Phase 2 · The filament · L16
Speeds and accelerations
Last, deliberately. Everything above has already decided how fast you may go — this lesson is about spending that budget in the right places.
Speed is the setting everyone reaches for first and it belongs eighth. By now the hotend's ceiling is measured (Lesson 15), the cooling brake is configured (Lesson 14), and the pressure compensation is tuned for a specific acceleration (Lesson 12). Those three between them have already set the envelope. What is left is a distribution problem: given a fixed budget, which features do you spend it on? The answer is not "all of them equally", and getting it right is the difference between a fast print and a fast print that looks fast.
1The hierarchy, in Prusa's own words
"Infill is usually printed the fastest, perimeters are printed slower and the external perimeter is printed the slowest for the best surface finish and accurate dimensions."1
That single sentence is the whole strategy. Nobody sees your infill. Everybody sees your external perimeter, and it is the surface your calipers land on when you check dimensional accuracy. So the external perimeter gets the slow, careful treatment and the infill gets whatever is left.
If you change exactly one number in this lesson, change this one. Keep it low even when everything else is fast — the time cost is small because external perimeters are a modest fraction of total extrusion, and the quality gain is the one that shows.4
The same logic applies to acceleration: keep external-perimeter acceleration well below infill acceleration. On the MK4S and Core One, Input Shaper lets you run high acceleration generally — but corners and seams still degrade first, so the exception still earns its place.4
2The speed you set is not the speed you get
Prusa are direct about this: print speeds are affected by the printer's acceleration and jerk settings, and when printing small models "it's possible the printer will never actually reach the target speeds".2
A 120 mm/s perimeter setting on a 20 mm cube is fiction. The toolhead spends the whole wall accelerating and decelerating and never touches 120. Which means two things worth internalising:
- On small parts, acceleration is the speed setting Raising the speed number changes nothing; raising acceleration changes everything — including surface quality, which is why it degrades faster than the speed number suggests it should.
- Use the Speed preview, not the profile, to know what you've built Switch the preview mode to Speed to inspect the actual print speeds after all accelerations are applied.2 This is the companion to the Volumetric flow rate view from Lesson 15 — same habit, different question.
- Benchmark on a real part, not a cube A calibration cube's speeds bear almost no relation to a 150 mm bracket's. Judge a speed profile on geometry you actually print.
Set travel high. It costs nothing in quality once retraction and advance are right, and on parts with many small islands it is a large fraction of wall-clock time.4 If raising travel speed does hurt quality, that is a retraction or advance problem showing itself (Lessons 12 and 13), not a reason to slow travel down.
3Two features that spend the budget for you
Dynamic overhang speed
Slows the print when printing overhangs so cooling has time to work. The algorithm calculates extrusion overlap with the previous layer and applies a speed from an overhang slowdown function that you shape via four control points — each point pairing an overlap percentage with a desired speed, linearly interpolated between.2
Turn it on. Without it, the only way to make overhangs survive is to slow the entire print down to the speed the worst overhang needs — which means paying for the hardest 2 % of the geometry across 100 % of it.4 Note the shared vocabulary with dynamic fan speeds in Lesson 14: both are functions of overlap percentage, both use four points, both interpolate linearly. They are the same idea applied to two different actuators, and they should agree with each other.
Pressure equalizer
Available from PrusaSlicer 2.5, at Print Settings → Speed. It smooths sudden changes in speed
between two features — infill to perimeter, for example — because extruders "love to extrude at a constant
rate for the best extrusion consistency" and a print is anything but
constant.3
- Max volumetric slope positive The acceleration limit going from lower to higher speed.3
- Max volumetric slope negative The limit going from higher to lower speed.3
- Lower value = smoother transition Prusa suggest values between 2 and 10, and note the transitions add a little print time.3
- If infill-after-perimeter costs too much time Set Max volumetric slope positive to zero to suppress that slowdown specifically.3
Prusa draw the line precisely. Linear Advance "predicts the pressure build-up in the extruder … and decreases the amount of filament extruded just before stopping and decelerating" — in short, it affects extrusion. The pressure equalizer "smooths the speed changes when transitioning from printing one feature to another" — in short, it affects print speed.3
They work together, and Prusa note the equalizer "can make the job of Linear Advance easier … by reducing the number of sudden fast movements".3 So if you enable it now, after tuning advance in Lesson 12, you have changed the conditions your advance value was measured under. Enable it, then look at your corners again.
