 s — down to the Min print speed floor, no further.
This is the case people misread as a slicing bug. The profile says 120 mm/s, the preview says 45, and nothing in the speed pane explains it. The explanation is in the cooling pane.
Drill 2 · Same settings. Layer takes 35 s.
Interpolated — roughly 51 %.
35 s sits between the two thresholds, so the fan runs proportionally between Min and Max.1 Roughly: 35 s is a bit under halfway from 60 s down to 15 s, so you land a bit above the midpoint of 40–60 %.
The exact number matters less than the shape: most of your layers live in this band, which is why Min and Max both need to be sensible values rather than one real setting and one placeholder.
Drill 3 · Same settings, but the layer contains a bridge. Bridges fan speed is 100 %.
100 % — while bridging. The rest of the layer uses whatever the time-based calculation gave.
Bridges fan speed overrides the calculated value for bridges and overhangs specifically.1 It is not a per-layer setting; the fan changes within the layer and changes back.
This is why bridges fan speed can sit at 100 % for PETG even when the normal band is 30–50 %. The two numbers describe different moments, and the delamination risk that keeps PETG's normal fan low simply doesn't apply to an extrusion with nothing underneath it to delaminate from.
6Check yourself
From memory. Question 2 is the one that saves you an evening.
overhang-bridge-test.stl — four fins leaning 30°, 40°, 50° and 60° from vertical plus a 25 mm bridge, on one small plate. The fin where the underside turns from clean to shaggy is your overhang limit at this fan band; the bridge underside is judged separately because Bridges fan speed overrides everything else there.
One overhang test, and read the sentence
- Open
Filament Settings → Coolingfor your material and paste me the live description text that PrusaSlicer generates. That sentence is the ground truth for your profile. - Print an overhang and bridge test at your material's starting band. Photograph the underside of the bridge and the steepest overhang that survived.
- Change only Min fan speed, by 10 points, in the direction you think helps. Reprint. Report what changed — including anything that got worse.
- Snap a piece off both prints and compare the fractures. More fan should make the break cleaner along layer lines; that is the cost you are paying.
- Core One only: record the chamber temperature at the start and end of each print, and the door and vent state.
→ Point 4 is the one people skip. Cooling is a trade, and if you can't show me what the extra fan cost you in layer strength, you've optimised one side of it blind.
What you can now say
- Fan speed is chosen by layer time, not geometry — Max below one threshold, Min above the other, interpolated between.
- Bridges fan speed overrides the calculation mid-layer, which is why it can be 100 % while the normal band is 40 %.
- The cooling brake can slow a print far below its profile speeds, and Min print speed is the floor that keeps it out of heat creep.
- Dynamic fan speed runs inverse to overlap — 0 % overlap is a bridge and wants the
most air. Verify it in
M106output; the KB's example image is backwards. - An enclosed machine needs a different band, not a scaled one — less fan for PETG, more for PLA, and a chamber state logged with every number.
§References
- Prusa Research. Cooling. The two cooling problems (printing on a solidified layer;
freezing overhangs and bridges); that a fan can with some materials "make things actually worse and cause
warping or layer separation"; "Keep fan always on" and its exception list of ABS, ASA, PC and some
flexibles; "Enable auto cooling" and its live-updating text description; the Min/Max fan speed rules
against the "Slow down if layer print time is below" and "Enable fan if layer print time is below"
thresholds with proportional interpolation between them; Bridges fan speed as an override; "Disable fan for
the first X layers" (typically 1–5); dynamic fan speeds with 100 % overlap = no overhang and 0 % = full
overhang, linearly interpolated; Min print speed and the note that "printing at extremely low speeds for
extended periods of time could lead to problems such as heat creep"; and that raising Min print speed too
high "may partially invalidate the minimum layer time". The misleading dynamic-fan-speed example image and
its correction by
M106inspection are from the article's reader comments, not from Prusa. Link. - Prusa Research. Max volumetric speed. That layer-time cooldown is one of the four limits on final print speed, alongside per-feature speed, MVS and overhang slowdown — the framing used to distinguish the cooling brake from an MVS cap. Link.
- Prusa Research. Stringing and oozing. Heat creep, and re-applying thermal paste between heatbreak and heatsink — the failure mode Min print speed exists to avoid. Link.
- This course. Master tuning guide, §2.5 and §4 — the per-material fan bands for all three machines, the PETG delamination-versus-drooping trade, and the Core One chamber guidance. The chamber numbers are community consensus and are flagged as such in section 4. Link.
Lesson L14 of the prusa-profile-tuning course · all lessons · The Calibration Card · glossary