Ringing and Ghosting: Fix Surface Ripples on 3D Prints

Ringing—those faint, wavy echoes that trail sharp corners—is one of the most persistent surface-finish problems in FDM printing. It appears right where the toolhead changes direction, when the printer’s motion system vibrates instead of stopping cleanly. The result looks like a ghost image of the corner repeated at decreasing intensity across the next flat wall. The good news is that ringing is almost always solvable, and the test print that reveals it also tells you exactly which adjustment to make first. I’d start by printing a dedicated ringing test model and reading the pattern before touching a single belt or firmware setting.

What Ringing and Ghosting Actually Look Like

Ringing shows up as a series of vertical ripples on a print’s surface, always on the face immediately after a sharp feature. If a cube has an embossed letter X on one side, the ripples will appear to the right of the X’s vertical edges, not before them. The amplitude decays with distance—the first wave is the strongest, and each subsequent wave is smaller. Ghosting is the same phenomenon, just named for the way it echoes the shape of a feature across an otherwise smooth wall.

These artifacts are distinct from layer shifts, inconsistent extrusion, or Z-banding. Layer shifts misalign entire layers. Inconsistent extrusion creates random horizontal bands. Z-banding gives a regular, repeating pattern tied to the leadscrew pitch. Ringing is tied to motion, not extrusion or Z-axis mechanics. If the pattern changes when you rotate the model on the build plate, you’re dealing with ringing from a specific axis. If it stays oriented the same way relative to the printer’s frame regardless of model rotation, the culprit is that axis’s motion system.

The Test Print That Tells You Everything

The most useful diagnostic is a ringing tower—a tall, narrow print with sharp 90-degree corners and a flat face on at least two sides. Many slicers include a built-in calibration shape, but a simple 20 mm square tower that’s 100 mm tall with a letter or notch on one face works perfectly. Print it at the speed and acceleration you normally use for outer walls. That’s critical: ringing is speed-dependent, so testing at 30 mm/s when you normally print at 80 mm/s will hide the problem.

Inspect the tower under raking light. Hold it so the light skims across the surface, and look for the telltale ripples emanating from the vertical corners. Measure the distance between the corner and the first ripple peak. Then measure the spacing between successive peaks. These distances, combined with the print speed, give you the ringing frequency. The formula is frequency (Hz) = print speed (mm/s) / ripple spacing (mm). If you printed at 60 mm/s and the ripple peaks are 1.5 mm apart, the ringing frequency is 40 Hz. That number is your target for input shaping later. Write it down.

Pay attention to which face shows the strongest ringing. If the X-facing wall has obvious waves but the Y-facing wall is clean, the X-axis motion system is the primary source. If both axes show similar artifacts, you’ll need to address both. Also check whether the ringing is worse near the top or bottom of the print. A tall, slender tower can amplify frame vibrations, so ringing that increases with height points to insufficient frame rigidity rather than belt or acceleration issues alone.

Belt Tension: Finding the Sweet Spot

Belt tension directly affects the springiness of the motion system. Too loose, and the belt stretches under acceleration, then snaps back, setting up oscillations. Too tight, and the belt acts like a rigid transmission line for motor vibration, while also increasing friction and bearing wear. CoreXY and Cartesian printers both need proper belt tension, but the symptoms differ slightly. On a CoreXY machine, loose belts often produce diagonal ringing artifacts because both motors contribute to any single-axis move.

The pragmatic way to set tension is by frequency, not by feel. Pluck the longest unsupported span of belt like a guitar string and measure the frequency with a smartphone app (a guitar tuner app works). For a typical 6 mm GT2 belt with a free span of about 150 mm, a frequency of 50–60 Hz is a good starting point for X and Y axes. That corresponds to roughly 2–3 kg of static tension, but the frequency measurement is far more repeatable across machines. If your free span is longer, the target frequency drops; if it’s shorter, it rises. The relationship is linear: doubling the span length halves the target frequency for the same tension.

After adjusting tension, reprint the ringing tower. If the ripple amplitude decreased but the spacing stayed the same, you’ve reduced the energy in the system without changing its resonant frequency. That’s a partial win. If both amplitude and spacing changed, the resonant frequency shifted—this often happens when going from very loose to properly tensioned. Note the new frequency. If ringing is still unacceptable, the next step is input shaping.

Input Shaping: Canceling the Wave Before It Starts

Input shaping works by modifying the motion commands sent to the motors so that the vibrations produced at the start of a move are exactly canceled by opposing vibrations a moment later. It’s not a smoothing filter; it’s a deliberate overlay of a second impulse timed to the ringing period. The result is that the toolhead arrives at the target position without residual oscillation. Firmware like Klipper and RepRapFirmware implement input shaping directly, while Marlin supports it through linear advance and junction deviation tuning, though the latter is less direct.

