Design solutions that pursue the sound quality and experience you need
# Production Audio Testing: Line Calibration and Pass/Fail Limits
A design that measures beautifully in the lab can still ship defects if the production line never calibrates its test station and sets pass/fail limits by guesswork. This article is written for brand owners and factory quality engineers: how to calibrate an audio test station, how to derive defensible limits, how to hold cycle time, and what data traceability is actually worth.
R&D testing proves design margin across every parameter. Production test answers a narrower question: is this unit consistent with the golden sample? Because of that difference in purpose, a line station should never sweep everything. It locks down three to five items that correlate most strongly with perceived sound quality and field complaints — rated sound pressure level, total harmonic distortion, frequency response deviation at a few key bands, Rub & Buzz rub-and-buzz detection, and operating current. The pass criterion is rarely an absolute specification, either. It is a deviation window against a reference unit, because line conditions, fixtures, and operators can never match a laboratory environment, and absolute limits only generate false rejects.
Drift on a production station comes from three places: the measurement microphone, the coupler seal, and ambient temperature and humidity. The practice is straightforward. At the start of every shift, run the golden sample and confirm system deviation sits inside the allowed band — typically within ±0.5 dB — before releasing the line. Calibrate the microphone on schedule with an acoustic calibrator at 1 kHz and 94 dB, otherwise its sensitivity will drift slowly with temperature and nobody notices. Use the coupler per IEC 60318-4 ear simulator practice; a poor seal leaks low-frequency energy, pulls the measured SPL down, and turns a good batch into a reported failure when the real problem is insertion depth. Design the fixture so insertion depth, clamping force, and driver orientation repeat exactly, which removes most human variation.
Limits should not be copied from another product, and they should not be lifted straight from the R&D spec sheet. The correct sequence: run a pilot build of at least 30 units, build the distribution from measured data, and set the first-cut limits at mean ±3σ with an allowance for measurement uncertainty. A GR&R study is worth the effort — gauge repeatability and reproducibility below 10% is good practice, and above 30% means the test system itself cannot be trusted, in which case tightening limits only converts noise into rejects. For SPL, which is highly fixture-sensitive, relative judgment against a reference unit from the same batch works better: subtract the shift's system offset first, then evaluate the individual unit.
Half the value of a line station is judgment, the other half is cycle time. Put fast tests first — open/short, DC resistance, and a quick Rub & Buzz sweep — and leave the slow frequency sweep for last, keeping total time per unit in the low tens of seconds with parallel stations. Upload results by serial number to MES or a local database and watch the mean with SPC rather than only counting rejects; a sustained mean shift usually means a loose fixture or an incoming-material batch change. Cap retests at two. Otherwise "test until it passes" hides a process problem permanently: the yield chart looks healthy while field returns keep climbing.
Q: Which items belong in a routine production audio test?
A: Rated sound pressure level, total harmonic distortion, frequency response deviation at key bands, and rub-and-buzz detection, plus operating and standby current. Every added item costs cycle time, so keep only what correlates with perceived quality and field complaints and push the rest to type testing and sampling.
Q: Some units in the same batch pass and some fail — is it the product?
A: Suspect the test system first. Inconsistent fixture sealing, an uncalibrated microphone, or limits set at the edge of the distribution all produce this random pattern. Run a GR&R and re-check the golden sample before concluding that unit-to-unit spread is the cause.
Q: How do we shorten test time without losing coverage?
A: Order tests by duration, run fast items first and sweeps last, and parallelize across stations. Switching to relative deviation limits also cuts repeat measurements. Do not drop rub-and-buzz detection to save seconds — abnormal noise is the defect customers hear first.
Shenzhen Liwei Electronic Technology Co., Ltd. designs audio headset electronics and complete PCBA solutions with over 10 years of experience, serving 15+ well-known brands and 300+ customers across eight product lines including hearing protection, shooting sports, live streaming, gaming, telephony, and LE Audio broadcast speakers. Production test item definitions and limit recommendations are delivered together with each design. Contact us to request a quote for a custom audio solution.
Shenzhen Liwei Electronic Technology Co., Ltd. designs audio headset electronics and complete PCBA solutions with over 10 years of experience, serving 15+ well-known brands and 300+ customers across eight product lines including hearing protection, shooting sports, live streaming, gaming, telephony, and LE Audio broadcast speakers. Production test item definitions and limit recommendations are delivered together with each design. Contact us to request a quote for a custom audio solution.
Keywords: production audio testing,line calibration,audio test limits,rub and buzz detection
