Membrane Switch Tests: Two Wires or Four?

Separate current delivery from voltage sensing

In a conventional two-wire resistance measurement, one pair of leads carries the test current and provides the voltage measurement. External series resistance in that path contributes to the reading. Keysight's resistance-measurement guidance identifies residual lead resistance as one factor when deciding whether this connection is suitable. The relevance is relative to the actual path and required measurement performance, not a universal membrane-switch threshold.

A four-wire, or Kelvin, connection uses a separate pair of sense leads to measure voltage while another pair supplies current. With negligible sense current and an appropriate arrangement, voltage drops in the force leads outside the sense points are largely excluded. Tektronix/Keithley's technical article also emphasizes placing sense connections close to the intended device terminals. More wires alone do not establish a complete measurement uncertainty or a passed switch.

Have the responsible test owner select the instrument, stimulus and safe circuit state. This guide concerns a passive, isolated resistance test, not a procedure for connecting an ohmmeter to powered equipment. Follow the applicable equipment and instrument instructions; a powered controller or parallel network changes the measurement problem.

Mark what remains between the sense points

For a membrane switch, distinguish the instrument lead, the fixture-to-tail interface, the printed routes, the functional switching contact and any mating connector required by the specification. Draw the force and sense connections on that same path. A four-wire measurement at two tail pins does not remove the resistance of the closed switching contact inside the keypad; that contact is part of the device being measured.

Moving the sense points inward can exclude a route or connection that the installed equipment still uses. That may be appropriate for a diagnostic question, but it produces a different characteristic. Conversely, if a required connector lies between the sense points, its contribution remains included. Retain separately named results rather than calling every reading contact resistance.

Do not assume that four separate contacts are available on every finished tail or that opening the assembly preserves its condition. Ask the converter and test owner how the specified measurement can actually be made without changing the specimen. A proposed test coupon or access feature needs its own relationship to the delivered circuit documented.

Worked example: one path, three different readings

This is a hypothetical linear series circuit, not measured hardware or a recommended resistance specification. Assume tail pin P1 connects through a 12 Ω outgoing route, a stable 0.8 Ω closed switching contact and an 8 Ω return route to tail pin P2. The external test path adds 1.2 Ω in total. Neglect thermal offsets, sense current, instrument error, parallel paths and changes in actuation. Assume the current is the same through every element.

Connection A is a two-wire reading that includes the whole described path: 1.2 + 12 + 0.8 + 8 = 22.0 Ω. Connection B uses four-wire sensing exactly at P1 and P2, with the external 1.2 Ω outside those sense points: the ideal result is 12 + 0.8 + 8 = 20.8 Ω. The 0.8 Ω switching contact has not disappeared.

For connection C, imagine separately accessible sense points C1 and C2 immediately on either side of only the switching contact, while current still traverses the same ideal series path. The sensed result is 0.8 Ω. That is a contact-region result, not the P1-to-P2 result. Internal access is an assumption for this example, not an HYR capability or permission to dismantle a sample.

All three numbers follow from their stated boundaries. Calling C the best keypad result would confuse a smaller measurement region with a better component. For supplier comparison, first align the required terminals and specimen state, then evaluate the relevant result using the agreed measurement method and acceptance rule.

A stored zero is not live fixture tracking

Read the instrument's correction terminology rather than treating zero, null and offset compensation as interchangeable. Keysight's BenchVue four-wire help defines its null result as the reading minus a stored null value and lists offset compensation separately. NI's low-resistance tutorial describes an offset-compensation technique using measurements with test current on and off to address an unwanted voltage. These descriptions apply to their stated implementations, not every meter's menu.

A second hypothetical calculation keeps the required P1-to-P2 path at 20.8 Ω. Suppose an applicable two-wire fixture correction was stored as 1.2 Ω, but the external series path later becomes 1.6 Ω. The raw reading is then 22.4 Ω; subtracting the old 1.2 Ω gives 21.2 Ω, leaving a 0.4 Ω difference from the assumed device path. This example does not assign a cause, drift rate or probability to the change.

Record what was used for the zero, when it was established and which configuration it represents. Ask the measurement owner how fixture changes, reconnection and stability are checked. A correction should remove an evaluated measurement contribution, not delete a functional contact or connector merely to produce a favorable result. Four-wire sensing also leaves other effects, including thermal offsets, for the test owner to evaluate.

Challenge five comparison gaps

These are plausible inspection-record gaps, not reported HYR defects. Resolve them against the actual fixture and controlled drawing before making a supplier or material decision.

• Four-wire label without sense locations: request the connection sketch; the method name alone does not identify included traces or interfaces.
• A diagnostic contact result compared with a complete-path limit: keep C1-to-C2 and P1-to-P2 readings separately named.
• Functional connector subtracted as fixture error: check whether that connector belongs inside the equipment's required path.
• Old zero reused after a fixture change: preserve the raw result and correction record, then evaluate whether the correction still applies.
• Lower reading after changed actuation or test stimulus: resolve the changed specimen condition instead of attributing the difference solely to wiring.

Put the connection boundary in the RFQ

Send a revision-controlled connection record with the electrical requirement. Have suppliers identify omissions and proposed alternatives, rather than return an unexplained resistance number. These fields are an editorial handoff worksheet, not a standardized test procedure or a promise of available HYR measurement equipment.

Use the separate acceptance guide when a relevant result approaches a limit. Use the installed controller check to verify actual input behavior. Neither a Kelvin label nor agreement between two instruments establishes equipment compatibility, durability or project approval.

• Required characteristic: key, pin pair, pressed or released state, included tail routes and mating connections.
• Connection sketch: force-current entry and exit, voltage-sense points, fixture elements and any diagnostic-only boundary.
• Test conditions: instrument/method, stimulus, actuation, timing, environment and sample state selected by responsible engineering.
• Correction evidence: raw result, stored zero or other correction, reference configuration and relevant fixture checks.
• Disposition: reported value and measurement information, applicable limit and decision rule, owner and installed functional evidence.