Phase 2 · The filament · L13

Retraction

A printer setting with a material-dependent answer — which is why it gets overridden per filament, and why more is emphatically not better.

⏱ ~45 minutes• 🧪 Spike towers, 0.1 mm at a time• Phase 2

◎ Why this lesson

Molten plastic under pressure keeps coming out of a nozzle that has stopped printing. Retraction pulls filament back to kill that pressure before a travel move. It is the setting people reach for first when they see strings — and it is very often the wrong one. Prusa name high printing temperature and moisture as causes of stringing alongside retraction settings.1 Coming fourth in the calibration order is what lets you rule those out first, so that what's left really is retraction.

Retraction lives in Printer Settings → Extruder 1 — it describes a mechanism, not a material.1 But the correct value is entirely material-dependent, so you override it per filament rather than editing the printer profile.

Filament Settings → Filament Overrides

# tick the box beside each value you want to override,
# and only those. Unticked = inherit from the printer.

Retraction length          # almost always override
Retraction speed           # override for flexibles
Retract lift (Z hop)       # override for PETG
Minimum travel after retraction

The tickbox matters. An unticked override inherits, so improvements you make to the printer profile still reach that filament. A ticked one is frozen at whatever you set. Override the values the material genuinely changes, and leave the rest alone.

1Rule out the other three causes first

Before you touch a single retraction setting, walk this list. Every item on it produces strings that look exactly like a retraction problem and will not respond to retraction changes.

Moisture Prusa's own diagnostic is to simply try a different spool — filament that "might have gathered moisture … will cause a lot of stringing".1 In a humid climate this is the first suspect, not the last.
Temperature Prusa suggest dropping the nozzle 5–10 °C and checking whether stringing reduces.1 If it does, your Lesson 10 answer was too hot — go back and fix it there, not here.
A dirty nozzle Printing a long time in one material, PETG especially, leaves a film on the nozzle that strings on its own.1 Clean it before you measure.
Heat creep Prusa point at heat dissipation in the hotend, and at re-applying thermal paste between heatbreak and heatsink.1 This is a Phase 1 machine fault wearing a Phase 2 costume.

Only when all four are ruled out is the remaining stringing yours to tune.

2The settings, and what each one costs

Prusa document the retraction family as follows.1 Read the cost column as carefully as the setting — every one of these buys you something at a price.

Retraction length How far filament is pulled back. The main lever. Cost: too much on a direct-drive machine grinds the filament, drives heat creep, and leaves a gap where extrusion restarts.
Retraction speed How fast the extruder motor pulls. "A higher value improves stringing, but if it is too high it may skip steps in the motor."1 Cost: skipped steps are silent and cumulative.
Lift Z (Z hop) Raises the nozzle during travel so it can't drag through the print. Cost: Prusa note that disabling it may cause the nozzle to hit the printed part1 — so this one has a cost in both directions.
Minimum travel after retraction The travel distance that triggers a retraction at all. PrusaSlicer's preset is 1 mm, which Prusa describe as "a low amount"; raising it lowers print times but increases oozing and stringing.1
Retract on layer change · Wipe while retracting Prusa recommend leaving both on.1 Treat them as defaults you need a measured reason to change.

⌁ Two print settings that reduce stringing without touching retraction

Only retract when crossing perimeters (Print Settings → Infill → Advanced) skips retraction when the travel stays inside the upper layer's perimeter — any ooze lands where it can't be seen. Avoid crossing perimeters (Print Settings → Layers and perimeters → Quality) routes travels to cross fewer walls in the first place.1

Both are geometry-level fixes. They're worth knowing because they solve the visible problem without asking more of the extruder — which is always the better trade on a direct-drive machine.

3Measure it

Print four to six tall thin spikes, spaced about 40 mm apart. Every layer forces a travel between them, so a single short print gives you dozens of retraction events to judge. Adjust length only, in 0.1 mm steps, until the strings go. Do not touch speed until length has stopped improving things.

⬇ Download · the spikes

retraction-spikes.stl — five tapered spikes, 40 mm tall, centres 40 mm apart on a shared 1.5 mm rail so the slender towers stay stuck to the bed. Every layer is four travel moves; twenty minutes of printing gives you hundreds of retraction events to judge.

