What Is a Laser Interferometer?

• 1
  How a laser interferometer turns light into numbers
• 2
  Single-frequency vs two-frequency laser interferometers
• 3
  Accuracy limits: air, wavelength and geometry
• 4
  What a laser interferometer actually measures on a CNC machine
• 5
  When a laser interferometer is the wrong tool
• 6
  Comparing measurement tools for a CNC shop floor

A laser interferometer splits one beam into two coherent paths. One path leaves the head, hits a fixed reflector, and comes straight back. The second path hits a moving reflector mounted on the axis under test and returns along the same line. Because the two beams started from the same source, they recombine at the detector and interfere.

The detector sees bright and dark bands, called fringes. Each fringe corresponds to a change of one half wavelength in the path difference. At 633 nm that is roughly 0.316 μm of travel per fringe. Electronics count fringes and interpolate between them, so the readout resolves a small fraction of a fringe rather than the full band.

This is why the instrument reads displacement, not position in space. It compares the moving reflector against the fixed one and reports the difference. Mount the fixed reflector on a wobbly stand and the numbers drift with the stand. Optical alignment matters more than the display resolution.

• 1
  SourceStabilized helium-neon laser at 633 nm, or a two-frequency head
• 2
  Beam splitterSends half the light to the moving path
• 3
  ReflectorCube corner or plane mirror on the axis under test
• 4
  Counted unitOne fringe = half wavelength = about 0.316 μm

A single-frequency head sends one optical frequency down both paths. When the reflector moves, the returning light is shifted by the Doppler effect and the count rate tells you speed. Direction is worked out from the phase relationship between the two returning beams. Alignment is fussy, and the count can be lost if the axis moves fast enough to blur the signal.

A two-frequency head splits the beam into two slightly different frequencies, usually by Zeeman splitting in the tube or by an acousto-optic modulator. The two paths now differ in frequency, so the detector reads a beat signal instead of a static pattern. When the reflector moves, the beat frequency shifts up or down, and the sign of that shift gives direction directly.

That difference matters on a machine tool. On a machine with rapid traverse of 20 m/min or more, a two-frequency head keeps a clean count through acceleration and reversal. A single-frequency setup can still work, but you should slow the traverse and expect to re-align if you bump the optics.

Cost and drift follow the same split. Single-frequency systems are simpler and cheaper but drift with air temperature and pressure unless you compensate. Two-frequency systems hold their calibration longer and tolerate more dirt on the optics, which is why most service groups carry them for machine acceptance testing.

• 1
  Single-frequencyCheaper, simpler optics, more alignment-sensitive
• 2
  Two-frequencyDirect direction sensing, stable count at high traverse
• 3
  Common useTwo-frequency units dominate machine tool calibration
• 4
  Both needAir temperature, pressure and humidity compensation

The wavelength of light in air is not constant. It changes with temperature, barometric pressure and humidity. A 1 °C rise over a 1 m path shifts the reading by roughly 0.9 μm. On a large gantry with a 4,000 mm travel, that is a real error, not a rounding note. Good practice is to let the sensor settle, log the environment, and feed the compensation into the software.

Wavelength compensation handles the air. It does not handle geometry. Straightness, squareness and angular errors of the axis itself still show up as part of the measurement unless you use the right optic set. A linear measurement with a plane mirror reflector picks up angular error as an apparent length change. A cube corner reflector is far more forgiving of small angular wobble.

Dead path is another quiet error. If the fixed reflector sits away from the zero point you care about, the uncompensated air in that gap adds to every reading. Keep the dead path short or enable dead path compensation in the software.

Finally, resolution is not accuracy. A readout that resolves 1 nm on a stable bench still carries the uncertainty of the whole optical chain. Treat the specification sheet as the best case, then add the mounting, the environment and the operator.

• 1
  Air compensationLog temperature, pressure and humidity for the whole run
• 2
  Reflector choiceCube corner for linear work, plane mirror for straightness
• 3
  Dead pathShorten it or compensate it in software
• 4
  Resolution vs accuracySmall digits do not fix a wobbly mount

The most common job is linear positioning error on each axis. The moving reflector rides on the table, spindle head or saddle. The machine is commanded to a series of target points, usually forward and backward, and the interferometer records the true position at each stop. The software then fits a compensation table you can load into the control.

Beyond linear error, the same head with different optics reports straightness, squareness, pitch, yaw, and rotary axis error. Squareness between two axes is measured with a straightness or squareness optic set and a bit of patience. Rotary tables are checked with an angular interferometer or a polygon and autocollimator pair.

The instrument also measures velocity and vibration. Feed drive tuning, servo bandwidth and settling time can be checked from the same data. If a machine passes positioning but fails at the surface finish stage, the velocity trace often explains why.

Do not confuse this with a ballbar test. A ballbar circles two axes together and reports circularity, backlash and servo mismatch in one short run. A laser interferometer goes one axis at a time but reports absolute distance error over the full stroke. Most machine acceptance programs use both.

• 1
  Linear positioningMain use on mills and lathes, feeds compensation tables
• 2
  Straightness and squarenessNeeds different optics, not just a linear setup
• 3
  Velocity and vibrationUseful for drive tuning and settling checks
• 4
  BallbarBetter for contouring error across two axes

If the question is whether a finished part is in tolerance, the interferometer is the wrong instrument. It measures the machine, not the workpiece. A CMM, a height gauge or a micrometer answers the part question. The laser answers the machine question, and it needs a machine that is already in reasonable condition to give meaningful data.

If the travel is short, say 100 mm, and the feature tolerance is loose, a granite square and a dial indicator will do the job in ten minutes with no compensation software. Setting up an interferometer for that is overkill. Reserve it for long travels where the error accumulates.

Dirty or hot shops are a real problem. Airborne oil mist and grinding dust settle on the optics and scatter the beam. Direct sunlight on the laser head or a strong draught across the beam path both shift the reading. If the shop cannot hold a stable temperature for the duration of the test, the result will not be repeatable.

Alignment time is the other cost. Plan roughly an hour for a clean linear setup on one axis, more for straightness or squareness. That is why calibration is usually scheduled around a shutdown or a slow week, not squeezed into a running shift.

• 1
  Use it forMachine acceptance, compensation tables, long travel axes
• 2
  Skip it forPart inspection, short travels, loose tolerances
• 3
  Shop conditionsStable temperature and clean air are part of the method
• 4
  Time budgetAbout an hour per axis for a clean linear setup