When you measure a low resistance or a small voltage drop across a component, the resistance of your wires and connectors becomes the dominant error source. A 2-wire ohmmeter puts current through the same leads it uses to measure voltage — so the reading includes lead resistance, contact resistance, and even solder joint variability. Kelvin sensing (also called 4-wire or force-sense) separates the current-carrying path from the voltage-measurement path, killing this error.

The trick: use one pair of wires (force) to push current through the device under test, and a separate pair (sense) to measure the voltage directly at the device terminals. Because the sense wires connect to a high-impedance voltmeter (or ADC input), essentially zero current flows through them — so their resistance drops zero voltage and contributes zero error. You measure the true voltage across the device, not the voltage across "device plus everything in series with it."

Real-world example: current shunt measurement. Say you're measuring motor current with a 1 mΩ shunt at 50 A. The voltage across the shunt is 50 mV. But if your PCB traces to the shunt add just 0.5 mΩ each (very realistic for a 10-mil trace), a 2-wire measurement sees 2 mΩ total — double the reading. Use Kelvin connections: bring the sense traces directly to the inner edges of the shunt's sense pads, and route the high-current force traces to the outer pads. The sense voltage now reflects only the shunt element itself. This is why precision current-sense resistors have four terminals — two big ones for current, two small ones for sensing.

Rule of thumb: If your target measurement voltage is less than ~100× the expected lead+contact resistance drop, use Kelvin. For a typical PCB with 10 mΩ of parasitic path resistance, that means any measurement below ~1 V across a sub-ohm element deserves 4-wire treatment.

Practical layout tips:

• Sense taps go inside the current path, not on the pads that carry the current. The Kelvin connection must be at the resistive element itself.
• Route sense lines as a differential pair, tightly coupled, away from switching nodes. They pick up noise easily because they're high-impedance.
• Use a difference amplifier or instrumentation amp at the receiving end — the sense lines may sit at a common-mode voltage far from ground.
• Bench multimeters often have dedicated 4-wire ohms mode; use it for anything below ~10 Ω.

The same principle applies to characterizing PCB vias, measuring battery internal resistance, calibrating shunts, and any milliohm-scale work. It costs you two extra wires and buys you an order of magnitude in accuracy.