Plasma lives and dies by standoff: pierce too low and the torch eats spatter, cut too high and the kerf bevels. The height system that prevents both is really two systems, and the single most useful thing to learn about it is the boundary between them, because each owns different failures.
Job one: finding the plate before the pierce
Sheet stock is never where the program hopes: plates bow, slats sag, and yesterday’s cut warped the corner. So every pierce starts with initial height sensing, a probing move the post emits. On Mach-style controls that is G31; LinuxCNC dialects use the G38.2 probing family; the choreography is identical:
| Step | What happens |
|---|---|
| 1 | G31 Z-50 F200 style move: torch descends |
| 2 | Probe input trips: floating-head switch or ohmic contact |
| 3 | Control records the trip position |
| 4 | Back off the switch-travel offset: true surface found |
| 5 | Rapid to pierce height above that surface |
Two sensing styles trip the input: a floating head, where the torch mount slides up against a switch as the tip presses the plate, and ohmic sensing, where electrical contact with the plate closes the circuit, faster but fooled by rust, paint, and dross. Floating heads add the critical number in step 4: the switch travel, measured once and subtracted forever.
Job two: riding the plate during the cut
Once the arc is cutting, height control changes hands entirely. The THC reads arc voltage, which tracks standoff almost linearly, and nudges Z to hold the set voltage while the program supplies X and Y. None of that lives in your G-code: the program may toggle THC duty on and off where a controller expects commands for it, and that is the boundary. The division of labor in one line: the program finds the plate; the THC rides it. Cut-height complaints belong to voltage settings and consumable condition; pierce-height complaints belong to the probe sequence.
Where do the real-world failures hide?
Disproportionately in step 4. The switch-travel offset is measured once, trusted forever, and silently wrong after any head crash or probe-tab bend, at which point every pierce in every job inherits the same error while both the probe and the THC work perfectly. The diagnostic shortcut: if pierce height is wrong by a consistent amount everywhere, re-measure switch travel before touching anything else. The supporting cast of failures: probing too fast overruns the trip point and reads low; ohmic sensing on dirty plate trips late and reads high; and a post emitting probe feeds meant for a different table does either at scale, which is a post processor fix, one edit inherited by every future job.
A concrete afternoon saved: a table began cutting visibly high after a minor torch bump, and the operator spent an hour in THC voltage settings before measuring the floating head’s travel, which the bump had changed by a millimeter and a half. One offset update restored every cut. The bump had moved step 4; the THC had never been wrong.
How does this fit the rest of plasma code literacy?
The IHS block joins a short vocabulary: the torch-fire M-codes covered in plasma torch fire codes M03 vs M20, the motion core every table shares, and the dwell-and-pierce timing the post emits around each cut, all listed in the standard code references. Read your own table’s output once with the table above beside it, and the mystery block between cuts becomes five legible steps. The vocabulary underneath drills free on the G-code practice hub.
Bottom line
Torch height is two jobs: the program’s G31-style probe finds the plate, backs off switch travel, and sets pierce height; the THC rides arc voltage during the cut. Most failures live in the probe sequence’s offsets, especially switch travel after a bump, so diagnose there first and leave the THC settings alone until the pierce height is proven honest.
Sources
- LinuxCNC G-code reference (G38 probing family)
- Wikipedia: Plasma cutting
- CNCCookbook: G-code and M-code cheat sheet
Frequently asked questions
What does G31 do on a plasma table?
It is the probe move: the torch descends until the floating-head switch or ohmic sensor trips, the control records the position, backs off the switch-travel offset, and sets pierce height from the true surface. LinuxCNC dialects use G38.2 for the same job.
Does the G-code control torch height during the cut?
No. Live correction belongs to the THC reading arc voltage; the program finds the plate, sets pierce height, and at most toggles THC on and off.
Why does my plasma cut at the wrong height even though probing works?
Usually the switch-travel offset, silently changed by a head bump or bent tab; also probing too fast or ohmic sensing on dirty plate. Re-measure switch travel first.
What is the best way to learn the codes behind plasma probing and THC?
Drill the probing and motion vocabulary until the post’s IHS block reads at sight. A free app like G-Code Sprint quizzes the everyday codes and repeats whichever ones you miss.
G-Code Sprint is a study and practice tool only. Always follow your instructor, employer, machine manual, and shop safety procedures.