Witness Sample Testing: How to Read Reflection and Transmission Data

Quick answer

Witness sample testing uses a small, melt-controlled optic that travels through your coating run alongside production parts. After coating, you measure its reflection or transmission and compare the result to your target curve. If the witness matches the specification, the run is verified; if it drifts, the data tells you whether to adjust the process before shipping parts. The key is to use a witness that shares the same substrate lot as the production optics.

Witness samples are only useful if you know what you are measuring and why. The questions below cover the practical decisions: which witness to choose, how to read the numbers, and when a witness result should stop a shipment.

Reflection vs transmission witness samples: which one do you need?

Choose based on the coating's primary function. A reflection witness sample is polished on one side and ground on the second, so it is designed for measuring the coated surface's reflectance without backside reflections confusing the data. A transmission witness sample is polished on both sides, which lets you measure how much light passes through the coated optic. If your coating is a high reflector, start with a reflection witness. If it is an anti-reflective or bandpass coating, a transmission witness is usually the better match. Some coating runs need both, especially broadband designs where reflection and transmission must balance.

What does the scratch-dig and flatness spec mean for my data?

Scratch-dig and flatness specifications define the baseline quality of the witness itself, not the coating. A 60-40 scratch-dig and flatness better than one wave at 633 nm mean the witness is smooth and flat enough that any measured change comes from the coating, not from surface defects. If the witness has a dig or scratch in the measurement area, scatter will inflate your reflectance or reduce transmission in a way that looks like a coating problem. Inspect the witness under a low-power microscope before coating. Flatness matters most for narrowband coatings, where a curved witness can shift the effective angle of incidence across the aperture.

How do I compare witness data to the production parts?

The witness is a proxy, so the comparison only holds when the witness and production parts share the same melt lot and coating run. Melt lot control means the glass's refractive index is consistent, so the witness curve should match the parts within normal process variation. If the witness passes but the parts fail, check for differences in cleaning, mounting, or position in the chamber. If the witness fails, assume the run failed unless you have a specific reason to believe the witness was mishandled. Keep the melt data sheet with the witness record so you can trace a discrepancy back to the glass source.

What should I record for each coating run?

A useful record includes the witness sample's melt lot number, the coating run ID, the measurement date, the spectrophotometer used, and the measured reflection or transmission curve. Note the angle of incidence and polarization, because a curve measured at 0 degrees will not match a specification written for 45 degrees. Record the witness's pre-coating baseline if your process allows it; a before-and-after comparison separates coating performance from witness-to-witness variation. Store the data with the production traveler so a customer question months later can be answered without re-measuring.

When should a witness result stop a shipment?

Stop the shipment when the witness falls outside the agreed tolerance band, not when it looks slightly different from the previous run. A single out-of-band point at the edge of the curve may be measurement noise; a shift across the whole band usually means a process change. If the witness is marginal, re-measure it on a second instrument or at a second location before making a decision. Document the decision either way. The witness is evidence, and a clear record of how you interpreted it protects both you and your customer.

Witness sample FAQ: practical checkpoints

Which witness type matches my coating?

Use a reflection witness for high reflectors and a transmission witness for AR or bandpass coatings. If the design balances both, run both types. Match the witness to the coating's primary function before the run starts.

What makes a witness result valid?

The witness must share the production melt lot, travel through the same coating run, and be measured at the specified angle and polarization. A witness from a different lot or run is not verification evidence.

How do I handle a marginal witness?

Re-measure on a second instrument or a second location on the witness. If the result is still marginal, treat the run as out of tolerance until proven otherwise. Re-measure before you decide, and document the decision.

Witness samples for coating verification

• Witness Sample, Reflection, 1.400-1.899: Witness Sample, Reflection, 1.400-1.899 is precision polished on one side to 60-40 scratch-dig, with flatness < 1 wave @ 633nm at 80% CA and parallelism < 1 arc min. The ground second side prevents backside reflections from contaminating reflectance data, so you measure only the coated surface.
• Witness Sample, Transmission, 1.400-1.599: Witness Sample, Transmission, 1.400-1.599 is precision polished on both sides to 60-40 scratch-dig, with flatness < 1 wave @ 633nm at 80% CA and parallelism < 1 arc min. Two polished sides let you measure transmission through the coated optic without scatter from a rough back surface.