Separate Product Specifications from Production Test Limits

A product specification states acceptable product behavior. A station limit defines how one measurement is classified under particular test conditions. Copying the specification into the station without considering measurement uncertainty can create inconsistent decisions near a boundary. Tightening every limit without analysis can reject acceptable products for the wrong reason.

Define the measured quantity

Record the quantity, unit, measurement point and operating condition. A regulated output measured at a connector under load is not necessarily equivalent to the voltage at the regulator’s feedback node. Include warm-up time, input supply, load and any software state that affects the result.

Identify what the measurement path adds: instrument behavior, fixture wiring, contact resistance, switching, loading and environmental variation. Calibration of the instrument alone does not characterize the entire station. Check whether the measurement procedure compensates for some effects and how that compensation is maintained.

Agree on the decision rule

Keysight’s guardbanding overview explains that guardbanding offsets acceptance limits from specification limits to account for measurement uncertainty. The appropriate rule depends on the intended decision risk; one vendor’s service policy is not automatically the right production policy for another product.

As an illustrative calculation, suppose an agreed specification is 2.90 to 3.10 V and the team chooses a simple inward allowance of 0.02 V. The station acceptance interval becomes 2.92 to 3.08 V. This arithmetic does not establish that 0.02 V is a suitable uncertainty allowance or that this rule meets the product’s risk requirements.

Keep the uncertainty basis visible

Record how uncertainty was estimated, which conditions it covers and which assumptions could invalidate it. Separate random variation from known corrections where the chosen method requires that distinction. Avoid treating a component’s tolerance as though it were the station’s measurement uncertainty.

Near-limit results deserve a defined disposition. Options may include a more capable measurement method, engineering review or rejection under the approved rule. Repeating a noisy measurement until it passes changes the decision behavior and should not become an undocumented operator practice.

Verify the station before changing limits

Use suitable reference devices or controlled inputs to examine repeatability and differences between stations. Investigate systematic offsets before widening acceptance limits. A limit change that hides a damaged fixture or incorrect load setup reduces confidence in every subsequent result.

Version the specification, measurement procedure, uncertainty assessment and station limits together. Preserve raw values as well as pass or fail status. That record allows a later investigation to determine whether a product changed, the measurement system changed, or only the rule used to classify it changed.