RF retest rules should define when a new attempt is allowed and preserve all earlier results with their reasons.
Why this matters in the industry
Factories need a fair route for connection faults while maintaining reliable quality records.
The technical reasoning
Retest rules should distinguish a valid repeat, a station error and a confirmed assembly failure. Repeated trials change the probability that a marginal assembly eventually passes, so unlimited retesting can conceal instability. Preserving the initial result and the reason for each repeat supports both process diagnosis and defensible yield metrics.
Using production data to separate product and station behavior
Manufacturing measurements contain variation from products and from the test process. Stable station offsets, contact wear, reference-device drift and inconsistent retest rules can alter apparent yield. A specification limit defines a product requirement; a process-control limit describes observed process behavior and serves a different purpose. Replacing one with the other can hide defects or create unnecessary rejection, so the data model must distinguish product, station and attempt identity.
How to structure the investigation
Distinguish incomplete tests, confirmed station faults and completed DUT failures. Define authorized retest conditions and escalation paths. Keep first-attempt values and review repeated intermittent failures instead of allowing unlimited attempts until a pass appears.
Preserve first-attempt records and classify confirmed station interruptions separately from completed DUT failures. Trend results by station, product revision, channel and relevant operating state. Use repeated reference checks and controlled cross-station comparisons to investigate shifts. Release corrections or test-limit changes through a documented review, and evaluate whether earlier results were affected when a station fault is discovered.
Worked example or engineering scenario
If a marginal unit has a 70 percent chance of passing each independent trial, its chance of passing at least once in three trials is about 97.3 percent. That outcome does not establish stable performance.
Evidence to collect
| Record | Purpose |
|---|---|
| Original record | Defines the tested state and scope of the comparison. |
| Retest reason | Makes the stimulus or route condition reproducible. |
| Attempt limit | Supports interpretation of variation and possible confounding effects. |
| Intermittent failures | Connects the observation with the stated engineering decision. |
Trade-offs and common interpretation errors
Passing a retest does not erase an intermittent failure, and improved yield does not automatically show improved radio performance. A golden device can itself change. Interpret station trends alongside reference stability and measurement uncertainty, retaining enough detail to distinguish a real process improvement from a changed test decision.
What the result can support
Define retest eligibility and limits before testing, and retain attempt-level outcomes for later analysis.
Repeated passing attempts do not erase evidence of an unresolved intermittent defect.
Further technical reading
Related industry knowledge
- How to Design RF Fixture Interfaces for Frequent Product Changeovers
- Qualifying an RF Measurement Route for Factory Release
Numerical scenarios are illustrative assumptions, not reported measurements of a supplied product or installation.