Four-Wire (Kelvin) Resistance Measurement

A practical reference for sub-milliohm resistance testing in battery pack hardware: cell tab welds, busbar joints, contactor and relay contacts, and BMS current shunts.

Cell tab-to-busbar welds, module interconnects, contactor and relay contacts, and BMS shunt resistors all sit in the same resistance band: tens of micro-ohms to a few milliohms. At this scale, a standard two-wire DMM reading is dominated by the test leads and probe contacts, not the joint being tested. The four-wire, or Kelvin, technique removes that error and is the only reliable way to characterise resistance at battery pack current-carrying interfaces. This reference covers the underlying principle, practical probing technique, the main battery-specific error sources, and how production and lab instruments compensate for them.

1. Why Two-Wire Measurement Fails at Pack-Level Resistance

A two-wire measurement passes current and senses voltage through the same pair of leads, so everything in that loop: lead resistance, probe contact resistance and any spreading resistance at the probe tip adds directly to the reading. A typical 0.5–1m test lead pair, including probe contact resistance, contributes on the order of 50–300 mΩ. Set against typical battery hardware joints of 10µΩ to ~300µΩ and this parasitic resistance can be 100 to 1000 times larger than the value being measured.

Take a laser-welded cylindrical cell tab joint with a true resistance of 80 µΩ. A two-wire measurement with 200 mΩ of combined lead and contact resistance reads roughly 200.08 mΩ, the joint signal is completely buried. The same joint measured four-wire, with current and voltage sensing physically separated, reads the true 80 µΩ regardless of lead length, because the voltmeter’s input impedance is typically in the giga-ohm range: sense-lead current is negligible (picoamps to a few microamps), so the IR drop along the sense leads themselves is negligible.

Parameter	Two-Wire	Four-Wire (Kelvin)
Typical usable resistance floor	~1 Ω and above	Down to low micro-ohms
Lead resistance influence	Included directly in the reading	Cancelled; independent of lead length
Contact resistance	Added directly to the DUT value	Excluded via separate sense path
Current path	Shared with voltage sensing	Geometrically separated from sensing
Typical battery applications	Harness continuity, high-resistance pre-checks	Cell tab/busbar welds, contactor & relay contacts, BMS shunts, HV connector contacts

2. The Kelvin Principle

The method is named after Lord Kelvin, who added a second pair of ratio arms to the Wheatstone bridge in the 19th century to stop bridge-arm resistance from swamping low-resistance measurements. The modern equivalent is a DMM or source-measure unit with separate Force (current) and Sense (voltage) terminal pairs, or a dedicated micro-ohmmeter / battery internal-resistance tester.

Because sense-lead current is negligible, the voltage drop along the sense leads is negligible by Ohm’s law, so the voltmeter reads only the drop across the DUT itself: R = V(sense) ÷ I(source). This holds regardless of sense-lead length, which is why Kelvin clips on several feet of cable can still resolve micro-ohm joints.

3. Probing Technique for Battery Hardware

Kelvin clips or four-point probes bring force and sense into contact as close to the DUT as possible. One half of each jaw carries current; the isolated other half senses voltage. Hence current injection and voltage sampling occur at slightly different points on the same contact pad.

For round conductors: battery cables, busbar tails or weld nuggets on cylindrical terminals, current distribution is non-uniform near the injection point. Sense probes should sit at least 1.5 cylinder-circumferences away from the current injection points, so current has homogenised across the cross-section by the time it reaches the sense point. On a flat busbar joint this typically just means landing the sense probes on the joint itself, with force probes further out on each conductor.

This setup is widely used in HV harness and cable qualification: individual strands of a multi-core conductor are progressively severed and the four-wire reading retaken at each stage, revealing hidden defects, eg: poor crimps, recessed pins, contamination or stress-corroded strands, well before they would show up on a continuity check. On the production line, the same technique is used to trend cell-tab-to-busbar weld resistance, either 100% in-line or by sample audit, since weld resistance correlates closely with weld quality and predicts long-term joint heating.

4. Contact Resistance in Pack Hardware

Every current-carrying joint in a battery pack (main and pre-charge contactors, balancing relays, bolted busbar joints, disconnect (service plug) contacts, and HV connector pins) depends on metal-to-metal contact. Joule heating at a degraded joint follows I²R, so even a small resistance increase produces disproportionate local heating at pack-level currents: at 200 A, a contact that degrades from 100 µΩ to 300 µΩ goes from roughly 4 W to 12 W of dissipation. Often concentrated at a single point rather than spread through the conductor, a common precursor to localised overheating at connector or busbar joints.

Test current matters: most contactor and busbar manufacturers specify a minimum DC test current (commonly in the 1–10 A range for low-voltage and small-signal contacts) high enough to break through thin oxide or sulphide films on the contact surface. Too low a test current under-stresses the film and reports an artificially high resistance.

Condition	Resistance vs. Baseline	Likely Cause	Action
Healthy	Within ±20% of baseline	Clean, well-mated contact	Continue routine monitoring
Watch	2–5× baseline, or a step change	Early erosion, contamination, slight misalignment	Increase test frequency; inspect at next service
Critical	>5× baseline, or imbalance between poles	Significant erosion, carbon tracking, misalignment	Schedule replacement before next duty cycle
Failure imminent	Erratic or unstable readings	Thermal damage, welding risk	Remove from service immediately

5. Error Sources at Micro-Ohm Signal Levels

5.1 Thermoelectric (Seebeck) Offsets

Below a few hundred micro-ohms, the IR-drop signal itself falls into the microvolt range, putting it in direct competition with thermoelectric EMF generated at every junction between dissimilar metals in the test loop. A temperature gradient across such a junction generates a voltage proportional to the Seebeck coefficient of the metal pair.

