Laser Interferometer: How Machine Tool Accuracy Gets Measured

• 1
  What a laser interferometer measures
• 2
  Five degrees of freedom from one axis
• 3
  Six error sources that ruin good data
• 4
  How to run a calibration you can defend
• 5
  Where the method falls short
• 6
  Laser interferometer vs other machine checks

A laser interferometer does not measure position the way a caliper does. It counts interference fringes. A stabilized laser beam is split into a reference path and a measurement path. The measurement beam travels to a retroreflector mounted on the moving machine axis, comes back, and recombines with the reference beam. Where the two beams are in phase, you get a bright fringe; where they cancel, a dark one.

Because the laser wavelength is known and stable, each fringe crossing equals a precise slice of travel. A helium-neon source at 632.8 nm is common. Count the fringes, apply the refractive index of the air the beam passed through, and you have displacement. That is the whole trick.

The resolution comes from fringe subdivision, not from the wavelength alone. Good electronics can split one fringe into hundreds of parts, which is how a system reads to 1 nm or better over a metre of travel.

This is why the laser interferometer sits at the top of the metrology chain. It is not comparing your axis to a physical master that itself wears. It is comparing your axis to light.

• 1
  Reference beam stays inside the headIt sets the phase baseline the measurement beam is judged against.
• 2
  Retroreflector rides the axisA corner cube returns the beam parallel to its incoming path, so small misalignment does not kill the signal.
• 3
  Fringe count = displacementWavelength and air refractive index convert counts into millimetres.

A single linear setup gives you positioning error along the axis of travel. Add optics and you can pull more out of the same beam. The standard kit reads five of the six degrees of freedom for one axis: linear positioning, horizontal straightness, vertical straightness, pitch, and yaw. Roll usually needs a separate electronic level or a different optical arrangement.

That matters for a machine tool because the errors are not independent. Yaw on a long X axis turns into a straightness error at the tool tip. Pitch bends the axis into an arc. If you only check positioning, you can pass a machine that still cuts a taper.

For a 5-axis machining centre the picture gets harder. Rotary axes add their own angular errors, and the linear and rotary errors stack. We calibrate our 16 simultaneous 5-axis centres this way because a part that is within ±0.005 mm on a test block can still be out of tolerance when the rotary table swings.

Not every shop needs all five. A 3-axis mill doing flat work often gets enough from linear positioning plus straightness. A mill-turn centre or a machine running long parts usually needs the angular terms too.

• 1
  Linear positioningError along the direction of travel. The baseline check.
• 2
  Straightness, two planesHorizontal and vertical drift of the axis from a true line.
• 3
  Pitch and yawAngular tilt in the vertical and horizontal planes.
• 4
  RollRotation about the axis of travel. Needs extra hardware.

The instrument is better than the environment around it. Most bad laser interferometer results trace back to one of a handful of causes, and most of them are thermal.

Air temperature is the big one. A 1 °C change in ambient temperature shifts the refractive index enough to move the reading by roughly 1 ppm. On a 1 m axis that is about 1 µm of phantom error. The compensation sensor must sit in the beam path, not on the machine column, or it corrects the wrong air.

Air pressure and humidity act the same way. A storm front moving through during a two-hour calibration run will bend your curve. Log the environmental data and discard runs where it drifted.

The machine itself is the second problem. Steel lead screws expand at about 10.8 ppm/°C, so a 1 °C rise on a 1 m screw pushes the end nearly 10.8 µm. A machine that has not soaked to room temperature will report errors that are really just thermal growth. Let it soak overnight before you start.

Dead path is the third. That is the distance between the interferometer and the measurement start point when the reflector is at its zero position. If the air in that dead path changes but the beam is not moving, the count drifts anyway. Keep the dead path short and keep it in the same air as the measurement path.

Finally, cosine error. If the beam is not parallel to the axis of travel, every reading is scaled by the cosine of the misalignment angle. At 1° it is negligible, at 5° it is not. Align the optics with a dial indicator before you trust a micron.

• 1
  Air temperatureAbout 1 ppm per °C. Sits in the beam path, not on the casting.
• 2
  Pressure and humidityWeather changes during a run corrupt the curve.
• 3
  Machine thermal growthSteel screws move about 10.8 ppm/°C. Soak the machine first.
• 4
  Dead path driftAir changes in an unmoving beam still count as travel.

Set the machine to a known thermal state. If it has been running, let it idle to a stable temperature, or better, start cold and log the warm-up. Record the ambient temperature, pressure, and humidity at the start, middle, and end of the run. If those numbers move, the run is suspect.

Mount the optics rigidly. A magnetic base that shifts by a few microns mid-run will show up as a step in the data. Align the beam to the axis with a dial indicator and check the return spot on the target.

Take multiple runs and compare them. A single pass tells you what happened once. Three passes tell you what is repeatable. If the positioning curve changes shape between runs, you have a thermal or mechanical problem, not a measurement problem.

Feed the results back into the machine's error compensation table. Most modern controls accept pitch, straightness, and positioning corrections per axis. Re-measure after compensation to confirm the fix, and recheck at a later date to see whether the machine is drifting over months.

For our own shop, this is the same discipline that backs the ±0.005 mm tolerance we quote. A machine that is not calibrated cannot hold that number, no matter how good the operator is.

• 1
  Log environment three timesStart, middle, end. Movement means the run is weak.
• 2
  Align before you measureDial indicator on the optics, return spot on target.
• 3
  Three passes minimumRepeatability matters more than one clean curve.

A laser interferometer is a displacement instrument. It measures motion along a beam, not the geometry of a finished part. If you want to know whether a bore is round or a face is flat, you need a CMM or a roundness tester. The two tools answer different questions.

It is also slow. A full five-axis volumetric calibration can take a day or more, and the machine is not cutting during that time. For a shop running three shifts, that downtime has a real cost, so you schedule it around production rather than the other way round.

The beam needs line of sight. Long machines with guards, way covers, or a rotary table in the path may need the covers moved or the setup split into segments. That adds time and adds opportunities for the optics to shift.

And it only tells you about the machine at that moment, in that thermal state. It is a snapshot, not a guarantee. Regular re-measurement is what keeps the number honest.

• 1
  Displacement, not formUse a CMM for roundness, flatness, and true position.
• 2
  Machine is down during the runSchedule it against production, not during it.
• 3
  Line of sight requiredGuards and rotary tables may force a segmented setup.