Open your machine. The starting values and, on the MK3S, the hard ceiling both differ.

MK3S · Bondtech direct drive, and a documented 2 mm ceiling

▲ Prusa's stated maximum

"On the MK2.5/S and MK3/S/+, the retraction length should be a maximum of 2 mm."1 That is a ceiling, not a target. If you find yourself pushing toward it, the problem is somewhere else — go back to section 1.

Working starting points4 — begin here and step down as well as up:

PLA 0.8 mm at 35 mm/s · PETG 1.0–1.2 mm at 35 mm/s · TPU 95A 0.4–0.8 mm at 20 mm/s

Z-lift: 0.4–0.6 mm suits PETG on this machine. Consider dropping it for PLA if you see travel scarring on top surfaces — but remember Prusa's warning that removing it entirely risks the nozzle striking the part.1

For scale, and to show how much geometry matters: the Bowden-fed MINI/+ ships a default retraction of 3.2 mm.1 Your MK3S needs a third of that because the filament path from drive gear to melt zone is a few centimetres rather than half a metre. Never take a retraction number from a machine with a different extruder arrangement.

MK4S · Nextruder, and stock settings worth respecting

The 2 mm figure above is stated for the MK2.5/MK3 family specifically and is not a Nextruder spec1 — but the underlying reason (a very short direct-drive filament path) applies here just as strongly. Expect the answer to be small.

Stock behaviour to know: MK4S profiles ship with Z-lift enabled at 1.5 mm and wipe-on-retract disabled.4 Note that this is the opposite of Prusa's general "leave wipe on" advice1 — the machine-specific profile wins over the general article, and it's worth understanding why before you change it. Confirm both against your own installed profile rather than trusting either source.

Working starting points4:

PLA 0.6–0.8 mm at 35 mm/s · PETG 0.8–1.0 mm at 35 mm/s · TPU 95A 0.4–0.8 mm at 25 mm/s

Keep Z-lift for PETG. PETG oozes stringy and welds to whatever the nozzle drags through; the 1.5 mm hop is doing real work. For PLA you can often drop it to 0.2–0.4 mm and gain a little speed.4

Core One · as MK4S, plus the enclosure's contribution to oozing

Same starting values and same reasoning as the MK4S. Same extruder, same firmware.

What the enclosure adds: warmer ambient air means the extrudate stays soft longer after it leaves the nozzle, so strings that would have snapped on an open-frame machine stay attached. You may measure a slightly longer retraction here than on the MK4S for the same material — and that difference is real, not measurement error.

Run the spike test in the door and vent state you actually use for that material. A retraction value measured with the door open is a value for a machine you don't have.

And keep the diagnostic order straight: on this machine especially, a hot chamber can push the hotend toward heat creep, which strings for a Phase 1 reason.1 If more retraction makes things worse, suspect that before you suspect the number.

4TPU: where two good sources disagree

⚠ A genuine conflict, stated rather than hidden

Prusa say: "Flexible filaments usually need longer retractions, because the material stretches while being pulled back to the nozzle."1

This course's own master guide says: keep retraction on flexibles as low as you can tolerate, because long retracts buckle the filament in the drive path.4

Both mechanisms are real. TPU is elastic, so part of a retraction is absorbed by the filament stretching rather than by the melt zone depressurising — that argues for more. TPU is also soft, so a long fast pull can buckle it inside the extruder path, where it jams — that argues for less. Which one dominates depends on your specific extruder geometry and your specific filament's shore hardness.

How to resolve it: start low and walk up in 0.1 mm steps, watching the extruder as much as the print. The reason to approach from below is asymmetric risk — under-retracting gives you strings you can see and cut off; over-retracting gives you a jam and a strip-down. Log the value you land on and which direction you approached from.

Two more TPU-specific notes, both from this course's guide rather than from Prusa:4 the spool must unwind with almost no resistance — mount it so there is no side pull — and the PTFE path from drive gear to hotend must be smooth and gap-free. A flexible filament will happily coil up in any gap it finds rather than move forward. Neither of these is a slicer setting, and both will masquerade as a retraction problem.

5Find the fault

Click each. Every one of these is stringing that will not respond to a retraction change.

It's stringing. Is retraction actually the problem?