This matters particularly in battery packs because aluminium busbars are frequently bolted or welded directly to copper cell terminals or tabs to save weight. A copper-aluminium junction with even a small thermal gradient (eg: from a warm hand, a fan or sunlight on the test bench) can generate a few microvolts of offset, which is not negligible against a signal of similar size.

Junction	Approx. Thermoelectric Offset
Copper–Copper	Near zero (well under 1 µV/°C)
Copper–Aluminium	Low, but non-negligible — a few µV/°C
Copper–Brass / Copper–Solder	Low to moderate
Copper–Oxide film	Can be large — a sign of a degraded, non-ohmic contact

Because thermoelectric offsets are essentially DC and roughly constant over a short measurement window, they are removed mathematically (Section 6) rather than by trying to eliminate every temperature gradient on the bench.

5.2 Non-Ohmic Contacts

Aluminium forms a thin, fast-growing, highly insulating oxide layer (Al₂O₃) almost immediately on exposure to air. This is the main reason Cu–Al busbar joints in battery packs need surface preparation, jointing compound, and adequate clamp force, and why test current must be sufficient to break through any residual film. A non-ohmic contact can rectify stray RF or switching noise into a DC offset; the simplest diagnostic is to change the test current or measurement range and confirm the reading stays consistent. A stable reading indicates an ohmic contact; a reading that shifts with current indicates a non-ohmic one.

5.3 Self-Heating and Temperature Coefficient

Passing test current through a joint generates I²R heating in the joint itself. Copper’s temperature coefficient of resistance is approximately 0.39%/°C, so even a few degrees of self-heating during a sustained high-current test measurably inflates the reading. This is one reason pulse-based internal-resistance testing (HPPC-style DCIR pulses) uses short current pulses with fast sampling rather than continuous DC. The same logic applies when characterising busbar or contactor resistance at high test currents. It’s also why BMS current-shunt resistors are made from low-TCR alloys rather than plain copper: current-sensing accuracy needs to hold steady as the shunt self-heats under load.

This post has been built based on the support of and sponsorship from: illumion, Custom Power, Quarto Technical Services, Eatron Technologies and About:Energy

6. Compensating for Offset in Practice

Most dedicated micro-ohmmeters and battery internal-resistance testers (Hioki BT/RM series, Chroma, Keysight/Keithley source-measure instruments) implement one of the following methods internally. It’s worth understanding what they’re doing so odd readings can be interpreted correctly.

Current reversal: source +I then −I and average the two voltage readings; the constant thermal offset cancels, leaving only the IR-drop term.
Offset-Compensated Ohms (OCO): alternate the source current on and off and subtract the two readings. Because this rejects any pre-existing DC offset, it’s the method used to measure resistance on a live, energised busbar or cable without disconnecting load current. This is relevant for in-pack or in-vehicle diagnostics where de-energising isn’t practical.
Delta mode: a three-point alternating-polarity sequence that cancels not just a constant offset but a linearly drifting one. Mainly seen in lab-grade nanovolt instrumentation rather than production testers, where the longer measurement cycle is acceptable in exchange for sub-microohm resolution.

For most pack-level production testing, knowing which method a tester uses is enough to interpret occasional odd readings, for example: a contact reading that shifts when test polarity is reversed usually points to a non-ohmic contact or a real thermal gradient on the bench, not instrument error.

7. BMS Shunt Resistor PCB Layout

Current-sensing shunt resistors for BMS designs are themselves four-terminal devices, and the PCB layout around them needs to preserve that. If the sense traces connect to the outer edge of the shunt’s mounting pads, the measurement includes the resistance of the copper pad and solder joint as well as the shunt element — errors of 10–20%+ are common with a naive two-pad layout.

Route sense traces from the inner edge of the shunt’s pads, not the outer edge, to exclude pad and solder resistance from the measurement loop.
Use a symmetrical, dual-sided pad layout so current splits evenly either side of the shunt, keeping the sense points in a quiet, low-noise location.
Drop sense traces to a separate signal layer via vias close to the pads, routed as a tight, parallel pair, to minimise loop area and noise pickup.
For shunts connected in parallel (common in high-current BMS designs), keep sense connections local to each individual shunt. Bridging sense lines across parallel shunts couples current-sharing imbalance directly into the measurement.

8. Recommendations Checklist

Use four-wire Kelvin connections for anything below roughly 1 Ω. Essentially every current-carrying joint in a battery pack.
Land sense probes directly on the joint or component body, with force probes further out, observing the 1.5×-circumference spacing rule on round conductors.
Use current-reversal or OCO compensation where thermal gradients on the bench are likely, and prefer pulsed/short-duration test currents to limit self-heating error.
Track contact and joint resistance against a healthy baseline over time rather than relying on a single absolute pass/fail threshold.
Specify a true four-terminal Kelvin layout (inner-pad sense routing) for any BMS current-shunt design from the outset. Far cheaper to fix at layout stage than after current-sensing accuracy issues show up in the field.