Membrane Switch Tests: Readings Near the Limit

Name the quantity before discussing its limit

Identify the measurand: the quantity intended to be measured. For a membrane-switch resistance review, mark the endpoints and included path. A contact-only value, a pin-to-pin closed path and an installed path including a mating connection are not interchangeable records. Tie the result to the specimen and drawing revisions.

Record the actuation condition, timing, environment and electrical test conditions needed to define that quantity. Ask the measurement owner which effects are included in the reported uncertainty and which changes would make the evaluation inapplicable. A number established on a reference resistor does not automatically describe a flexible switch held in a different fixture.

NIST explains measurement uncertainty in terms of the dispersion of values attributed to the measurand using the available information. For procurement, that means uncertainty belongs with a defined result. Display resolution alone is not the missing result-specific evaluation, and changing the displayed digits does not resolve an undefined circuit boundary.

Separate a specification limit from an acceptance limit

JCGM 106:2012 distinguishes specification limits, acceptance limits and documented decision rules. Its discussion includes simple acceptance and rules using guard bands. Moving an acceptance boundary changes the treatment of false acceptance and false rejection; it does not eliminate both risks. The document does not prescribe a universal resistance limit for membrane switches.

Keep the drawing requirement visible even if an agreed rule uses a more restrictive acceptance boundary. A buyer should be able to trace both numbers to their purposes. Do not quietly rewrite the specification to match the instrument or apply a new guard band only to a supplier whose result is inconvenient.

Check what the uncertainty symbol means. NIST describes expanded uncertainty U as a coverage factor k times combined standard uncertainty. Its approximate 95% interpretation for k = 2 has conditions, including an applicable normal distribution and a reliable standard-deviation estimate. Do not treat every plus/minus label as expanded uncertainty or as a guaranteed bound on error.

Worked example: five readings and an explicit equality rule

Hypothetical HYR editorial worksheet, not measured product data. Invent a single upper specification limit T = 100 Ω, with equality permitted. Assume the measurement owner has supplied a valid expanded uncertainty U = 3 Ω for each result y under the same defined conditions. This page does not calculate or validate that U. The chosen three-branch rule is an example for discussion, not a JCGM-mandated membrane-switch procedure.

For this worksheet, support a conformity statement when y + U ≤ T. Support a nonconformity statement when y - U > T. Otherwise, label the interval as overlapping the limit and send it to the pre-agreed disposition process. Overlap is not an automatic pass or permission to reject an entire lot. The actual applicable standard, method or contract can require a different decision rule.

The table applies this rule to five invented readings. Notice the two equality cases: 97 Ω gives 94–100 Ω and meets the inclusive conformity branch. A 103 Ω result gives 100–106 Ω; its lower endpoint still includes the permitted value, so the worksheet labels it overlap rather than nonconformity.

The boundary cases matter. With U fixed at 3 Ω, the conformity branch is y ≤ 97 Ω; the nonconformity branch is y > 103 Ω; the remaining values overlap. These are decision thresholds for this invented worksheet, not replacement product specifications. An interval-based statement still has the limitations of the supplied measurement and coverage information; it is not certainty about the true resistance.

Now hold the displayed result at 97 Ω but change the assumed uncertainty. With U = 2 Ω, the interval is 95–99 Ω and supports conformity under the stated rule. With U = 4 Ω, it is 93–101 Ω and overlaps. This arithmetic alone does not show which laboratory is correct, whether either uncertainty evaluation is valid or whether their specimen conditions match. It identifies the evidence sourcing must request before comparing their decisions.

Compare supplier reports without choosing a convenient result

Request the measured value and unit, uncertainty type, coverage basis, method and specimen conditions, plus the rule used for any conformity statement. Keep an instrument accuracy specification, a calibration result and the uncertainty of the actual switch measurement separately identified. A supplier should not need to disclose confidential internal details publicly, but the responsible reviewer needs enough controlled evidence for the decision.

Agree how recorded precision and rounding enter the comparison. Retain the available unrounded result when the process supplies it; do not recover imaginary digits from a rounded certificate. If the available precision could change the branch, obtain the authorized clarification rather than rounding toward a preferred answer.

Define the retest and overlap process before testing. A changed fixture, conditioning state or actuation method can answer a different measurement question. Preserve the original result, reason for a repeat, changed conditions and responsible disposition. Repeating until a favorable reading appears is not a demonstrated uncertainty reduction or an agreed acceptance plan.

One result also remains one result. A measurement-uncertainty interval is not a sampling plan, a zero-failure confidence statement or evidence of every key and unit in a shipment. Keep lot release and installed controller behavior tied to their own approved evidence.

Five ways the acceptance record can answer the wrong question

These are review scenarios, not reported HYR defects or customer results.

• The supplier measures contact-only resistance while the order limits a longer closed path: align the measurand first.
• A meter's last displayed digit is substituted for a result-specific uncertainty evaluation: request the missing measurement basis.
• A drawing maximum is silently replaced by an acceptance threshold: retain both quantities and their controlled purposes.
• An interval touching the inclusive specification boundary is classified using the wrong inequality: test equality cases explicitly.
• A near-limit result is repeatedly retested or rounded until it passes: preserve the first record and use the agreed disposition process.

What OEM engineering and sourcing should put in the RFQ

Send the controlled drawing, pin map and specimen state with the proposed numerical requirement. Identify the exact characteristic and unit, upper or lower limit, and equality convention. Have the responsible engineering and quality owners resolve which method or contractual rule applies before asking suppliers for comparable acceptance statements.

Ask the measurement owner to specify the result and uncertainty information it can support for that configuration. Record the rule, reporting precision and the response to overlapping evidence. If a different measurement capability is proposed, review the resulting decision boundary and its suitability rather than treating a larger number of displayed digits as an improvement.

Use HYR's existing secure RFQ to submit the membrane-switch drawings and acceptance-evidence questions. This requests a scoped engineering review; it does not promise a resistance capability, accredited test service, delivery date or final equipment approval. Retain the approved record and identify which changes require a fresh measurement or decision review.