Attribute Sampling
Understanding Attribute Sampling in RF Quality Control
When a factory produces 10,000 SMA connectors, testing every single one is impractical. But accepting the entire lot without testing is risky. Attribute sampling provides the statistical middle ground — inspect a carefully selected sample, classify each as pass or fail, and use the results to make a statistically valid acceptance decision for the entire lot.
How Attribute Sampling Works
The process follows Z1.4:
- Define the lot size (e.g., 10,000 connectors).
- Select the AQL (e.g., 1.0% for non-critical parameters).
- Look up the sample size code letter and corresponding sample size (e.g., 200 units for this lot/AQL combination).
- Randomly select 200 connectors from the lot.
- Test each: VSWR ≤ 1.25:1 at 18 GHz? Pass or fail.
- Count the number of failures. If failures ≤ acceptance number (e.g., 5), accept the lot. If failures > rejection number (e.g., 6), reject.
Key Equations
Attribute Sampling is a statistical quality control inspection method where each sampled item from a production lot is classified into one of two categories —...
Key specifications:
20 dB | 1.0 % | 1 a | 18 GHz | 0.3 dB
Yield: Y = e−AD (Poisson defect model)
Comparison
| Aspect | Attribute Sampling Spec | Typical Range | Impact | Design Note |
|---|---|---|---|---|
| Primary function | Unlike variable sampling (which measures... | Application-dep. | Critical | Verify in sim |
| Operating range | Understanding Attribute Sampling in RF Q... | Application-dep. | Critical | Verify in sim |
| Performance | But accepting the entire lot without tes... | Application-dep. | Critical | Verify in sim |
| Integration | How Attribute Sampling Works The process... | Application-dep. | Critical | Verify in sim |
| Trade-off | Select the AQL (e.g., 1.0% for non-criti... | Application-dep. | Critical | Verify in sim |
Frequently Asked Questions
What is the difference between attribute and variable sampling?
Attribute sampling classifies each item as pass/fail (binary). Variable sampling measures the actual value (e.g., insertion loss = 0.32 dB) and evaluates the sample's statistical distribution (mean and standard deviation) against the specification limits. Variable sampling is more statistically powerful — it extracts more information per sample — but requires calibrated measurement equipment and more complex data analysis. Attribute sampling is simpler to execute but requires larger sample sizes to achieve the same statistical confidence.
What is the Operating Characteristic curve?
The OC curve plots the probability of accepting a lot as a function of the lot's actual defect rate. An ideal OC curve would be a step function: 100% acceptance for lots at or below AQL, 0% acceptance for lots above AQL. Real OC curves are S-shaped: lots near the AQL have approximately 95% acceptance probability, lots with much higher defect rates have near-zero acceptance probability. The steepness of the OC curve depends on the sample size — larger samples produce steeper curves and more discriminating sampling plans.
Can attribute sampling catch systematic RF failures?
Attribute sampling is designed to detect random defects distributed throughout a lot. It is less effective at catching systematic failures that affect a specific subset — for example, all connectors from one cavity of a multi-cavity mold, or all amplifiers from one wafer lot. Stratified sampling (deliberately sampling from each subset) or lot segregation (treating each subset as a separate lot) improves detection of systematic issues.