If you run a fast machine in a shared makerspace, you have probably watched it go through a long straight line smoothly, only to shudder violently as it changes direction at the end of the wall. The table rattles, prints pick up faint ripples near corners, and you find yourself quietly lowering every speed setting just to keep the noise down.
The frustrating part is that slowing the printer usually makes it look tame again — but it also throws away the throughput you bought the machine for in the first place.
Here is the part most troubleshooting guides get wrong: your printer isn't shaking because it's going too fast. It's shaking because of how quickly it accelerates. Top speed and acceleration are two completely different things, and understanding the difference is the key to fixing high-speed shaking without sacrificing speed.
Key Takeaway: High-speed shaking is a resonance problem, not simply a "too fast" problem. Acceleration injects the energy that excites resonance, and top speed is largely irrelevant to it. Fix it with a reproducible order: get the mechanics right, tune belts by frequency, then let Klipper's Input Shaping cancel what's left.
Why Acceleration — Not Speed — Makes a Printer Shake
When your toolhead travels at 300 mm/s in a straight line, the motion system isn't working particularly hard. It's just holding a steady state, like a car cruising on a flat motorway.
The violence happens at the start, the corners, and the stop. Every time the toolhead changes direction, the motors have to reverse the moving mass of the gantry, the toolhead, and (on a bed-slinger) the print bed. That reversal injects a pulse of force into the frame. The bigger the acceleration, the harder that pulse.
This is why a printer can run at a low speed with aggressive acceleration and shake hard, while a different machine can run much faster with gentle acceleration and stay calm. As one of the clearest explanations of the phenomenon puts it in this practical guide to fixing ringing and ghosting, you could run at 200 mm/s with low acceleration and see almost no ringing — or run at 80 mm/s with aggressive acceleration and get noticeable ghosting.
Once those force pulses hit the frame, something else takes over: resonance.
Every machine has a natural frequency it wants to vibrate at
Tap the corner of a long aluminium extrusion and it rings at a specific pitch. Pluck a belt and it sings. That pitch is the part's natural resonance frequency, and every part of your printer — the gantry beams, the belts, the Z frame, even the print bed — has one.
When the force pulses from acceleration arrive at a rate that lines up with one of those natural frequencies, the vibration amplifies instead of dying out. The frame keeps bouncing after the motion command has already finished. Those residual oscillations carve themselves into the part as the classic ringing (also called ghosting or rippling) — the repeated ripples you see near sharp corners on fast walls.
Two things control how bad this gets:
-
How stiff the structure is. A rigid frame and tightly coupled gantry store and release far less vibration energy than a flexible one.
-
How much moving mass there is. A heavier toolhead or bed is harder to stop and reverse cleanly, so it excites more vibration with every direction change.
On a CoreXY machine, this matters even more. Because the X and Y motors both act through the same belt system, the two axes can share and mix resonances — which is why the shaking on a fast CoreXY often feels more pronounced than on a more basic cartesian printer.
Get the Mechanics Right First
Before you touch a single firmware setting, the machine has to be mechanically sound. Input Shaping cancels a resonance it measures; it does not tighten a loose belt, square a racked gantry, or reseat a wobbly rail. Skipping this step is the most common reason a "tuned" printer still shakes.
Work through these three checks in order. Each has a clear pass/fail, so you can verify the fix before moving on.
1. Square the frame and tighten every joint
A racked or loose frame is like a warped guitar neck — no amount of electronic tuning fixes the notes. Check that the gantry is square to the frame with a machinist square, ensure the bed carrier sits level and flush on its rails, and go around the machine torquing every structural screw you can reach.
On a printer with V-slot wheels, also check the eccentric nuts that set wheel tension. A wheel that's too loose lets the carriage wobble; one that's too tight adds drag and can bind. The carriage should slide smoothly with no perceptible side-to-side play.
Verify: grab the gantry and try to twist it. If you can feel any give at the joints, tighten them. The goal is a structure that feels like one solid piece.
2. Tune the belts by frequency, not by feel
Belt tension is where most high-speed shaking actually hides. A belt that's too loose adds compliance and slop, so motion can't transmit cleanly. When you're setting up and calibrating your machine, a common practical step is to standardise your 3D printer setup around repeatable checks like this before chasing "magic" slicer settings.
The reliable way to set belt tension is the frequency method: pluck the belt like a guitar string and read its pitch with a free tuner app. Tighten or loosen the tensioner until the belt hits the target fundamental frequency for its span.
-
Most desktop CoreXY machines sit comfortably around 90–110 Hz for a typical 6 mm GT2 belt measured at the midpoint of its longest unsupported span.
-
Larger CoreXY spans often land lower — roughly 60–90 Hz — because the longer belts resonate at a lower pitch.
-
The critical rule is balance: tune the left and right belts to within about 3 Hz of each other. An unmatched pair sends the toolhead slightly off its intended path on every diagonal move, which reads as extra vibration and echo on corners.
Verify: grab the toolhead and rock it side to side at the gantry midpoint. It should move with solid resistance, not flopping slack, and the two belts should read nearly the same frequency when plucked.

