Fiber Laser Cutting Torch Head Sensor: How Height Control Actually Works
This page explains what the fiber laser cutting torch head sensor on an FK-4020 measures, how its signal closes the height loop, and which mounting and material choices keep it stable. Written for engineers and maintenance leads who must judge whether a sensor fault is the sensor, the mount, or the machine.
What the fiber laser cutting torch head sensor measures
A cutting torch head sensor is the feedback element in the Z axis. It does not cut anything and it does not move anything. It reports one number, the gap between the nozzle and the sheet, and the CNC uses that number to drive the servo. On an FK-4020 style machine with a 4 m x 2 m bed, that gap decides whether the cut is clean or the sheet is scrap.
On most fiber heads the sensing element is capacitive. The nozzle forms one plate, the workpiece the other, and a ceramic ring insulates them. As the gap closes, capacitance rises. The controller converts that shift into a height number and commands the Z servo. A second, contact-based channel usually sits alongside it as a backup for non-conductive work.
Standoff is the number that matters. On 1–3 mm mild steel it typically sits around 0.8–1.2 mm. On thicker plate the head runs higher so spatter clears the nozzle. On thin stainless or aluminium, where dross forms easily, you want the low end of the 0.5–1.5 mm window and tight control over it.
The sensor also watches the pierce. During piercing, molten metal and vapor sit directly under the nozzle, and the capacitance reading goes noisy. Firmware blanks the signal for a set pierce time, then reacquires. If that reacquisition window is wrong, the head either dives into the part or drifts up and loses the cut.
How the height loop closes, and how fast
The loop is short. Sensor measures gap, controller compares it with the setpoint, servo moves the head. Total response time on a well-tuned FK-4020 class head sits under 10 ms. That is fast enough to follow mild sheet distortion at normal feed rates, but not fast enough to follow a loose sheet that is flapping in the slats.
Bandwidth matters more than resolution here. A sensor that resolves 1 μm but updates at 100 Hz will lose a wavy 4 m sheet at 20 m/min. Sampling at 1 kHz and resolving a few micrometres is the useful combination. Anything beyond that is rarely the limiting factor in the cut.
The ceramic ring and the nozzle are part of the sensing circuit, not just hardware. Change nozzle brand and you change the effective electrode area. That shifts the calibration constant. It is why a nozzle swap can throw the height off by 0.2 mm until the head is recalibrated.
Thermal drift is the other slow error. A 4 kW source heats the head over a shift. Aluminium mounts expand roughly 23 × 10⁻⁶ per °C, so a 100 mm bracket grows about 0.05 mm over 20 °C. On a 1.0 mm standoff that is a real fraction of the tolerance, and it shows up as a slow height creep through the shift.
Two more things sit inside the loop. First, the machine has to know the sheet surface is where it thinks it is. A bowed sheet or a burr left over from shearing reads as a height error. Second, look-ahead: the controller uses the tool path to predict where the sheet will be, so the servo acts before the gap actually changes. If the nesting software does not share that path, the head reacts late.
Mounts, housings, and the parts we machine for these heads
Most field failures we see on these heads are not sensor electronics. They are the bracket. A mount that flexes 0.02 mm under cutting vibration adds 0.02 mm of apparent height error, and no amount of controller tuning removes it. So the first question is always whether the bracket is actually rigid.
Housings that sit near the head see a real thermal gradient. We machine these from 6061 or 6063 aluminium when heat has to move away from the sensor, and from 316 stainless when the shop is humid and corrosion is the bigger risk. 316 carries a PREN above 25, which is why it holds up in a wet storeroom or near a waterjet.
Tolerances on the mating faces are where the accuracy lives. A sensor mount with a ±0.005 mm face flatness and a repeatable dowel location keeps the calibration constant across a nozzle change. If the mount face is out by 0.05 mm, every recalibration fights the hardware.
We hold ±0.005 mm on these parts, with surface finish down to Ra 0.2–0.8 μm on sealing and mating faces and Ra 1.6–3.2 μm on general faces. The 16 simultaneous 5-axis machining centers and the Ø400 mm rotary table handle the angled sensor pockets in one setup, which keeps the datum consistent.
Fifteen years of this work, three wholly-owned plants, 127 high-precision CNC machines, and 100% inspection before shipment. Material certificates and inspection reports come with the parts on request. Uploads stay confidential, and an NDA is available if your drawings are sensitive.
