The reference sample is mounted. Acquire a scan, then inspect a pixel on each terrace.
Move through focus, record the interference signal, and find the centre of its modulation envelope. Each valid peak becomes one height measurement.
Use the mounted reference terraces, 10× objective, −3 to +3 µm scan and 2 ms exposure. Acquire, select pixels on both plateaus, then inspect the measured step. Try a narrower scan or higher exposure to see why measurements fail.
Scalar CSI signals, seeded camera noise (Gaussian approximation to shot noise at high counts), clipping, envelope localization, stage position, calibration scale errors, invalid masks, region analysis and a common-tilt step fit. Results come from acquired signals; changing the view does not change a measurement.
This browser prototype uses a 128 × 128 sensor crop, one surface element per pixel and a Gaussian coherence envelope. Diffraction, multiple scattering, material phase changes, films, high-NA effects and edge artifacts are not quantitatively modeled. The 3D instrument is schematic. This is training data, not certified metrology.
PSI, confocal profiling and focus variation remain later milestones. The lab and learning checkpoints save in this browser when storage is available. Export a measurement for a portable copy.
Save the height arrays, validity masks, settings, processing history, selected signal and a labeled preview together.
All results are labeled simulated. The archive is prepared on your device.