4Starting values per machine
These are entry points to test against the stock system profile, not answers. Always compare against the stock Prusa profile for that printer before deviating — and if a value below differs a lot from stock, trust stock and tune from there.4 PrusaSlicer has a built-in preset comparison tool for exactly this job.5
Open your machine.
MK3S · accept that it is the slow one, and win on quality
A bedslinger with no Input Shaper. The Y axis is carrying your part, and the moving mass grows as the part does. There is no filter for mass.
Starting values (0.4 nozzle, 0.2 mm layer)4:
- PLA — external perimeter 30, perimeters 45, solid/infill 50/70, travel 180 mm/s
- PETG — external perimeter 25, perimeters 40, solid/infill 40/55, travel 180 mm/s
- TPU 95A — external perimeter 15, perimeters 20, solid/infill 20/25, travel 120 mm/s
The strategic advice for this machine: accept ~45–60 mm/s perimeters and get quality from there, rather than pushing it to MK4S numbers and fighting ghosting.4 An MK3S printing beautifully at 45 mm/s is a more useful machine than an MK3S printing acceptably at 90.
Bridges: 25–30 mm/s, with bridge flow ratio ~0.95 for PLA and ~0.9 for PETG.4
And remember from Lesson 15: at 0.2 × 0.45, this machine's flow ceiling corresponds to roughly 140 mm/s. You will hit the motion limit long before the flow limit. On this printer, MVS is a safety net rather than the binding constraint.
MK4S · Input Shaper buys acceleration, not corners
Still a bedslinger — the bed and your part still move on Y — but with Input Shaper cancelling the frame resonances that would otherwise limit acceleration.6
Starting values4:
- PLA — external perimeter 45, perimeters 120, solid/infill 150/200, travel 400 mm/s
- PETG — external perimeter 40, perimeters 100, solid/infill 120/160, travel 400 mm/s
- TPU 95A — external perimeter 25, perimeters 35, solid/infill 35/45, travel 250 mm/s
Notice the ratio. External perimeter is 45 while perimeters are 120 — a factor of nearly three. On the MK3S the same ratio is 30 to 45. The faster the machine, the wider the gap should be, because the external perimeter's requirements haven't changed at all while everything else's ceiling has risen.
On acceleration: Input Shaper lets you run high accelerations, but corners and seams still degrade first.4 Keep external-perimeter acceleration well below infill acceleration even here. And note Prusa's own diagnostic from Lesson 8 — an Input Shaper frequency outside 35–70 Hz means go back and check belts, not raise acceleration.6
Bridges: 50–80 mm/s.4
Core One · CoreXY, where the acceleration budget is genuinely larger
CoreXY: two fixed motors, only the light toolhead accelerates, and the part's mass never enters the motion equation. This is the one machine where the motion system is not your first constraint.
Starting values4:
- PLA — external perimeter 45, perimeters 130, solid/infill 150/220, travel 450 mm/s
- PETG — external perimeter 40, perimeters 110, solid/infill 120/170, travel 450 mm/s
- TPU 95A — external perimeter 25, perimeters 40, solid/infill 40/50, travel 300 mm/s
The interesting consequence: this is the machine most likely to be limited by MVS rather than by motion. Its flow ceiling and the MK4S's are comparable, but it can actually reach the speeds that demand them. Check the Volumetric flow rate preview here first — on this printer it is the binding constraint far more often than on the other two.
And the gantry-squareness caveat from Lesson 6 comes back: CoreXY squareness depends on the two belts being tensioned equally, within 8 Hz of each other. High acceleration on an unequally tensioned gantry produces dimensional error that no speed setting fixes. If you're about to raise acceleration substantially, re-measure the belts first.
Bridges: 50–80 mm/s.4 And remember the chamber — bridges in warm still air sag more than bridges in room air.