Setup and check sequence for a suspect head
- 1Check the mount firstIndicate the sensor bracket face with a dial test indicator. Anything past 0.05 mm runout or flex under hand pressure is a hardware problem, not a tuning problem.
- 2Recalibrate after every nozzle changeFit the nozzle, then run the head calibration routine at a known standoff. Expect the constant to shift by roughly 0.1–0.2 mm if the nozzle brand changed.
- 3Verify the standoff windowSet 0.8–1.2 mm for 1–3 mm mild steel and confirm with a feeler gauge or the machine's own offset screen before running production.
- 4Watch the pierce reacquisitionLog the height trace through a pierce. If the head does not settle within the programmed pierce time, extend the blanking window or lower the pierce power.
- 5Check thermal creep over a shiftMeasure standoff cold and again after two hours of cutting. More than 0.05 mm of creep points at the mount material or the cooling path, not the sensor.
- 6Inspect the ceramic ringCracks, metal spatter, or a chipped seat change the sensing circuit. Replace the ring before chasing electronics.
Capacitive vs contact sensing on an FK-4020 head
Pick the channel by material and surface condition, not by habit.
| Condition | Capacitive channel | Contact channel |
|---|---|---|
| Mild steel 1–3 mm | First choice, standoff 0.8–1.2 mm | Backup only |
| Thin stainless, aluminium | Works, keep gap low and stable | Use when dross or oxide confuses the reading |
| Non-conductive sheet | Will not read | Required |
| Slag or heavy dross on top | Reading drifts, needs blanking | More tolerant of surface debris |
| High feed, wavy sheet | Needs 1 kHz sampling to keep up | Follows the surface physically |
| Piercing window | Signal blanks, then reacquires | Unaffected by plasma noise |
| Nozzle change | Needs recalibration | Less sensitive to nozzle wear |
When to fix the sensor, when to fix the mount
If the height error appears suddenly and repeats at every nozzle change, replace or recalibrate the sensor. If it drifts slowly across a shift, or grows with feed rate, the mount and the thermal path are the real problem, and a new sensor will not fix it.
Common questions on this head
Can a fiber laser cutting torch head sensor work on non-conductive material?
A capacitive channel needs a conductive workpiece to form the second plate of the circuit. On plastics, composites, or coated non-metals it will not read a stable gap.
The usual answer is a contact or mechanical follower channel, which touches the surface and follows it directly. That costs you speed and adds wear parts, so most shops only fit it for jobs that genuinely need it.
Why does the height drift after a few hours of cutting?
Heat. The head, the bracket, and the ceramic ring all grow with temperature, and the aluminum parts grow fastest. A 100 mm aluminium bracket expands about 0.05 mm over a 20 °C rise.
The fix is usually mechanical: a shorter load path, a stiffer material near the sensor seat, or better cooling around the head. Recalibrating mid-shift treats the symptom and costs you production time.
How often should the mount be replaced?
There is no hour count that fits every shop. Check it whenever you see a height error that survives recalibration.
Indicate the mating face and look for fretting, dents from spatter, or a dowel that no longer holds. A mount that has lost its face flatness will not hold calibration no matter how carefully you set it.
Does nozzle quality change the height reading?
Yes. The nozzle is part of the sensing electrode. A different orifice size, a different brand, or a worn tip all change the effective area and shift the calibration constant.
Keep to one nozzle specification where you can, and recalibrate after any change. That single habit removes a large share of unexplained height faults.
What tolerance do the machined sensor parts need?
Mating faces and dowel locations are the critical features. We hold ±0.005 mm on those, with Ra 0.2–0.8 μm on faces that seat against the sensor body.
General outer profiles and clearance holes can sit looser. Putting the tight tolerance only where it changes the reading keeps the part affordable.
Can you machine a replacement mount from our drawing?
Yes. Send the drawing or a sample and we return a quotation with a free DFM analysis within 12 hours. Production can start within 24 hours, and parts ship in 3–5 days.
There is no minimum order quantity, so a single replacement bracket and a 10,000-part run go through the same process. Uploads are secure and confidential, and an NDA is available on request.
Send us the mount drawing, get a real quote
Upload your sensor bracket or housing drawing and we return a quotation with a free DFM analysis within 12 hours.
12-hour quote±0.005 mm tolerance100% inspection
More from the shop floor
We publish setup notes, tooling trials and inspection data from the factory floor.
