In manufacturing quality control, measurement data is used to accept or reject parts. However, the measurement system itself introduces variation. Gage Repeatability and Reproducibility (Gage R&R) is the statistical method used to isolate and quantify this variation.
This guide outlines how to apply Gage R&R to optical metrology and automated optical inspection (AOI) systems, emphasizing key sources of optical variation.
1. Terminology: Variation vs. Accuracy
A critical distinction must be made between a measurement system’s variation and its accuracy:
- Accuracy (Bias): How close the average of your measurements is to the true reference value (calibrated golden standard).
- Gage R&R (Precision/Variation): The spread or dispersion of measurements when the same features are measured repeatedly. A system can be highly repeatable (low variation) but completely inaccurate if it has a constant offset (high bias).
- Gage R&R is designed to analyze measurement system variation, not measurement accuracy.
2. Components of Gage R&R in Optical Systems
Total observed process variation ($\sigma^2_{\text{total}}$) is the sum of the actual part-to-part variation ($\sigma^2_{\text{part}}$) and the measurement system variation ($\sigma^2_{\text{GRR}}$):
$$\sigma^2_{\text{total}} = \sigma^2_{\text{part}} + \sigma^2_{\text{GRR}}$$
Measurement system variation is further split into two primary components:
$$\sigma^2_{\text{GRR}} = \sigma^2_{\text{repeatability}} + \sigma^2_{\text{reproducibility}}$$
Repeatability (Equipment Variation - EV)
Repeatability is the variation observed when the same operator measures the same part multiple times using the same equipment under identical conditions. In optical systems, repeatability variation typically stems from:
- Sensor Noise: Thermal noise in CMOS/CCD sensors affecting pixel intensity values.
- Illumination Drift: Minor fluctuations in LED or halogen light source intensity or spectrum.
- Auto-focus Repeatability: Subtle axial positioning errors of the focus stage, leading to variations in edge sharpness.
- Vibration: High-frequency mechanical vibrations in the environment causing pixel jitter.
Reproducibility (Appraiser Variation - AV)
Reproducibility is the variation observed when different operators (or different test stations, fixtures, or shifts) measure the same parts using the same system. In automated optical inspection, reproducibility variation typically stems from:
- Part Loading / Fixturing: How a human operator places the part in the nest. Minor tilt or rotational offsets can change the light path and shadow placement on the camera sensor.
- Calibration Protocol: Differences in how operators perform the daily dark-frame or white-balance calibration.
- Multi-Station Matching: Subtle hardware discrepancies between “identical” test stations (e.g., lens-to-lens distortion differences or camera-to-camera sensor gains).
3. Evaluation Criteria (AIAG Guidelines)
In accordance with standard automotive and consumer electronics quality guidelines (AIAG), a metrology system is evaluated based on its %GRR relative to the product specification tolerance width (USL - LSL):
- $%GRR < 10%$: Excellent. The measurement system variation is negligible relative to the part specifications.
- $10% \le %GRR < 30%$: Acceptable depending on the criticality of the measurement and the cost of the gage.
- $%GRR \ge 30%$: Unacceptable. The measurement system’s variation is too large to reliably distinguish between good and bad parts.
Synthetic Data Example
Suppose a critical waveguide parameter has a tolerance width of $1.0\text{ mm}$ (e.g., spec is $12.0\text{ mm} \pm 0.5\text{ mm}$). If a Gage R&R study computes a measurement system standard deviation ($\sigma_{\text{GRR}}$) of $0.03\text{ mm}$:
$$\text{Study Variation (6-sigma width)} = 6 \times \sigma_{\text{GRR}} = 0.18\text{ mm}$$ $$%GRR = \frac{0.18\text{ mm}}{1.0\text{ mm}} \times 100% = 18%$$
The system is classified as acceptable but leaves room for improvement.
4. Optical Gage R&R Caveats and Best Practices
- Avoid Dynamic Fixture Over-tightening: Clamping parts too tightly can introduce transient stress birefringence in glass components, changing their optical performance temporarily during measurement.
- Thermal Stabilization: Run your light sources for at least 15-30 minutes before starting a Gage R&R study to minimize early thermal luminance drift.
- Randomize Measurement Order: Always randomize the order of parts presented to the operators during the study to prevent temporal drift or operator fatigue from biasing the statistical analysis.