3. Give the mass a solid home
A printer bolted to a flimsy desk will shake even if everything above it is perfect, because the desk itself becomes part of the resonance system. For a machine in a shared space, this doubles as a safety and stability matter. Sovol's guidance on placing a large printer on a stable, flat surface makes the same point: a firm, level base reduces vibration and improves print quality.
A dense concrete paver or solid workbench is usually enough. Rubber dampening feet can help isolate motor whine from a desk, but be careful — too much compliance under a fast machine can add a slow wobble of its own. The aim is a rigid, heavy base that absorbs force rather than a springy one that amplifies it.
Verify: lean on the printer while it's idle. The frame should barely move. If the whole table sways, the base is part of the problem.
Let Firmware Cancel the Vibration
Once the mechanics are sound, you can stop fighting the vibration and start cancelling it. This is exactly what Input Shaping — also called resonance compensation — does in Klipper.
Input Shaping is an open-loop control technique. Instead of simply commanding the toolhead to move, the firmware shapes the motion command so the motor applies force in a pattern that cancels the machine's own vibrations before they reach the nozzle. As Klipper's resonance compensation documentation explains, it creates a command signal that cancels its own vibrations — so you suppress the shake without making the printer slower.
Use an accelerometer to find the real frequencies
You could try to guess your machine's resonance frequencies, but you'd be guessing through the fog. The accurate route is to measure them with an accelerometer.
In Klipper, that usually means a small ADXL345 board wired to the toolhead (and, if your setup supports it, to the bed). The workflow is well documented in this ADXL345 input shaping calibration guide and looks like this:

-
Mount the accelerometer firmly on the toolhead. A secure, rigid mount matters — a loose sensor reads garbage.
-
Add the
[adxl345],[resonance_tester], and[input_shaper]blocks to yourprinter.cfg. -
Run the resonance test. Issue
TEST_RESONANCES AXIS=XthenTEST_RESONANCES AXIS=Y. Klipper sweeps the axis through a range of frequencies while the accelerometer records how strongly each one resonates. You should watch this the first time — it shakes the toolhead hard enough to look alarming, but that's expected. -
Save the result with
SHAPER_CALIBRATE. Klipper analyses the sweep and recommends a shaper type and frequency for X and Y, then youSAVE_CONFIGso it writes those values straight into your config.
Once shaping is active, ringing at corners typically drops out almost completely, and you can raise acceleration back to a level that was unthinkable before.
A couple of practical notes that save real time in a shared machine:
-
Turn off the part-cooling fan during the resonance sweep. Airflow from a fan corrupts the accelerometer reading.
-
Re-tune after any mechanical change. If you replace a belt, tighten a gantry screw, or swap the toolhead, the resonance frequencies shift — so re-run the calibration.
-
CoreXY tips: since X and Y share the belt path, run and tune both axes even if only one seems to shake. And keep an eye on the recommended maximum acceleration Klipper reports, because it marks the ceiling before quality degrades.
Raise Acceleration — Carefully — From There
Input Shaping removes the echo of the vibration, but it has a physical limit. Beyond a certain point, the mechanical force genuinely overpowers the shaping and quality drops again.
So after calibration, treat acceleration as something you tune upward in controlled steps rather than a number you max out:
-
Print a small test part with sharp 90° corners (a ringing test tower is ideal).
-
Raise acceleration by a few hundred mm/s² and print again.
-
Inspect the corners for the first sign of ringing returning.
-
Back off to the last clean setting.
You'll often land far higher than the conservative default that was holding the machine back — that's the whole payoff. This is the "put it all together" stage where a machine that used to shake at modest settings runs fast and clean, exactly the outcome a busy workspace needs.
Pro Tip: Keep a written changelog for a shared machine — baseline filament, profile, belt frequency readings, and the saved Input Shaping values. When three people tune the same printer over a weekend, the changelog is what keeps the setup reproducible.
What This Means for Your Workspace
The practical upshot of all this is a machine that runs fast without shaking, which translates directly into what a makerspace leader cares about: higher throughput, fewer failed prints, and less late-night "why is it wobbling?"
-
Throughput: keeping acceleration high means faster prints and quicker prototype turnaround for your members.
-
Uptime: a stable motion system has fewer ringing artefacts and layer issues, so fewer prints need re-runs.
-
Teachability: the frequency method and
SHAPER_CALIBRATEare reproducible, documented checks any experienced member can perform — not dark magic reserved for one person. -
Open-source fit: Input Shaping and the ADXL345 workflow run natively on Klipper, so there's no proprietary lock-in, which matters when you're choosing hardware for a community.
Next Steps
If you're evaluating a high-speed printer for a shared space, the combination of a rigid CoreXY frame and true Klipper support is what makes the difference between a machine that shakes and one that runs clean at speed. The Sovol SV08 is a good example — an open-source CoreXY kit built for high-speed positioning that runs Klipper, so the Input Shaping workflow above applies directly. If you'd rather compare several fast machines side by side, Sovol's high-speed 3D printer comparison breaks down speed and build-volume trade-offs, and their firmware and profiles hub is where model-specific configuration and calibration files live.
Start with the mechanics, tune the belts, run the resonator, and then push the acceleration back up. Your printer will stop shaking — and your print times will thank you for it.


Share:
Why Large Build Volume Matters for Serious Makers
How to Print Large Models Successfully on an FDM Printer