Prototype to Production: A Custom Optics Timeline

"Two weeks for a prototype" gets quoted carelessly in optics, then missed. Here is the honest timeline of a custom filter or lens program from a factory that runs its own coating chambers — the one that lets you plan a launch instead of defending a delay.

Week 0 — the design conversation

A drawing gets a DFM review: which tolerances are inflating cost, which metrology we can verify and which we would have to buy. A functional paragraph gets a coating-design proposal — center wavelength, edges, blocking budget, tolerance plan — before any commitment. This week is free and it is where bad programs die cheaply. Skipping it is how a program pays twice.

Weeks 1-2 — standard-substrate prototypes

If the design sits on stocked substrate and format, the proto goes on the next bench run: grind, polish, coat on a small witness load, measure on the production spectrophotometer and interferometer, ship express with the measured curves. This is the fast lane most buyers believe is the whole process — it is only the proof.

Weeks 2-4 — the coater campaign

Production coatings run in campaigns: the chamber loads with like designs, a recipe per process, witness coatings between runs. Custom designs queue for a campaign slot. This is where a factory's schedule — not its capability — sets lead time, and why an honest quote states the slot it is aiming at rather than "15 business days."

Pilot lot and the data pack

The first production tranche ships with the full data pack: per-lot spectra, surface maps, dimensionals. This is the pilot whose numbers become your incoming-inspection spec and your support team's answers. Read the pack before releasing the ramp; programs that skip this step inherit surprises at ten times the volume.

Ramp and second sources

From pilot, price breaks are published in the quote and the recipe is frozen. Qualified second sources are encouraged — drawings stay yours, the process travels on paper. The optics programs that fail are rarely capability failures; they are timeline fantasies. Plan against the campaign calendar, demand the measured data at every gate, and the calendar stays boring, which is exactly what a launch needs.