Most shops blame the plasma first.
It's the obvious suspect. It's the part you can hold in your hand, it's the part with a price on it, and it's the part somebody will happily sell you more of.
Meanwhile the table underneath it is twenty years old and the rack and pinion has never been touched. Not serviced, not adjusted, not replaced. Just cutting, five days a week, for two decades.
That machine is going to make bad parts no matter what you put in the torch.
We sell consumables. And the first thing we'd tell you is to stop buying them until you know that's actually your problem — because a motion fault diagnosed as a consumable fault costs you the parts and leaves the real fault in place. You get to buy a second stack and still have the problem.
Here's how to tell which one you've got.
The one question that splits it
Cut five of the same part. Lay them side by side.
Does the defect land in the same place on every one?
That single question does most of the work.
A geometry-locked defect — the same corner rounded on all five, the same hole out of round, the same short side undersized — is a motion or a program problem. The arc doesn't know where it is on the part. It has no way to fail in the same spot every time.
A progressive defect — the first part is clean, the fifth is worse, and it keeps getting worse — is a consumable or an arc problem. Wear accumulates over time, not over geometry.
Almost everything else is detail hanging off that one distinction.
What a motion problem looks like
Corners that round over or overshoot. Your table accelerates and decelerates on direction changes. If the acceleration settings are too aggressive for the machine, the gantry can't make the turn cleanly. Too conservative and you get a dwell mark where it slows more than it needs to. Either way it shows up at corners specifically, and it shows up on every part.
Holes out of round, or small holes that come out undersized and tapered. Small-diameter holes are the hardest thing a plasma table does — the machine has to move slowly enough for the arc to keep up but consistently enough not to distort. When hole quality falls off while straight cuts stay fine, look at motion and at your hole-cutting routine before you look at the nozzle.
A repeating pattern at a fixed interval. Run your hand down a long straight cut. If you feel a regular ripple that repeats at consistent spacing, that's mechanical — something in the drive is contributing a periodic error. Rack and pinion wear, a worn bearing, debris packed into the rack, a belt going off.
Error on one axis only. If X is clean and Y isn't, or the part is dimensionally correct in one direction and off in the other, that's not the arc. The plasma doesn't care which direction it's travelling.
Offset that appears on direction reversal. Slop in the drive train shows up as a small step where the machine changes direction, because the motor turns before the gantry does. Grab the torch carriage with the machine off and try to move it by hand. Anything you can feel, the machine can too.
Dimensional error that scales with part size. A part that's out by a little on a small piece and out by a lot on a big one usually points at calibration rather than mechanics.
The maintenance nobody did
Here's the uncomfortable part. A table doesn't announce that it's drifting. It degrades slowly enough that the shop recalibrates its expectations year by year, and nobody notices the baseline moved.
The tell isn't a measurement. It's memory: did this table ever cut better than it does now? If somebody who's run it ten years says yes, the answer isn't in the torch.
And the schedule isn't about the age of the machine. It's about hours under the arc.
Running two or three shifts, cutting around the clock — you want a full service every three to six months. That table is putting on more hours in a quarter than a one-shift shop does in a year.
Running one shift, three or four hours of actual cutting a day — an annual full service is usually enough. New racks, a full lube job, the table re-squared, and a proper clean end to end. Not a wipe-down. End to end.
Weekly, regardless of hours:
- Blow the racks out. They pack with dross and dust, and that debris is what chews the teeth and opens up your backlash. Blow them clean, then lubricate — lubricating over grit just grinds it in.
- Clean the slide on the torch height controller and lubricate it. A sticky Z is a height problem, and a height problem looks exactly like a consumable problem.
- Check the bolts on the gantry. Everything tight. A loose gantry makes defects that no amount of new parts will fix.
Daily or weekly, depending on how hard you're running:
- Clean the slats, and replace them when they're chewed up. Plate needs to sit as flat as possible on the slats while it's being cut. Warped, notched, built-up slats mean the plate isn't where the machine thinks it is — and every height reference downstream is working off a bad assumption.
None of that is expensive. Most of it is a compressor, a grease gun, and half an hour. It's just work that nobody schedules, because the machine keeps running and the consumables are right there to blame instead.
What an arc problem looks like
For completeness, because the contrast is the point:
- Angularity that changes over the run — the edge was square on part one and isn't by part five. That's the nozzle.
- Bevel on both sides equally — height. Too high, and the arc widens before it reaches the plate.
- Dross that changes character — heavy rolled dross at the bottom, or a thin hard line welded to the underside. That's speed against amperage, not motion.
- Top-edge rounding and a wide kerf that gets worse — the nozzle orifice is opening up.
- Everything got worse at once, across several parts — think air or power before you think mechanics.
If you want the full wear picture, the wear-ID pillar walks each part and what a used one should look like.
The one that sits in the middle
Height control. It's motion — a Z-axis doing a mechanical job — but every symptom it produces reads as an arc problem. Bevel, inconsistent kerf, dross, failed pierces, consumables dying early.
If your torch height isn't right, no consumable in the world fixes it, and you can burn through a lot of money proving that. Watch the Z actually move through a cut. If it hunts, lags, or parks at the wrong standoff, stop looking at parts.
The CYA post covers standoff in more detail.
And the one that isn't the machine at all
Before you start adjusting anything mechanical, check the program.
Lead-ins and lead-outs. A lead-in that starts too close to the profile puts the pierce damage in your finished edge. A missing lead-out leaves a divot where the arc shut off on the part instead of off it.
Pierce delay. Too short and the arc is still blowing molten metal when the machine starts moving — you get a gouge at the start of every profile. Same place, every part. Geometry-locked, and nothing to do with the hardware.
Cut direction and kerf compensation on the wrong side. Produces parts that are consistently the wrong size by about one kerf width, which is the kind of thing that gets blamed on everything else first.
Program problems are geometry-locked, same as motion problems — so when the repeatability test points that way, check the program before you touch the machine. It's free and it's faster.
A sane order to work through
- Cut five parts and compare them. Geometry-locked or progressive? That's your fork.
- If progressive — consumables, air, power. Start with a fresh stack and dry air.
- If geometry-locked — check the program first. Lead-ins, pierce delay, kerf compensation.
- Then watch the machine cut. Stand there for a full part. Corners, direction changes, the Z.
- Then go mechanical. Rails, bearings, rack, backlash, squareness.
- Change one thing at a time. Two changes and a better result tells you nothing.
And when you land on settings that work, write them down — machine, material, thickness, amperage, speed, standoff. Your own record of a known-good cut is worth more than any chart when something drifts.
Why this matters for what you spend
The reason to get good at this split isn't tidiness. It's that a motion problem diagnosed as a consumable problem costs you a stack of parts and leaves the actual fault in place — so you get to buy a second stack and still have the problem.
Knowing which side of the line you're on is what stops that.
Need parts matched to your machine and amperage? Our guided finder walks you from your machine to the complete stack — in assembly order, with current pricing. Find your consumables →
TorchandTip.com sells Thermacut® aftermarket plasma and welding consumables across Canada. Thermacut® is an independent manufacturer. TorchandTip.com is not affiliated with, authorized by, or endorsed by Hypertherm®, ESAB®, Lincoln Electric®, Kjellberg®, or any other original equipment manufacturer. All OEM names, model numbers, and part numbers are used for identification and reference purposes only.
0 comments