Low-resistance connector measurements: what four-wire sensing, dry-circuit conditions and offset compensation each address

First mark the resistance you intend to measure

Current flows from the instrument’s Force HI through the fixture, wire and connection under test to Force LO. A two-wire setup also measures voltage along this path, so the reading includes the lead and connection resistances in the current path. A four-wire setup uses separate Sense HI and Sense LO leads to read the voltage between two points. The current at the sense terminals is very small, so the voltage drop along the sense leads can usually be neglected. Keysight’s explanation of four-wire measurement describes the current-sourcing and voltage-sensing paths separately.[1]

The key is where the Sense leads connect. If the two sense leads connect to the outer ends of two wire lengths, the reading still includes the sum of those wire lengths, the terminations and the mating contact. If the target is the mating contact, the sensing boundary must match the measurement points in the product specification. If the target is the complete termination path, retain that path in the measurement and identify it accordingly rather than calling it a single contact point.

The relationship in the schematic can be written as:

1. Force HI→
2. Fixture→
3. A↑ Sense HI→
4. Wire / termination / mating contact→
5. B↑ Sense LO→
6. Fixture→
7. Force LO

The diagram shows the measurement boundary. What lies between A and B depends on the actual measurement points. A four-wire setup usually excludes the voltage drop along the current leads outside A and B. It does not automatically exclude a wire length or another connection inside A and B.

One product example shows why measurement points cannot be universal: TE’s AMP MONO-SHAPE specification 108-20065-1 Rev. C3 groups the wire-to-IDC-slot connection, the contact-to-TAB connection and the complete connection separately. Section 5.1.1 and Figure 2 specify the measurement points for its IDC path and the 20 mV / 20 mA conditions.[2] These locations and values apply only to the items covered by that specification; they cannot serve as default limits for other connectors.

Before comparing results, both parties therefore need the same measurement-point diagram, the same path under test and the same product specification revision. Having “four-wire” on both instruments is not enough to make two values comparable.

Keep the three controls separate

Four-wire connections and dry-circuit conditions describe different dimensions of a measurement. An instrument that supports four-wire measurement may use a higher open-circuit voltage on a particular range. A setting that limits excitation may still use two-wire sensing. When preparing a measurement, record the connection method, the test current for the selected range, the source voltage limit and the compensation method separately.

Keithley’s low-resistance white paper explains the purpose of dry-circuit conditions: stronger test excitation may alter the contact surface and affect the state that the measurement was intended to observe. The paper gives common conditions of no more than 20 mV open-circuit voltage and no more than 100 mA short-circuit current. It recommends obtaining dry-circuit results before other electrical tests that may alter the contact state.[3] This is background on the method. A product specification may require a lower current or prescribe its own test sequence; the TE example’s 20 mA in the previous section is a condition for a specific item.

Check the open-circuit and current-limiting conditions here. The small voltage drop measured across a normal connection alone does not establish that the setting meets dry-circuit conditions. Check the instrument manual and product test method for the conditions used during range changes, contact opening and contact closing. If these functions have not been confirmed, pause this comparison and leave the unconfirmed fields recorded as such.

A milliohm difference after four-wire sensing: check the microvolt offset

The voltage produced by current through a low resistance is small to begin with. The following is a constructed calculation example; no physical test was performed. Assume a path under test of 10 mΩ and a forward current of 10 mA. The voltage drop across the resistance is then 100 μV. If the sense circuit also has a stable +10 μV offset, a single reading is 110 μV. Dividing directly gives 11 mΩ, which is 1 mΩ above the assumed value.

The four-wire setup has not removed the offset at this step. Section 3.3.2 of the seventh edition of Keithley’s handbook describes polarity reversal and offset compensation. Successful cancellation requires, among other conditions, that the offset remain sufficiently stable while the paired measurements are taken.[4] With a system that has positive and negative current capability and is suitable for the item under test, keep the same sensing polarity and reverse the excitation current:

For equal-magnitude currents in opposite directions:

R = (V+ − V−) / (2I) = (110 − (−90)) μV / 20 mA = 10 mΩ

E = (V+ + V−) / 2 = 10 μV

Keep the signs in the subtraction. Do not take the absolute value of −90 μV first, and do not retain only the instrument’s final resistance display while discarding the two raw voltage readings. If the actual positive and negative currents are not exactly equal in magnitude, use the signed measured currents under the assumptions of linear resistance and a stable offset:

R = (V+ − V−) / (I+ − I−)

This solves two equations of the form V = IR + E. The accompanying calculation file uses this general formula and does not treat the set current as the measured current. When the inputs use μV and mA, the resulting numerical resistance value is in mΩ.

The two-point calculation alone cannot establish that these assumptions hold. If the offset is 10 μV during the forward reading but has drifted to 14 μV during the reverse reading, the same example’s two voltages become +110 and −86 μV. The calculated result is 9.8 mΩ. Retaining raw readings, sampling times and repeated pairs helps reveal drift. Averaging suspicious readings does not establish that the error has disappeared.