There are several shaper types, each with different trade-offs between vibration reduction, corner smoothing, and maximum acceleration. The three most common are ZV (zero vibration), MZV (modified zero vibration), and EI (extra insensitive). ZV is the simplest: it uses two impulses separated by half the ringing period. It eliminates vibration completely if the frequency estimate is perfect, but it’s sensitive to errors. MZV adds a third impulse to widen the effective frequency range at the cost of slightly more smoothing. EI is the most robust to frequency mismatch and is often the best choice for printers where belt tension drifts with temperature or where the resonant frequency changes with axis position.

To apply input shaping, you need the ringing frequency from the test print and the shaper type you’ve chosen. In Klipper, for example, you’d run a calibration test that moves the toolhead back and forth while an accelerometer measures the actual vibration. The firmware then recommends shaper type and frequency. Without an accelerometer, you can use the manual tower method: print a ringing tower, measure the frequency, and set the shaper parameters directly. For a 40 Hz resonance on the X-axis, you’d set shaper_type_x = ei and shaper_freq_x = 40 in Klipper’s config. Reprint the tower. The ripples should be dramatically reduced, often to the point of being invisible.

One critical detail: input shaping adds a small delay to motion commands because the shaper needs to look ahead in the move queue. This can affect dimensional accuracy on very small features, but for most prints, the effect is negligible. If you’re chasing absolute precision on parts with tight tolerances, print a calibration cube with and without shaping and measure the difference. You may need to adjust horizontal expansion slightly.

Acceleration and Jerk: Turning Down the Energy Source

Belt tension and input shaping address the system’s response to vibration, but reducing the energy that causes vibration in the first place is equally valid. Acceleration and jerk (or junction deviation) control how abruptly the toolhead changes velocity. High acceleration forces the motors to apply large torque instantly, exciting the belt’s natural frequency. Reducing acceleration lowers the initial impulse, making the system easier to damp.

A practical workflow: find the highest acceleration that produces acceptable ringing with your chosen input shaping. Start at your current outer-wall acceleration—often around 500–1000 mm/s². Print a ringing tower. If ringing is present, enable input shaping first. If it persists, reduce acceleration by 200 mm/s² increments and reprint. Many printers reach a point around 1500–2000 mm/s² where ringing becomes negligible, while still delivering reasonable print times. Going below 500 mm/s² usually yields diminishing returns and makes prints unnecessarily slow.

Jerk, or the newer junction deviation, controls the instantaneous speed change allowed at corners. Classic jerk is set in mm/s; a value of 8–10 mm/s is typical for direct-drive printers, while Bowden setups sometimes need lower values to avoid overshoot. Junction deviation in Marlin replaces jerk with a single parameter that automatically scales with acceleration. A value of 0.03–0.05 is a safe starting range. If ringing is concentrated right at the corner and decays within a millimeter or two, lowering jerk will help more than lowering acceleration. If the ripples extend far from the corner, acceleration and input shaping are the primary levers.

Frame and Motion System Rigidity: The Foundation

No amount of tuning will fix ringing if the printer’s frame flexes under the toolhead’s inertia. A lightweight printer with unsupported uprights can oscillate at a low frequency that belts and shapers can’t fully cancel. The test print gives a clue here: if ringing is worse near the top of a tall print, the frame is likely twisting. Bracing the gantry, adding diagonal supports, or simply bolting the printer to a heavy, stable surface can shift the resonant frequency higher, where belts and input shaping are more effective.

Check all motion-system fasteners. Loose linear rail bolts, wobbly bearing blocks, or a print head that can rock on its mount all introduce play that masquerades as ringing. The difference is that play usually produces a more irregular pattern, sometimes with a double image rather than a clean decaying wave. Tighten everything, then reprint the test tower. If the pattern changes from irregular to a uniform ripple, you’ve removed slop and exposed the true belt resonance, which input shaping can then handle.

Putting It All Together: A Systematic Sequence

I’d recommend this order, because each step builds on the last and avoids masking root causes:

1. Print the ringing tower at your normal outer-wall speed and acceleration. Record the ripple spacing and frequency for each axis.
2. Check frame rigidity and tighten all motion hardware. Reprint if anything was loose.
3. Set belt tension to the 50–60 Hz target (adjusted for span length). Measure the frequency, not just the feel. Reprint and note the new ringing frequency.
4. If ringing is still visible, enable input shaping. Start with EI shaper at the measured frequency. If your firmware supports accelerometer-based calibration, use it; otherwise, enter the manual frequency.
5. Fine-tune acceleration and jerk/junction deviation. Reduce acceleration until the ripple amplitude is acceptable, then nudge jerk lower if corner-specific ghosting remains.
6. Print a final test tower and a real-world part with similar geometry to confirm the surface finish holds up on longer prints.

The whole process rarely takes more than an afternoon, and the payoff is a printer that produces smooth, professional-looking vertical surfaces even at speeds that would have been unthinkable without it. The ringing tower is the thread that ties all the adjustments together—print it after every change and you’ll never guess at what worked.