5The one legitimate loop back
Advance compensates for a pressure lag whose magnitude scales with how hard the machine accelerates. You tuned it in Lesson 12 at production accelerations precisely so this wouldn't happen — but if this lesson has moved your accelerations substantially, that assumption has broken.
Re-run the advance tower. It is the only step in the calibration order that is allowed to loop, and this is the loop. Everything else stays forward-only: change a step and everything below it is void.
The cheap version of this check: print your corner-test object at the new accelerations and look at one corner. If it's still clean, you're done. If a bulge or a gap has appeared, run the full tower.
6Sign Phase 2 off
One material, one machine, one spool. Everything measured, everything logged, and every number attributable to a print you can point at.
That last item is the one that matters most in six months, and it is what Lesson 19 is about. A profile full of numbers where you can't tell which were measured and which came from a table is a profile you cannot debug.
Phase 2 ends here. Next comes turning this pile of measurements into something durable: a named, machine-scoped preset that can't end up on the wrong printer, a validation set that tells you when it's finished, and a log that keeps three printers honest.
7Check yourself
From memory. Question 3 interleaves the whole of Phase 2.
Close Phase 2 on one machine and one material
- Set your speed values from the panel for your machine. Then open PrusaSlicer's preset comparison against the stock system profile for that printer and report every value where you differ by more than 20 %, with your reason for each.
- Slice a real part you actually print. Screenshot both the Speed preview and the Volumetric flow rate preview.
- Tell me which of the four limits from Lesson 15 is binding, and where.
- Print one corner-test object at your new accelerations. Is the advance value from Lesson 12 still right? Show me a corner.
- Complete the sign-off checklist above and paste the finished Phase 2 block from your log.
→ Point 1 is the assessment. Deviating from stock is fine — that's the whole course — but every deviation should have a measurement behind it. If you can't name the measurement, take the stock value back.
What you can now say
- External perimeter slowest, infill fastest — and the faster the machine, the wider that gap should be.
- Acceleration is the real speed setting on small parts, where the target speed is never reached at all.
- The Speed and Volumetric flow rate previews tell you what you actually built, which the profile does not.
- Pressure equalizer affects speed; advance affects extrusion. They cooperate, and enabling one changes the conditions the other was measured under.
- Changing acceleration is the one sanctioned loop back in the calibration order. Everything else stays forward-only.
§References
- Prusa Research. Pressure equalizer. "Infill is usually printed the fastest, perimeters are printed slower and the external perimeter is printed the slowest for the best surface finish and accurate dimensions." Link.
- Prusa Research. Speed settings. That print speeds are affected by the printer's acceleration and jerk settings and that on small models "it's possible the printer will never actually reach the target speeds"; the Speed preview mode for inspecting actual speeds after accelerations are applied; and dynamic overhang speed with four control points pairing extrusion overlap percentage with a desired speed, linearly interpolated. Link.
- Prusa Research. Pressure equalizer. Available from PrusaSlicer 2.5; located in
Print settings → Speed; Max volumetric slope positive and negative; that lower values give smoother transitions and Prusa "suggest using values between 2-10"; setting the positive slope to zero to suppress excessive slowdown of infill after perimeters; that it needs no firmware changes; and the distinction that Linear Advance "affects extrusion" while the pressure equalizer "affects print speed", with the equalizer able to "make the job of Linear Advance easier … by reducing the number of sudden fast movements". Link. - This course. Master tuning guide, §2.7, §3 and §4 — the per-machine per-material speed tables, the external-perimeter and acceleration rules, dynamic overhang speed, travel speed, bridge speeds and flow ratios, the MK3S strategic advice, and the instruction to compare against the stock system profile before deviating. Link.
- Prusa Research. Compare presets. PrusaSlicer's built-in tool for diffing two presets — the practical way to audit your profile against the stock one. Link.
- Prusa Research. Input Shaper. What it cancels, that it is coordinated with pressure advance compensation, and the 35–70 Hz expected frequency band whose violation points back at belt tension and assembly. Link.
Lesson L16 of the prusa-profile-tuning course · all lessons · The Calibration Card · glossary