Table of Contents
Two quotations for testing the same IC can produce reports with very different decision value. One may provide only a pass stamp; another may identify the test conditions, measured values, uncertainty and the rule used to decide conformity.
When buying electronic component testing services, define the acceptance question before comparing price per unit. The specification limit belongs to the device requirement; the laboratory’s acceptance limit depends on the agreed decision rule.
What should a component testing quotation include?
A testing quotation should identify the exact devices, parameters, conditions, sample quantities, method, deliverables and decision rule. It should also state exclusions that could affect interpretation, such as untested temperature corners or unsupported functions.
For each parameter, retain the datasheet revision and the conditions attached to the requirement. A threshold measured at one supply voltage or temperature may not support a conclusion about another condition.
Distinguish destructive analysis from electrical screening and distinguish either from a full qualification program. The quotation should make clear which units can be returned and what further use, if any, the handling permits.
How should measurement uncertainty change a pass/fail decision?
Measurement uncertainty should be considered through an agreed decision rule, particularly when a result lies near a specification limit. A guard band is one approach: it places the acceptance limit inside the specification boundary by a defined amount.
Keysight’s explanation of guardbanding describes this offset between specification and acceptance limits. Its example service uses an expanded uncertainty for that offset; this does not make that exact rule mandatory for every component test.
The buyer and laboratory should agree how uncertain boundary results will be reported. An uncertainty interval is not absolute proof that the true value lies inside it, and the selected rule changes the balance between false acceptance and false rejection.
A boundary result under two decision rules
Consider a hypothetical parameter with a maximum specification of 10.00 units. The laboratory reports 9.94 units with expanded measurement uncertainty of 0.10 units under its stated coverage assumptions.
| Item | Simple measured-value acceptance | Illustrative uncertainty-sized guard band |
|---|---|---|
| Specification maximum | 10.00 | 10.00 |
| Acceptance maximum | 10.00 | 9.90 |
| Measured result | 9.94 | 9.94 |
| Decision under the stated rule | Accepted | Does not meet guarded acceptance |
The same measured value produces different dispositions because the decision rules differ. The guarded result does not, on its own, prove that the device’s true value exceeds 10.00. The report should use terminology that matches the agreed rule rather than disguising a boundary case as an unqualified statement of physical failure.
For a two-sided specification, both boundaries need an appropriate decision rule. If uncertainty is too large relative to the usable tolerance, a tighter method or a different measurement approach may be more useful than adding more test repetitions without understanding the dominant error sources.
What records make laboratory offers comparable?
Comparable offers specify the measurement scope and evidence supplied, including how uncertainty and exceptions are handled. A low unit price is difficult to interpret when fixture development, temperature control or detailed reporting is excluded.
Request the method identifier, relevant calibration status, fixture and connection approach, environmental conditions and applicable uncertainty statement. Where accreditation is required, verify the scope relevant to the actual service; do not assume a laboratory’s general accreditation covers every offered test.
Ask whether the report will include individual results, unit identifiers, applicable limits, decision rules and deviations. Determine who owns the fixture and whether a later repeat can use an equivalent setup.
Keep unit measurements separate from lot acceptance
A well-defined measurement decision applies to the units and parameters actually tested. It does not establish the quality of every untested unit in a shipment.
Use a separate incoming inspection sampling plan to connect tested units with a lot-level disposition. Sample selection, lot identity and acceptance numbers answer a different question from measurement uncertainty.
Before issuing the service order, resolve how inconclusive or disputed results will be handled: retest conditions, additional methods, retained units and any authorized escalation. Buying those rules with the test prevents the acceptance decision from being improvised after a near-limit result arrives.
Frequently Asked Questions (FAQ)
Does an instrument calibration certificate establish uncertainty for the entire test?
No. The measurement also depends on the method, fixture, connections, environment and other relevant contributors. Ask for uncertainty applicable to the reported result.
Can two laboratories report different pass/fail decisions for similar measured values?
Yes. Different uncertainty estimates or decision rules can produce different dispositions near a specification boundary. Compare those inputs before treating the difference as an error.
Does a passing electrical test prove component authenticity?
No. It establishes performance only for the tested parameters, conditions and units within the method's limits. Authenticity and lot acceptance require their own evidence.
