---
title: "Lathe Cutoff Leaving a Nub: The G-Code and Setup Fixes"
description: "A cutoff nub has three parents: tool above or below center, an X endpoint that stops short, and centerline speed physics. The fixes, code and setup both."
url: https://gcodepractice.com/journal/g-code-for-part-cutoff-lathe-leaving-a-nub/
canonical: https://gcodepractice.com/journal/g-code-for-part-cutoff-lathe-leaving-a-nub/
author: "Lawrence Arya"
authorUrl: https://www.linkedin.com/in/vibecoding/
published: 2026-06-07
updated: 2026-06-07
category: "Practice"
tags: ["cutoff", "parting", "lathe", "troubleshooting"]
lang: en
---

# Lathe Cutoff Leaving a Nub: The G-Code and Setup Fixes

> **TL;DR** A nub or pip left after lathe cutoff has three parents, and the nub's own shape points at them: a tool above or below center leaves material the edge physically could not reach (the dominant cause, fixed at setup, not in code); an X endpoint that stops at or short of centerline leaves geometry by arithmetic, the code fix being to program X slightly past center; and centerline physics, where constant surface speed demands infinite RPM at zero diameter, argues for capping RPM with G50, switching to G97 near center, and easing the feed for the last fraction of a millimeter.

The nub is evidence, not just annoyance: a little cone or pip on the parted face is a physical record of where the cutting edge actually was when the slug let go, and reading it correctly sorts the three parents of the problem faster than any forum thread. Two of the parents live in the setup and one in the code, which is worth knowing before editing anything, because the most common cause cannot be programmed away.

## What cutoff asks of the machine

Parting is the lathe's strangest operation: a thin blade buried in a deepening groove, cutting at a diameter that shrinks toward zero, on a part whose connection to the spindle is being deliberately destroyed. Every other turning operation enjoys stable geometry; cutoff's geometry degrades by design, which is why it concentrates so many failure modes into one pass and why the little cone on the finished face has three different possible authors. The good news is forensic: each author signs differently.

## The three parents, and what each nub looks like

| Parent | The mechanism | The nub's testimony |
| --- | --- | --- |
| Tool above or below center | The edge geometrically cannot reach the axis | A cone that survives even past-center endpoints |
| X endpoint short of center | The pass stops before the axis by arithmetic | A consistent stub matching the missing distance |
| Centerline physics | Surface speed collapses; material tears, slug sags | A torn, irregular pip, often with a heat mark |

## Parent one: center height, the setup truth

On a [lathe](https://en.wikipedia.org/wiki/Lathe), a tool above or below the spindle axis leaves an uncuttable core: the edge sweeps a path that simply never intersects the center, and no X endpoint fixes geometry that the tool's height already forfeited. This is the dominant nub source in practice, and its fix is physical, shim or adjust the holder until a light facing cut dies clean at center, the facing test being the honest referee since it runs the same geometry. Cutoff tools also deflect under load, so a tool perfectly on center at rest can ride up in the cut; rigidity, stickout, and sharpness all testify here.

The diameter-mode arithmetic that makes lathe X values [mean diameters](/journal/why-is-my-cnc-lathe-cutting-twice-as-deep/) matters in this conversation constantly, because the next parent is about exactly that number.

## Parent two: the endpoint, the code truth

A pass programmed to X0 ends the edge at the theoretical axis, assuming perfection everywhere else; a pass programmed slightly past, a small negative X in diameter mode, sweeps through center and buys insurance against the sum of small errors. The practice is common, the amount modest (on the order of half a millimeter to a millimeter past zero, scaled to the tool and machine), and the limit is the machine's own geometry and the tool's nose hitting the far side of nothing. The [lathe-user documentation](https://linuxcnc.org/docs/html/lathe/lathe-user.html) covers the coordinate conventions; the shop habit is reading the cutoff line of every inherited program for exactly this endpoint, because X0-exactly is a very popular way to manufacture pips.

## Parent three: the physics at zero diameter

Constant surface speed mode keeps cutting speed constant by raising RPM as diameter falls, and at center the request becomes infinite: the control hits its or [G50's cap](/journal/g50-max-spindle-speed-crash-lathe/), surface speed then collapses toward zero, and the final material tears rather than cuts while the slug's support vanishes under it. The working pattern, documented across references like [Helman CNC's G96/G97 pages](https://www.helmancnc.com/cnc-lathe-g96-and-g97-g-code/): cap RPM always, and for the final approach many programmers [switch to fixed-RPM G97](/journal/spindle-speed-calculation-for-g97-lathe/), trading ideal surface speed for predictable behavior at separation, with feed eased over the last fraction of travel. A parts-catcher or a gentle hand on the slug finishes the story politely.

## The combined fix, in order

Set center height by the facing test first, because nothing downstream survives its absence. Program the endpoint past center second. Tune the speed strategy third, cap, switch, ease. Then read the next nub, if any, as evidence again: cone means height moved (check deflection), stub means endpoint, tear means physics, and the [error-species discipline](/journal/how-to-read-g-code-to-find-errors/) applies to chips as well as code. The lathe vocabulary that keeps this audit fast, G96, G97, G50, diameter-mode X, lives in the standard core, free to drill in the 60-second rounds on the [G-code practice page](/g-code-practice/), and cutoff is where the lathe family's fluency pays in the most literal way: the last millimeter of every part, every cycle, forever.

## Sources

- [LinuxCNC: Lathe user information](https://linuxcnc.org/docs/html/lathe/lathe-user.html)
- [Helman CNC: G96 and G97 on the lathe](https://www.helmancnc.com/cnc-lathe-g96-and-g97-g-code/)
- [Wikipedia: Lathe](https://en.wikipedia.org/wiki/Lathe)

## Frequently asked questions

### Why does my lathe cutoff leave a nub on the part?

Three parents, often cooperating: the parting tool above or below centerline (the dominant cause, fixed at setup), an X endpoint at or short of center (fixed in code by programming slightly past), and centerline physics where surface speed collapses (managed with G50 caps, a G97 switch, and eased feed).

### What X value should a cutoff pass end at?

Slightly past the spindle centerline, a small negative X in diameter mode, so the edge sweeps fully through center. Ending exactly at X0 assumes perfect center height.

### Should cutoff run in G96 or G97?

Start in G96 if you like, cap RPM with G50 always, and switch to G97 for the final approach: predictable behavior at separation beats ideal surface speed at a vanishing diameter. Ease the feed for the last fraction.

### How do I tell if the parting tool is on center?

The facing test: a light facing cut leaves a center pip when height is off. Shim until the facing pass dies clean, and the nub's own shape, cone versus stub versus tear, tells you which parent to revisit.

---

Source: https://gcodepractice.com/journal/g-code-for-part-cutoff-lathe-leaving-a-nub/
Author: Lawrence Arya — https://www.linkedin.com/in/vibecoding/