If the system cannot reverse current, check the instrument’s supported “current-on / zero-current” offset compensation. The Keysight BenchVue four-wire settings document lists this function as Offset Compensation.[5] Use the mode and connections specified in the manual for the current model. The presence of this name in software alone does not establish that it is usable on a particular range and item under test.

Choose the next step from the reading pattern

This table gives a troubleshooting order. It does not directly identify a reading pattern as oxidation, a poor contact or an instrument fault. State a cause in the conclusion only after finding the corresponding evidence.

Start a sample evaluation with a defined scope

First isolate the sample from external power, use a measurement system suitable for the item under test and follow the current product method. This note concerns static low-resistance readings. Use the prescribed procedure for energized systems or other specialized tests.

1. Define what is being compared. Record the product number, specification revision, sample condition and A/B measurement-point diagram. State whether the target is the complete path, a termination or the mating contact; everyone comparing results must use the same boundary.
2. Check the available instrument functions. Obtain the four-wire connections and test current for the selected range from the manual for the current model. Also obtain the source limits, compensation method and stabilization or sampling settings. Obtain any missing information before proceeding; display digits do not replace it.
3. Arrange the test sequence. Follow the product specification to distinguish initial readings, readings after mechanical or environmental treatment, and tests with higher excitation. Mark every result group with its actual history. The applicable specification supplies the sample count, limits and stress conditions; this note gives no universal pass threshold.
4. Check the measurement chain before reading the samples. Check the method with a reference resistor suitable for the range and having a traceable value and uncertainty. Record the reference conditions and the measured result. This checks the measurement chain; it does not establish connector acceptance. Then retain the sample’s raw signed current and voltage readings, times and measurement-point diagram.
5. Separate repeatability from the effect of reconnection. Observe reading changes with the connection undisturbed, then reconnect in the planned manner. If you find drift, source-limit operation, inconsistent measurement points or conditions outside the specification, retain the abnormal results and pause the affected comparison. Do not repeatedly mate and unmate the connection to seek a lower reading.

This preparation is complete when another engineer can reconstruct the path under test, excitation, compensation and sample history from the records and explain how the final resistance was calculated. Decide product acceptance using the criteria in the same specification and the required uncertainty. Without these materials, the conclusion remains measurement preparation or method evaluation.

Record and calculation files

The blank CSV retains the sample, specification revision and measurement-point diagram reference. It also retains the method, range and source limits. Sample history and the two signed readings remain in the record. The demonstration CSV uses the constructed values above and also includes the offset-drift example. These examples are not test data or acceptance samples. The calculation script solves only for the resistance and offset corresponding to paired current/voltage readings. It leaves missing measurements blank, rejects invalid polarity and non-finite values, and does not output “pass”.

Run it with python kelvin_pair_calc.py demo-readings.csv result.csv. You can also calculate by hand using the formulas in the article. After entering real data in the blank template, first confirm that the stable-offset and linearity assumptions suit this measurement, then use the calculated results. The calculation file does not control an instrument or validate dry-circuit settings, mechanical state or the product specification.

If the review moves on to rated current and temperature rise, continue with Connector current derating: from specifications to equipment evaluation conditions in the resource library. Low-level resistance results and temperature rise under load are two different forms of evidence. If you first need to distinguish housing retention, terminal retention and the contact interface, read Connector locking review: what to check at the housings, terminals and contact interface in the same column.

Sources and conditions of use

The sources were read on October 10, 2026. This article is an AI-assisted review of technical material and compilation of measurement methods. The diagram and numerical examples are constructed illustrations; no MTTJ physical tests were performed.

• [1] Keysight: How to Measure Resistance Using Four-Wire Measurement. Used to explain the two measurement paths; it does not establish the dry-circuit capability of a particular model.
• [2] TE: AMP MONO-SHAPE Product Specification, 108-20065-1 Rev. C3, Sections 4 and 5.1.1 and Figure 2. The public text was used to check the grouping of items, measurement-point requirements and 20 mV / 20 mA. Criteria for those items do not apply to other products.
• [3] Keithley / Tektronix: Accurate Low-Resistance Measurements Start with Identifying Sources of Error, Contact Resistance Measurements. Used for the dry-circuit background and sequence explanation.
• [4] Keithley: Low Level Measurements Handbook, seventh edition, 3.3.1–3.3.5. Used for the boundaries of sensing, offset compensation, nonlinearity and self-heating. This article is explicitly tied to this readable edition; page titles for other editions are not treated as checked revisions.
• [5] Keysight BenchVue: Resistance (4-Wire). Used for the settings and meaning of Offset Compensation; check the manual for the instrument in use to establish the actual functions.

When the product specification, instrument manual, measurement points or sample history changes, repeat the affected comparisons. Also recheck excitation after a range change rather than directly carrying over the previous conditions.