A 0.4 mm Step Where the Cutter Comp Engaged
A 2.5D finish pass around a 60 × 40 mm pocket wall in 6061 uses a 12 mm, 3-flute carbide end mill with G41 D21. The entry is G1 G41 X20.0 Y-4.0 D21 F600 - a 4.0 mm linear move perpendicular to the wall at Y0 - followed by G1 X20.0 Y0.0 and then the contour. The finished part shows a 0.4 mm deep witness step and a slight gouge exactly where the tool entered; every other point on the wall is within 0.02 mm. The D21 register holds 6.02 (6.00 mm radius plus 0.02 mm wear allowance), and the tool was measured on the presetter. The same program ran clean on the older control in the other bay.
1. Explain the mechanism that produced the step and gouge at the entry point.
Read the worked answer
The control had less lead-in distance than the offset it had to build.
With G41 the control must place the tool center 6.02 mm to the left of the programmed path. On a 4.0 mm lead-in, the compensation vector has to grow from zero to the full 6.02 mm and rotate to become perpendicular to the wall within a move shorter than the tool radius. The control resolves this by extending/looping the compensated path - the classic cutter-comp interference condition. Some controls alarm ("interference" / PS alarm on a Fanuc), some simply cut the loop, which shows up as a witness step and a gouge at exactly the entry point.
Supporting evidence to confirm the diagnosis:
- The defect is localized to the entry; the rest of the wall is within 0.02 mm, so the D value and the tool size are right.
- 0.4 mm is not a size error - it is a path-shape error.
- The older control may use a different comp type (A vs B), a disabled interference check, or simply produce the loop silently.
Rule: the first move after G41 must be at least the tool radius in length - practically one full tool diameter - and the entry should be perpendicular or tangential, never a short stub.
2. Rewrite the entry and exit logic to be robust on any control.
Read the worked answer
Long lead-in, long lead-out, overlap, and cancel away from the wall.
(approach clear of the part, comp cancelled)
G0 X-15.0 Y-15.0
G1 Z-8.0 F300
G1 G41 X20.0 Y-14.0 D21 F600 (14 mm lead-in ~ 1.2 x D, perpendicular to the Y0 wall)
G1 Y0.0 (contact the wall)
... contour ...
G1 Y-14.0 (lead-out, same length)
G1 G40 X-15.0 Y-15.0 (cancel on a move away from the wall)
Design rules:
- Lead-in/lead-out length ≥ 1 × tool diameter (minimum acceptable is the tool radius; one diameter gives margin for the wear allowance on D).
- Enter perpendicular to or tangent to the first contour element. Never engage comp on an arc smaller than the tool radius, and never on a move shorter than the offset.
- Add overlap (1–2 mm) at the contour closure so the entry mark is blended out.
- Cancel G40 after the tool has left the finished surface.
- Keep D equal to the true radius plus a small wear allowance (0.02 is fine); do not use D to correct a feature that is off-size for other reasons.
Verify before cutting metal: run the program with a positive Z clearance in single block on the control's graphics or a dry run, and watch the comp build-up at the lead-in.
3. The operator proposes dropping D21 from 6.02 to 6.00 and slowing the feed to 300 mm/min. Does that fix it?
Read the worked answer
No - neither change touches the cause.
- D: 6.02 → 6.00 moves the whole wall 0.02 mm. It is a size correction. The 0.4 mm step is a path-geometry error produced by a lead-in shorter than the offset; it will still be there, roughly 0.4 mm, regardless of D.
- Feed: 300 mm/min changes surface finish and chip load, not the geometry of the compensated path. It may make the step look slightly smoother, which is worse - it hides the defect.
What actually changes the outcome:
- Lengthen the lead-in (one diameter) and add overlap - the real fix.
- Check the control's cutter-comp interference parameter and comp type so both machines behave the same way.
- If a short entry is unavoidable, apply comp on the positioning move well clear of the part (or use a control that supports a type-B style comp with a defined approach), never at the wall.
Judgement point: a proposal that changes the symptom's appearance but not its magnitude is a rejection, not a compromise. Ask the operator to demonstrate acceptance with a first-article check on a scrap slug before the change goes back into the program.