CNC Tour Coordinate System: How Machine Position Becomes a Cut

• 1
  The three frames inside a CNC tour coordinate system
• 2
  How X, Y and Z map to real machine motion
• 3
  Work offsets and tool offsets: where the numbers live
• 4
  How Fanuc and other controls set the part frame
• 5
  Frame errors and what they cost at the tool tip
• 6
  Which frame to trust for which job

A CNC machine does not know where the part is. It only knows where its own axes are. The CNC tour coordinate system is the chain of frames that connects the control's internal position to a point on the workpiece. Three frames do the work: machine, work and part. Each one is defined at a different time and for a different reason.

The machine frame is fixed at build time. It comes from the home position of each axis, set by the machine builder or by the encoder zero after a reference return. Nothing on the shop floor should move it. When a machine loses its reference, the machine frame is the first thing to rebuild before any job runs.

The work frame is what the programmer thinks in. It sets the origin where the drawing sets it, often a corner, a bore center or a datum face. On a vertical mill this is usually G54 through G59. The part frame is the same idea applied to one specific workpiece, and on lathes it is often called the program origin.

These three layers stack. A move in the part frame is transformed into the work frame, then into the machine frame, then into axis commands. Get one layer wrong and the error appears at the tool tip, not in the numbers on the screen.

Nearly every machining center uses a right-handed Cartesian set. X runs along the longest travel, Y across it, and Z along the spindle axis. On a vertical mill, Z is the spindle direction. On a horizontal mill or a lathe, the same letters describe different physical axes, which is why the machine type matters as much as the G-code.

Right-handed means a simple rule: point the thumb of your right hand along +X, the index finger along +Y, and the middle finger gives +Z. This matters for rotary work. A four-axis job that indexes around X behaves differently from one that indexes around Y, and a sign flip sends the cut into the fixture instead of the part.

Direction also depends on whether the tool moves or the table moves. On many mills, a positive X command moves the table left so the tool appears to move right relative to the part. Operators who read position from the table rather than from the tool tip misjudge offsets by the full travel distance.

For most prismatic parts, three linear axes and a clear datum cover the job. Add a fourth or fifth axis only when the feature cannot be reached or when one setup replaces three. More axes mean more frames to keep aligned.

A work offset stores the distance from the machine origin to the work origin. On a Fanuc-style control, G54 to G59 hold those values, and G10 or the offset screen edits them. The tool length offset is separate: it stores the distance from the spindle gauge line to the tool tip. Two different families of numbers, two different screens.

This split is where many first-run crashes start. The work offset is correct, the tool length is correct, but they are stored in the wrong register. The control then adds a 12 mm face mill length to a 3 mm drill and the rapid move lands in the vise. Check the active offset number on the position screen before the first rapid.

On turning centers, geometry and wear offsets are split the same way. Geometry sets the nominal tool position, wear trims it after measurement. A common mistake is to bury a large error in the wear column instead of correcting geometry, which then drifts across every part in the batch.

Keep one rule: geometry offsets carry setup, wear offsets carry drift. Mixing them leaves no clean record of what changed between the first part and the last.

Fanuc controls offer several routes to the part frame. You can write tool values into parameters, define the frame with G50, or touch off in MDI and store the result in G54. Each route writes to a different place, so the offset screen may look correct while the active frame is not.

Other control families use G92 to define the current tool position as a coordinate value. That is a shift, not an absolute frame. Run a G92 block twice and the origin moves twice. Use G54-style offsets for repeat work and keep G92 for short manual sequences.

The practical difference shows up after a power cycle. Absolute offsets survive. A frame built from a G92 shift may not, depending on the control and the parameter set. On a production run, an origin that moves after a restart costs a whole first part.

Whichever route you use, prove it. Bring the tool to a known datum, read the position screen, and compare against the drawing. One minute of checking beats a scrapped casting.

A frame error does not stay small. It adds directly to every feature cut in that frame. If the work origin is off by 0.05 mm, every hole in the batch is off by 0.05 mm. The machine may hold ±0.005 mm repeatability and still produce a bad part, because repeatability only describes how well it repeats the wrong position.

Thermal drift works the same way. A spindle that grows 0.02 mm over a four-hour run shifts the part frame as the tool length changes. On tight work, warm up the machine, then touch off. On long runs, re-check the datum at intervals and let the wear offset absorb the drift.

Rotary axes add a second source. Any runout or center-height error on a fourth-axis table appears as a position error that grows with the distance from the rotary center. Index the part close to the center when the tolerance is tight.

The engineering meaning is simple: the frame is part of the tolerance budget. It is not a setup detail that sits outside the drawing.