Map the Complete Measurement System

List device under test state, fixture, locating datums, clamps, acoustic opening, microphone and preamp, distance and angle, room or booth, absorption, background, audio interface, amplifier or load where used, cables, software, algorithm, signal, gain, limit file, database, operator steps, and environmental conditions.

The in-house testing article explains the broader role of factory test equipment. Correlation shows whether different stations make the same decision.

Define the Measurand and Product Decision

State what each result represents: response band, relative level, polarity, rub and buzz, channel balance, microphone path, input function, or another characteristic. Link it to a drawing, specification, approved sample, risk, and disposition.

Do not use a dense test graph with no decision meaning. Limits should detect defects that matter while allowing normal product and measurement variation. Keep engineering characterization separate from fast production screening where the methods differ.

Study Repeatability on Each Station First

Measure the same stable units repeatedly without changing setup, then with normal unload/reload and operator steps. Randomize order where practical. Review variation by frequency band and decision output, not only an overall correlation coefficient.

Run a Station-to-Station Correlation Study

Select stable units covering normal variation and known targeted defects. Move the same units through all stations in a controlled, randomized sequence. Use the same conditioning and product state. Compare bias, spread, frequency-dependent difference, and pass/fail agreement.

A strong linear relationship can still hide a constant offset that changes decisions near a limit. Correct physical causes where possible. If station-specific compensation is used, document its basis, validation, access, and review trigger.

Build Daily and Shift Reference Checks

Use more than one controlled reference when practical so one damaged unit does not define the system. Store references with identity, protected condition, battery or supply rule, firmware, use history, expected results, and retirement criteria. Include a known defect or electronic/acoustic check device when it improves fault detection.

Define start-of-shift, after-maintenance, after-software-change, after-microphone-change, and after-abnormal-event checks. A failed reference should automatically stop release or define a quarantine window since the last valid check.

Control Limits and Software as Released Configuration

Limit files, signal assets, scripts, algorithm parameters, smoothing, channel mapping, gain, database fields, and software versions need approval and access control. Keep effective date, station scope, approver, reason, validation, and rollback file.

The repeat-order sound-consistency guide helps connect production limits with the approved acoustic reference across orders.

Link Failures, Rework, and Retest

Store original result, failure code, station, time, product serial or lot, operator or shift where permitted, repair action, changed parts, and retest result. Repaired units should repeat the full risk-relevant sequence, not only the failed point. Prevent repeated retest from turning an unstable unit into a pass by chance.

The pre-shipment inspection guide can use test records as supporting evidence, but shipment inspection does not replace station capability.

Measurement-System Release Record

• Measurand, product requirement, failure risk, method, output, limit, and disposition are defined.
• Fixture, position, microphone, room, background, interface, cable, software, signal, and operator steps are mapped.
• Instrument noise, reload variation, operator variation, and known-defect detection are studied.
• Common units establish station bias, spread, frequency differences, and pass/fail agreement.
• Reference units have identity, storage, use history, expected results, checks, and retirement criteria.
• Software, algorithms, signals, limits, compensation, database, and access are version controlled.
• Failed references create a stop and quarantine window; maintenance requires revalidation.
• Original failure, rework, changed parts, and full retest remain traceable.

References

• International Organization for Standardization, ISO 10012:2026 — requirements for measurement management systems.
• International Electrotechnical Commission, IEC 60268-21:2018 — acoustical output-based measurements.