Input shaping vs manual speed tuning

Input shaping vs manual speed tuning

Every printer that runs fast eventually shows the same artifact: corners that echo. Print a thin-wall box and the surface ripples just past every corner, or a sharp edge repeats itself two or three times like a fading shadow. In the shop we call it ringing or ghosting, and it is the most common reason a print looks "almost right but not quite." There are two honest ways to fight it: slow the machine down, or teach the firmware to cancel the vibration. Klipper's input shaper is the second option, and this post is about when it is worth the setup — and when it simply is not.

Why corners ring

A printer's head does not glide through corners; it accelerates. To change direction, the motors reverse torque and the toolhead snaps the other way. That snap excites whatever natural frequencies the frame, gantry, and toolhead have — the machine rings like a struck bell. The nozzle keeps moving while the frame vibrates, so the vibration gets drawn into the print as surface ripples and repeated corner echoes.

Anything that changes the machine's stiffness or mass changes that frequency: a heavier toolhead, a longer belt run, a loose panel. That is worth remembering later.

Lever one: manual speed tuning

The traditional fix is to slow the motion down. Lower the acceleration, lower the cornering speed (junction deviation or jerk, depending on firmware), and if necessary cap print speed. Gentler moves excite less vibration, so the ringing fades.

Manual tuning has real virtues. It works on every printer ever made, on any firmware, with no extra hardware. It is free, reversible, and it teaches you how your machine actually behaves. The cost is print time: acceleration is a large share of real-world speed on small parts, so dropping accel from 5,000 to 2,000 mm/s² can add hours to a batch of small brackets. In a shop that cost compounds; at home it just tests your patience.

Lever two: input shaping

Input shaping attacks the problem at the source. Instead of asking the motors to reverse as hard as possible, the firmware shapes the acceleration pulse — splitting and timing the move so the energy it injects cancels the machine's own ringing instead of amplifying it. The head still moves fast; the frame never gets excited in the first place.

Klipper implements this as the input_shaper section, configured per axis (X and Y separately, because they rarely ring the same way). You can pick a shaper type by hand — zv, mzv, zvd, ei and the multi-hump ei variants trade corner sharpness against residual vibration — but the right way is to measure. Bolt a cheap ADXL345 accelerometer to the toolhead, run Klipper's resonance test, and the firmware identifies the machine's actual resonant frequencies and recommends a shaper. No accelerometer? There is a visual calibration method: print a ringing test tower, measure the echo spacing, and configure from that. Less precise, still effective.

Once calibrated and verified, input shaping is set-and-forget — until you change hardware.

When input shaping is worth it

  • You run Klipper (or a machine that already ships with it configured). This is the whole ballgame: input shaping lives in firmware, so if you cannot configure the printer, it is not on the menu.
  • Ghosting actually limits you. If you print functional parts with sharp corners and echoes are visible at your current speeds, shaping is the highest-value tuning available.
  • Your mechanics are sound. Tight belts, tight pulleys, solid frame. Shaping cannot rescue a loose machine.
  • You want speed without hand-tuning everything. A calibrated shaper holds clean corners at accelerations you would never dare set by trial and error.

When it is not worth it

  • Corners are already clean. If your test print shows no ringing at your normal settings, shaping buys nothing visible. Do not tune for a problem you do not have.
  • Your bottleneck is not motion. Overhangs limited by cooling, small detailed models you print slowly on purpose, a hotend running at its volumetric flow ceiling — faster motion fixes none of these, so shaping is wasted effort.
  • You cannot configure the firmware. Stock machines without accessible config get manual tuning or nothing.
  • You change hardware constantly. Every toolhead change — new hotend, different fan mount, a heavier spool setup — shifts the resonance. A shaper calibrated last month on different hardware is a guess, not a measurement. If you re-kit the printer weekly, recalibration is real overhead.
  • The real problem is mechanical. Worn belts, loose pulley grub screws, sloppy idlers show up as layer shifts and lost steps, which input shaping does not address — it cancels vibration, it does not recover missed motion. Worse, calibrating on top of a mechanical fault bakes the fault into the shaper profile.

The decision path we use

  1. Print a ringing test at your normal settings — a thin-wall box or cylinder with sharp corners, tall enough to show the whole speed range.
  2. Clean corners? Stop. No tuning needed; spend the time elsewhere.
  3. Ghosting? Check mechanics first. Belt tension, pulley grub screws, frame bolts, loose panels. Cheapest fix, and in our experience the most common root cause.
  4. Mechanics solid, Klipper available? Run the accelerometer calibration (or the visual method), apply the recommended shaper, and re-print the test. Expect corners to clean up at speeds that previously failed.
  5. No Klipper? Step acceleration down in roughly 20–25% increments until ghosting fades to acceptable, and keep the speed you can live with.
  6. Write everything down. Speeds, acceleration, shaper type and frequencies, per machine. When quality changes later, a written baseline is the difference between "the tuning changed" and "something came loose."

Mistakes we see

  • Chasing "ghosting" that was actually over-extrusion or a partial clog. Surface artifacts have several causes; confirm the pattern is corner-echo before tuning motion.
  • Calibrating input shaping while a mechanical fault is still present.
  • Never recalibrating. Hardware changed, resonance changed, shaper went stale.
  • Expecting shaping to cancel vibration coming up from the desk or floor. A wobbly stand transmits motion no firmware can shape away; a heavy, stable base is part of the tuning.
  • Assuming the most aggressive shaper is automatically best. Higher-order shapers suppress vibration harder but can slightly soften fine external detail. Measure, print, compare — then choose.

Bottom line

If you are on Klipper, your mechanics are sound, and speed actually matters to you, calibrate input shaping — it is the better tool, and an afternoon of calibration pays back on every print. If your corners are already clean, or the firmware is out of reach, manual speed tuning costs nothing but time and works everywhere. And neither lever fixes a machine with loose belts — fix the hardware first, then tune the software. That order saves more print hours than any setting in this post.

Sources for this week's post

The weekly research digest for this cycle (research/weekly-digest.md, covering 2026-10-20 → 2026-10-26) contained no input-shaping or speed-tuning items, so no external sources were used. Everything above is first-hand shop knowledge — the tuning and diagnostics practices we apply to Klipper and Marlin-family machines at Advanced 3D Printing — not sourced reporting, and no URLs are cited because none were consulted.

Back to blog

Leave a comment

Please note, comments need to be approved before they are published.