Nano Laser Interferometer: Non-linear Error Symptoms, Causes, Correction

• 1
  Symptom, likely cause, first action
• 2
  How a nano laser interferometer produces non-linear error
• 3
  Optical, electronic and environmental sources
• 4
  How to separate non-linear error from machine error
• 5
  Step by step: correcting non-linear error

A single-frequency nano laser interferometer counts interference fringes to resolve displacement down to the nanometer. The count should be a straight line against true travel. Non-linear error is the small curve on top of that line: a periodic deviation that repeats with the optical path rather than with the distance moved. On a 633 nm helium-neon source the classic period is one fringe, about 316 nm of displacement, though the amplitude you see depends on the receiver, the polarizing optics and the alignment.

The error is not random. If you plot deviation against position, a genuine non-linear source gives a repeatable sine-like shape. That shape is the reason the fault survives a fresh calibration. You can zero the axis, re-run the test and get the same curve within a few nanometers. Random scatter means something else is wrong, usually noise or vibration.

In practice you meet this error in two places. First, when a machine is bought off after a linear positioning test and the numbers look worse than the machine's own repeatability. Second, when a compensation table is built from bad data and then applied to the control. The second case is worse, because you are now correcting a measurement artifact with a real axis move.

Keep the scale in mind. Non-linear error on a healthy single-frequency system sits in the low nanometer range over a short travel. If your curve is tens of micrometers, you are not looking at non-linearity. You are looking at a mechanical or thermal problem that the interferometer is faithfully reporting.

• 1
  Periodic, not randomThe deviation repeats with optical path, so a re-run gives the same curve.
• 2
  Independent of distanceAmplitude stays roughly flat as travel grows; a pitch error does not.
• 3
  Small in absolute termsLow nanometer amplitude on a healthy receiver, not tens of micrometers.

Optical sources sit at the top of the list. A receiver that mixes the two polarizations imperfectly turns phase into amplitude, and the fringe count drifts as the beam intensity changes. Misalignment of the polarizing beam splitter or a waveplate rotated away from its design angle adds a second harmonic to the curve, so you see two cycles per fringe instead of one. Dirty optics do the same thing in a less repeatable way.

Electronic sources are easier to fix but harder to spot. Comparator threshold drift in the counter, limited bandwidth in the amplifier, and a reference clock that is not locked to the source all show up as period error. Cable noise adds spikes rather than a smooth curve, so you can usually separate the two by looking at the shape of the data rather than its size.

Environmental sources behave differently. Air temperature changes the refractive index of the beam path, which shifts the apparent wavelength. The result is a slow drift, not a periodic wobble, and it grows with the length of the beam in air. A 1 °C change over a 1 m path is enough to move the reading by roughly 1 μm, which will bury nanometer-level non-linearity in thermal noise.

For CNC work the practical split matters. Optical and electronic causes need the interferometer fixed. Environmental causes need the shop fixed: soak time, air shielding, and a stable foundation. Trying to correct a thermal drift with a compensation table on the control is a dead end, because the drift is not repeatable from morning to afternoon.

Run the same test twice, once with the optic mounted on the axis centerline and once offset by a known distance. Abbe offset error scales with the offset; true non-linearity does not care where the optic sits. This single test removes the most common false alarm on long beds, where the optic is often clamped to the table edge for convenience.

Then change the feed rate. Non-linear error from the receiver is independent of speed. A screw pitch error, a servo tuning problem or a resonance in the structure all change when you change speed. If the curve moves when you change feed, look at the machine first and the interferometer second.

Compare forward and reverse passes. A stable gap between the two directions points to backlash or reversal error. A curve that flips sign between directions points to a cosine or sign error in the interferometer or its setup, not to lost motion in the screw.

Finally, shorten the travel. Non-linear error keeps its shape over a short run; dead-path error and thermal drift usually shrink or vanish. If a 100 mm run is clean and a 1,000 mm run is not, you are dealing with the long path, not the optics.

Write all four results down before you change anything. Most bad compensation tables come from a technician who adjusted the machine on the first suspicious curve instead of running the checks that identify it.