Why Does My CNC Cut Have Chatter Marks?

Short answer: chatter is a feedback loop, not a setting. The tool deflects under cutting load, springs back, takes a bigger bite, deflects further — and that oscillation is what prints regular waves into the wall. You break the loop one of two ways: make the system stiffer (shorter stickout, larger diameter, less radial engagement) or move off the resonant frequency (change RPM by 10–15%). Feeding too lightly makes it worse, because a rubbing edge excites the loop instead of cutting through it.

What chatter actually is

A router bit is a cantilever spring. Under cutting load it bends away from the material; the load drops as it bends, so it springs back; springing back pushes it into the material again, the load rises, and it bends away once more. If the timing of that cycle happens to line up with a natural frequency of the tool, spindle or workpiece, the oscillation reinforces itself and grows.

What you see on the wall is the record of that oscillation — a regular, repeating pattern with a consistent pitch. That regularity is the diagnostic. Random roughness is a dull tool or torn grain. Regular, evenly-spaced waves are chatter.

A chattered wall with regular waves compared with a clean rigid cut Two cross-sections through a machined wall. On the left the tool oscillates as it deflects and springs back, printing evenly spaced waves into the surface. On the right a rigid setup with the same feed leaves a straight, flat wall. Chatter — the loop is running Tool deflects, springs back, deflects again tool Regular, evenly spaced waves Rigid — loop broken Same feed, shorter stickout, lighter stepover tool Flat wall, no pattern Chatter — the loop is running Tool deflects, springs back, deflects again Regular, evenly spaced waves Rigid — loop broken Same feed, shorter stickout, lighter stepover Flat wall, no pattern
The pitch of the waves is set by the oscillation frequency and the feed rate — which is why changing RPM often eliminates chatter without changing anything else.

Find your cause in one pass

What you observe Most likely cause Fix
Clean at shallow depth, wavy at full depth Tool deflection Shorter stickout, larger diameter, or lighter radial engagement
Chatter only when slotting full width Radial engagement too high Reduce stepover; add a finish pass
Waves plus burning or a squealing cut Chipload too low — rubbing excites the loop Raise feed into the published range for the series
Regular pitch that ignores feed changes Machine or workpiece resonance Change RPM by ±10–15%
Gets worse through the job Dull edge Replace the tool
Only on thin or unsupported areas Workpiece is vibrating, not the tool Improve hold-down; add tabs or support
One wall good, the opposite wall wavy Climb versus conventional direction Climb-cut the finish pass in wood

The single most effective fix: a finish pass

Most chatter lives in the roughing cut, where the tool is fully engaged. A light finish pass of 0.010″ to 0.020″ removes almost no material, so cutting force is small, deflection is small, and the loop never starts. The wall you keep is the one cut by that pass.

This is why a part can rough badly and still finish beautifully. If you are fighting chatter and have not added a finish pass, do that before changing anything else — it is the highest-value change available and it costs seconds of cycle time.

The RPM trick

If the wave pitch is regular and does not respond to feed changes, you are almost certainly sitting on a resonant node. Resonance is a property of the whole system — spindle, tool, holder, part, fixture — and you cannot design it away in the shop. But you can step off it.

Change RPM by 10–15% in either direction. If the chatter disappears, that was the cause. Then adjust feed proportionally so chipload stays constant: if you drop from 18,000 to 15,500 RPM, multiply your feed by 15,500/18,000 as well, or you will simply trade a chatter problem for a rubbing problem.

Chipload is set by the series, not the material

Under-feeding is a genuine chatter cause, so it is worth checking your numbers against the tool you are actually running. Published chipload for a 1/4″ bit in hardwood varies widely by series:

Series Flutes Chipload at 1/4″ in hardwood
Amana compression spiral 2 0.0031″
Onsrud 64-000 / 65-000 O-flute 1 0.004″ – 0.006″
Onsrud 57-200 downcut wood rout 2 0.005″ – 0.007″
Onsrud 60-100MW compression 1 0.014″ – 0.016″

That is a 5.2× spread across four legitimate published sources for the same material at the same diameter. There is no single "hardwood chipload" — only a chipload for the bit in your collet. Every product page in our catalogue lists its own published figures, and the same data drives our Cut Doctor and feed rate calculator.

Where rigidity actually comes from

  • Stickout. Deflection scales with the cube of unsupported length — halving stickout cuts deflection roughly eightfold. This is the cheapest fix available and most shops leave value on the table here.
  • Diameter. Stiffness scales with the fourth power of diameter. A 3/8″ bit is about five times stiffer than a 1/4″ one. If the geometry of the part allows a bigger tool, use it.
  • Collet condition. A worn or dirty collet grips unevenly, adds runout and makes every other fix less effective. Collets are consumables, not fixtures.
  • Hold-down. If the part can move, no tooling change will fix the finish.

Still chattering after a finish pass?

Tell us your material, bit, RPM, feed and stickout and get a specific diagnosis — including whether your chipload is below the published range for your exact series. Photo upload supported.

Diagnose my cut →

Bits that resist deflection

Onsrud 57-200 Series — 2 Flute Downcut Spiral Wood Rout
Larger core than a single flute; published data across five wood and panel materials
View bit
Onsrud 60-100 PLR — 2 Flute Polaris Compression
Production geometry rated 0.021″–0.023″ at 3/8″ — high feed without a heavy chip per tooth
View bit
Onsrud 60-100PLR — 3 Flute Polaris Compression
Three flutes spread the cutting force, which damps the oscillation
View bit

See more in hardwood router bits and compression router bits, or use the Bit Finder to match a bit to your machine's rigidity.

Common questions

What causes chatter marks on a CNC router?

Chatter is a self-reinforcing loop: the tool deflects under load, springs back, takes a larger bite and deflects again. When that cycle aligns with a natural frequency of the tool, spindle or workpiece it grows and prints regular waves into the wall. The usual underlying causes are excessive stickout, too much radial engagement, a chipload below the published range, a dull edge, or poor hold-down.

How do I get rid of chatter marks in wood?

Add a light finish pass of 0.010 to 0.020 inches — this is the single most effective fix because the finishing cut removes so little material that deflection never builds. Then shorten stickout, reduce stepover, and confirm your chipload is within the published range for your specific tool series. If the wave pitch is regular and ignores feed changes, change RPM by 10 to 15 percent and scale feed proportionally to hold chipload constant.

Does changing RPM fix chatter?

Often yes, when the cause is resonance rather than deflection. Resonance is a property of the whole machine, tool and workpiece system, and shifting RPM by 10 to 15 percent steps off the resonant node. Always adjust feed by the same proportion so chipload stays constant, otherwise you replace chatter with rubbing and burning.

Is chatter the same as a rough or torn surface?

No. Chatter produces a regular, evenly spaced pattern with a consistent pitch. Random roughness, fuzz or tear-out points to a dull tool, wrong cut direction, or grain behaviour instead. The regularity of the pattern is the key diagnostic.

Chipload figures are manufacturer application data stored per product in our catalogue and cited by series. Finish-pass depths and rigidity guidance are general starting points, not manufacturer specifications. Verify on scrap and adjust for your machine.