G92 Coordinate Offset: What It Does and Why G54 Is Safer

G92 is the old way to set a part zero: park the tool somewhere known and declare it. G92 X0 Y0 Z25 means "from now on, call this position X0 Y0 Z25" — the control back-computes an offset and applies it globally. No axes move.

It predates stored work offsets, and on hand-cranked or unhomed machines it was the only option. On a modern homed machine it survives mostly as a source of mystery shifts — because of two properties:

1. It stacks. The G92 shift adds on top of the active G54–G59 offset — and stays applied when you switch to another work system. Every fixture on the table moves together.
2. It lingers. The shift persists across program end and reset on many controls — FANUC stores it in parameters that survive power-down. The next program inherits it invisibly.

G54 G0 X0 Y0     ; tool over G54 zero: readout says X0 Y0
G92 X10          ; declare "this is X10" — a −10 shift now exists
G55 G0 X0 Y0     ; even in G55, everything is shifted by −10
G92.1            ; clear it; sanity restored

#Where G92 still earns its place

• Axes that can't home. A 3D printer's extruder (G92 E0 before every retract block) or a rotary axis treated as endless.
• Unhomed hobby setups where "zero is where I jogged it" is the entire workflow — GRBL's G92 exists for exactly this.
• Temporary shifts inside a program on controls without G52 local offsets — though if G52 exists, it's the better-behaved tool.

If the machine homes and the control has work offsets, the modern rule is blunt: set G54, don't set G92.

#Debugging a mystery shift

Parts cutting consistently off-location by the same amount, with correct work offsets, is the G92 signature. The checklist:

1. Look at the control's offset/parameter screen for nonzero G92 values.
2. Clear with G92.1 (or zero the parameters).
3. Find the program that set it — usually a hand-written probe or setup routine — and replace the G92 with G10 L20 writing to a real work offset.

A simulator that tracks the full offset stack shows a G92 shift the moment it happens, instead of after the third scrapped part.

#Common mistakes

• Using G92 to "touch off" on a homed machine. It works until anything — a different program, a reset behavior difference — reveals the shift was global and persistent. G10 L20 P1 X0 (write to G54) is the intended tool.
• Assuming reset clears it. Control-dependent; FANUC's survives power-down. Only G92.1 (or zeroing the parameters) is definitive.
• Mixing G92 into a multi-fixture job. The shift drags every G54–G59 system with it — the second vise is now wrong too, which makes the symptom look like a machine problem instead of a program one.