Compact CNC plasma cutting: how it works and when to use it

• 1
  What compact CNC plasma cutting actually does
• 2
  Gas, current and standoff: the three settings that decide cut quality
• 3
  Why a plasma edge is not a machined edge
• 4
  Where compact plasma cutting stops working
• 5
  From plasma blank to finished part: what the second operation does
• 6
  When a compact table pays for itself
• 7
  What a compact table realistically holds
• 8
  Seven steps to set up a compact table
• 9
  Compact CNC plasma cutting: thickness vs cut speed and kerf
• 10
  Compact plasma table vs fiber laser vs CNC milling

A compact CNC plasma cutting machine pairs a small gantry or cantilever frame with a plasma torch head driven by steppers or servos. The workpiece sits on a downdraft or water table, the torch moves on X and Y, and height control keeps the nozzle at a fixed standoff. That is the whole machine. The interesting part is what happens at the nozzle.

The arc starts as a pilot arc between electrode and nozzle, using a low-current DC circuit. Once the arc touches the workpiece, the main transferred arc forms: current flows from electrode, through the orifice, into the plate, and out through the work lead. Compressed air or nitrogen is forced through the same orifice, and the gas heats into a plasma that reaches roughly 20,000 °C at the core.

That temperature melts the metal locally rather than shearing it. Gas flow pushes the molten pool out of the kerf, and the fast travel speed keeps the heat-affected zone narrow. On 3 mm mild steel a typical cut runs at 3,000–5,000 mm/min with a 1.0–1.3 mm kerf. On 12 mm plate the same machine drops to 800–1,500 mm/min and the kerf widens to 1.5–2.0 mm.

None of this is new. Compact tables became practical when inverter power supplies replaced transformer units, cutting torch weight and letting a 40–60 A machine run from a single-phase supply. That is why a small shop can now cut plate without a 3-phase service or a 6 m footprint.

Current sets how much material you can cut and how square the edge stays. Running a 45 A torch at 45 A on 6 mm steel gives a near-vertical edge. Turning the same torch down to 30 A to save consumables produces a beveled, dross-heavy cut. Match the amperage to the thickness chart from the torch maker and leave it there.

Gas choice changes the cut face more than most operators expect. Compressed air is cheap and works well on mild steel up to around 10 mm. Nitrogen gives a cleaner, more oxidation-free edge on stainless and aluminum. On 304 stainless, air cutting leaves a dark oxide layer that has to be ground or pickled before any welding.

Standoff and pierce height are the two settings people get wrong most often. Pierce height is usually 1.5–2× the cut height so the molten splash does not clog the nozzle. Cut height is typically 1.0–1.5 mm. A torch that drags on the plate will lose the arc, and a torch that floats too high will lose the cut.

Air quality matters as much as air pressure. Moisture and oil in the line kill electrodes. A refrigerated dryer plus a coalescing filter at the machine inlet is not optional on a table that runs daily.

Plasma cuts by melting. The arc does not touch every point in the kerf equally, so the top edge is wider than the bottom edge on most cuts. That taper is the single biggest reason plasma parts usually need a second operation. On 6 mm steel, expect 3–6° of taper per side with a standard torch.

The heat-affected zone is the second limit. A 6 mm cut leaves roughly 0.5–1.0 mm of material whose hardness and microstructure have changed. On A36 and 1018 that rarely matters. On 4130 or 4140 the HAZ can harden enough to chip a cutter if you machine straight into it without a roughing pass.

Hole quality is the third limit. A plasma-cut 8 mm hole in 6 mm plate typically comes out 0.3–0.8 mm oversize with a lead-in mark on one side. If the drawing calls for H7 or a press fit, plasma is a blanking step, not a finishing step.

Kerf compensation in the controller helps but does not fix taper. CAM software offsets the toolpath by half the kerf width, so the nominal size is right at the top face. The bottom face is still smaller. On parts where the bottom edge is the datum, that error stacks with the taper.

