Analysis Areas Laser Plan: What the Interferometer Hardware Demands

A laser plan interferometer splits a beam, sends the two paths across a reference and a moving target, and recombines them. Every fringe you read is a distance measurement. A mirror mount sitting 20 μm off its nominal position makes the beam walk. A mount face tilted 30 arc-seconds throws the return beam past the aperture. The analysis area here is not the optics themselves; it is the metal that locates them.

Start with the datum scheme. A kinematic mount usually needs one flat face, one vee or slot, and one cone. That trio fixes all six degrees of freedom without over-constraining the optic. Machining all three features in one setup keeps them related to each other. Move the part to a second fixture and you re-introduce the locating error you were trying to remove. On our 5-axis centers we cut the seat, the vee and the cone in a single operation, which holds position to ±0.005 mm across the set.

Flatness matters more than most drawings admit. A mirror bed specified at 5 μm flatness over 50 mm will bend a 25 mm optic enough to shift the wavefront. We usually hold the seat flat within 3 μm and check it on a granite surface plate before the optic goes in. Surface finish works the same way. A Ra 0.8–1.6 μm seat is fine for a clamped mount. For an adhesive-bonded optic, we go to Ra 0.2–0.8 μm so the bond line stays uniform.

Threaded adjusters deserve their own note. A M6 × 0.5 fine thread gives about 0.5 mm of travel per turn, so a 5° turn moves the optic roughly 7 μm. Coarse threads make that kind of adjustment impossible. We cut adjuster threads on the same machine as the bore they sit in, which keeps the axis concentric and stops the optic from walking sideways as you turn the screw.

• 1
  One setupSeat, vee and cone cut together to keep the datum stack short.
• 2
  Flatness first3 μm seat flatness before wavefront error shows up in the fringe.
• 3
  Fine threadsM6 × 0.5 for adjusters; coarse pitch kills repeatability.

Thermal drift is the quiet failure mode. A 300 mm aluminum rail grows about 6.9 μm per 1 °C. Over a 20 °C shop swing that is 138 μm of apparent displacement, all of it fake. The fix starts with material choice, not with software correction.

Aluminum 6061-T6 and 7075 are the usual picks for mounts and brackets. They machine fast, take a good anodized finish, and hold tight tolerances. Their expansion coefficient is high, around 23 × 10⁻⁶ /°C. For a frame that has to stay put, 17-4PH stainless or Inconel brings that down to roughly 10 × 10⁻⁶ /°C and 13 × 10⁻⁶ /°C respectively. Invar would be better still, but it is difficult to machine and rarely worth the cost outside a metrology lab.

Mixing materials is where designs go wrong. A stainless post bolted to an aluminum base moves at two different rates. The joint sees shear, and the optic tilts. If a mixed stack is unavoidable, keep the dissimilar interface short and put the optic seat on the more stable side. We often cut a thin stainless insert and press it into an aluminum housing for exactly this reason.

Mass helps too, up to a point. A heavier base resists vibration and slows the rate at which the frame follows an air current. It also takes longer to reach thermal equilibrium. For a bench instrument that runs all day, mass wins. For a portable unit that gets moved between rooms, low mass plus a short warm-up is easier to live with.

• 1
  6061-T6Easy to machine, high expansion; good for mounts, poor for long rails.
• 2
  17-4PHRoughly half the expansion of aluminum; use it for the stable side.
• 3
  Avoid mixed stacksDifferent rates in one joint shear the optic seat.

Finish specifications on interferometer parts are not cosmetic. A turned face leaves a spiral that scatters light and can print into the fringe pattern. A ground face is better but still directional. For any surface the beam touches or reflects from, we aim for a non-directional finish and specify it explicitly on the drawing.

Bead blasting gives a matte, uniform surface that kills stray reflections. It also rounds sharp edges, which matters when an optic sits in a pocket and you do not want a burr chipping the edge. The trade-off is dimensional change. Blasting removes a few microns and blurs a sharp corner, so it goes on non-critical faces or after the critical dimensions are already met.

Anodizing adds 5 to 15 μm per surface and it is not perfectly uniform. A hardcoat on a bore shrinks the clearance. A clear anodize on a seat can lift the optic off its datum. When a face carries a tolerance tighter than ±0.01 mm, we mask it before plating or machine it after. Black oxide and electroless nickel behave the same way; both build thickness and both need masking on precision seats.

Laser marking is the last finish step. Minimum character height is 1.5 mm, which keeps part numbers readable without cutting deep enough to distort a thin wall. We mark on a non-functional face so nothing touches the beam path.

• 1
  Non-directional finishSpecify it on any face the beam sees; turning marks scatter light.
• 2
  Mask tight facesAnodize and plating add 5–15 μm; mask anything under ±0.01 mm.
• 3
  Shallow marking1.5 mm minimum character height, on a non-optical face.

An interferometer reads displacement down to a fraction of a wavelength. A frame that rings at 80 Hz will show that ringing in the data. Stiffness and damping are the two levers, and they pull in different directions.

Stiffness comes from geometry. A ribbed aluminum bracket is far stiffer than a solid block of the same mass. We add ribs where the load path runs and keep wall thickness uniform, because a sudden change in section creates a stress riser and a place for chatter during machining. Thin walls under 1 mm are possible but need light passes and support; below 0.8 mm the part starts to move under its own clamping force.

Damping is harder to design in. Cast iron and filled epoxy absorb vibration better than aluminum. Some builders fill a hollow frame with epoxy granite for that reason. The machined interfaces still have to be tight, because any bolted joint with a gap acts as a small spring. We hold mating faces flat and use dowel pins where a joint has to stay put under load.

Assembly fit is the final check. A bore for a 25 mm optic should be a light push fit, typically 10 to 20 μm clearance, so the optic seats without stress. Interference fits crack optics. Clearance over 40 μm lets the lens rattle and shift the beam. We measure bores with bore gauges and pin gauges and record the numbers, because the fit is what the beam actually sees.

• 1
  Ribs over massRibbed sections beat solid blocks for stiffness at equal weight.
• 2
  Dowel the jointsA loose bolted joint is a spring in the load path.
• 3
  10–20 μm fitLight push fit on optic bores; interference cracks glass.

Most interferometer hardware fails at the drawing stage, not the machine. A callout of ±0.005 mm on every dimension is expensive and often meaningless. Tighter than the datum stack can support, it just adds cost. The useful approach is to tighten only the features that steer the beam and leave the rest at general tolerance.

We inspect 100% of parts before shipment. That covers a raw material check, in-process monitoring on the critical features, and a final dimensional report. Reports are available on request, and for optical hardware we can include flatness and bore diameter data so the customer can match parts to optics before assembly.

For prototypes, the free DFM review usually catches two or three problems: a datum that cannot be reached in one setup, a wall thin enough to deflect under clamping, or a coating that will close a tight fit. Quotation and that review come back within 12 hours, and production can start within 24 hours once the drawing is settled.

• 1
  Tighten selectivelyTolerance the beam-steering features, not the whole part.
• 2
  100% inspectionRaw material, in-process and final checks before shipment.
• 3
  DFM firstFree review flags datum and coating problems before cutting.