Strings everywhere, with faint popping while printingMoisture
Fine hairs, and the surface is glossier than expectedTemperature
A gap where printing restarts, but no strings at allAdvance
Clicking from the extruder, and worsening under-extrusionGrinding
Oozing that gets worse the longer the print runsHeat creep

6Check yourself

From memory. Question 3 interleaves Lesson 12.

Question 1 of 3

Why does retraction come after temperature in the calibration order?

Question 2 of 3

On TPU, why does the lesson say approach the answer from below rather than above?

Question 3 of 3

A gap appears where extrusion restarts after travel, but there are no strings. Cause?

✎ Your task · do this with your tutor

Spike test, one material, one machine

Walk section 1's four causes and report on each: spool drying date, current nozzle temperature, when you last cleaned the nozzle, and whether stringing worsens through a long print.
Print the spike tower at your current retraction length. Photograph it.
Step length only, 0.1 mm at a time, reprinting each time. Report the value where strings stop and the value where the extruder starts complaining. Both numbers.
Only then, if strings remain, try speed — and say what changed.
Set the winner in Filament Settings → Filament Overrides, tick the override box, and tell me which fields you ticked and which you deliberately left inheriting.

→ I want both numbers from point 3. The distance between "strings stop" and "extruder complains" is your safety margin, and knowing how wide it is tells you how much this profile will tolerate a wetter spool or a warmer day.

What you can now say

Stringing has four common causes and retraction is only one — moisture, temperature and a dirty nozzle come first.
Retraction is a printer setting with a material answer, so it is overridden per filament, selectively.
Prusa cap MK3-family retraction at 2 mm; a Bowden MINI needs 3.2 mm, which is why values never transfer between extruder types.
Length first, in 0.1 mm steps; speed only after. One variable at a time, even inside a single setting family.
On flexibles, approach from below — the two credible sources disagree, and the downside risks are not symmetric.

📖

Primary source · read this next

Stringing and oozing — Prusa Knowledge Base

The full retraction setting list with Prusa's own cost notes, the 2 mm MK3-family ceiling, and — usefully — the non-retraction causes given equal billing. Read the order of their suggestions: nozzle cleanliness and temperature come before settings, which is the ordering this whole lesson is built on.

💬 I'm your tutor — ask me anything. Photos of spike towers are genuinely easy to compare, so send them — and send the pair, before and after, not just the good one. If you think you've hit the TPU conflict in section 4 from the buckling side, describe what the extruder sounds like and we'll work out which mechanism you're in. Confused beats stuck.

Linked: Glossary → retraction, filament overrides, Pressure Advance, heatbreak, extruder, hygroscopic, drying, calibration order, one-variable rule, tuning log. Keep it close with → The Calibration Card.

§References

Prusa Research. Stringing and oozing. That stringing "is caused by very high printing temperatures and/or using incorrect retraction settings"; the nozzle-film cause and the instruction to clean it; retraction settings living in Printer Settings → Extruder 1; that flexible filaments "usually need longer retractions, because the material stretches"; retraction length, Lift Z, retraction speed ("a higher value improves stringing, but if it is too high it may skip steps in the motor"), minimum travel after retraction (preset 1 mm, "a low amount"), retract on layer change and wipe while retracting (both recommended on); the 2 mm maximum for MK2.5/S and MK3/S/+ and the MINI/+ default of 3.2 mm; "Only retract when crossing perimeters" and "Avoid crossing perimeters"; lowering nozzle temperature by 5–10 °C; moisture as a cause; and heat dissipation / thermal paste on the heatbreak-to-heatsink threads. Link.
Prusa Research. Cooling. Referenced here for the interaction between fan settings and heat creep at very low print speeds — the failure mode that mimics an over-retraction jam. Link.
Prusa Research. Configuration and profiles. Filament Overrides as the mechanism for per-filament values that belong structurally to the printer profile. Link.
This course. Master tuning guide, §2.4 and §4 — the spike-tower method, the per-material starting lengths and speeds, MK4S stock Z-lift of 1.5 mm with wipe-on-retract disabled, and the flexibles guidance that conflicts with Prusa's. The conflict is stated in section 4 rather than resolved by picking a side. Link.

Lesson L13 of the prusa-profile-tuning course · all lessons · The Calibration Card · glossary