Thin sheet is a harder problem than thick plate on a compact table. Below about 0.8 mm, the heat input warps the sheet before the cut finishes, and the pierce blowout is larger than the kerf. A small fiber laser handles 0.5–3 mm sheet far better. If your parts are mostly 1 mm enclosures, plasma is the wrong tool.

Very thick plate is the other boundary. A 45 A air torch will cut 12 mm, but the pierce takes 2–3 seconds and the edge needs grinding before welding. Above 20 mm, most compact tables lose the cut on long straight runs because the gantry cannot hold speed.

Tight tolerances are the third boundary. Plasma holds roughly ±0.2 mm on a well-tuned table and ±0.5 mm on a light hobby frame. That is fine for brackets, gussets and weldments. It is not fine for bearing bores, dowel holes or sealing faces.

Material reactivity deserves a mention. Aluminum cuts fast but the oxide layer re-forms within minutes, so any welding should follow within a few hours. Stainless cut with air needs the oxide ground off before it goes into a food or medical environment.

Most plasma-cut parts arrive at a machining center because the drawing has features plasma cannot hold. A 6 mm plate bracket might need two Ø10 H7 bores, a counterbore and a flat sealing face. Plasma gives the outline; the mill gives the features. That division of labor is cheaper than cutting the whole part from solid.

The machining allowance depends on the measured taper. If the test coupon shows 0.4 mm of taper per side, leave 1.0 mm of stock on any face that must be square. On a 4,000 mm maximum processing size machine, a plasma blank can be loaded, clamped and faced in one setup.

Clamping is the part people underestimate. Plasma edges are not flat, so a blank that sits on its cut face will rock. Machine a reference edge first, or clamp on the top face and face the bottom after. On thin blanks, light passes at 0.3–0.5 mm depth avoid pulling the part out of the vise.

Finishing follows the same logic. Anodizing and powder coating hide nothing, so any dross or HAZ discoloration has to be removed first. Bead blasting at a fine grade removes the oxide without changing dimensions. For stainless that will be welded, pickling or a light grind on the cut face is usually enough.

The case for a compact CNC plasma cutting table is not precision. It is the cost of a blank versus the cost of a cut. A 1,000 × 2,000 mm sheet of 3 mm mild steel cut into twenty brackets takes about 15 minutes of table time. Outsourcing that to a laser shop adds freight and a week of queue time.

Consumables are the real running cost. An electrode and nozzle set on a 45 A air torch lasts roughly 1–2 hours of arc-on time on 6 mm steel, less if the air is wet or the pierce height drifts. Budget for that, plus slats and a water table top-up, before comparing the table to a cutting service.

Electricity is minor. A 45 A inverter supply draws around 6–7 kW while cutting, so a shift of cutting costs less than the consumables. The bigger hidden cost is fume extraction, which is not optional indoors.

The payback case weakens on parts with many small holes or tight tolerances. Those parts spend more time in secondary machining than they save in blanking. If more than half your parts need secondary features, price the whole job as a machined part and use plasma only where the outline is complex.

Realistic positional accuracy on a 1.5 × 3 m compact table is ±0.2 mm after calibration, and that assumes a squared gantry, a level slat bed and consistent plate. Repeatability across a batch is usually better than absolute accuracy, which is what matters for welded assemblies.

Speed control through corners is the weak point on stepper-driven tables. The controller slows on a 90° corner, heat input rises, and the corner rounds or drosses. Servo drives and a THC with corner-height lock reduce this, but no compact table corners as cleanly as a laser.

Nesting software changes the economics more than the machine. A good nest on a 1,000 × 2,000 mm sheet uses 75–85% of the material; a poor nest wastes a third of the sheet and doubles the cut length. On a job with 200 identical blanks, spend the hour on the nest.

Maintenance is simple: change consumables on schedule, keep the slats level, drain the air system, and check the torch lead for kinks. Tables that fail mid-job usually fail because one of those four was skipped.