Kelvin Sensing Cuts Measurement Errors

The venerable Kelvin-sensing circuit, or four-terminal measurement circuit, minimizes measurement errors that your test equipment and test leads can impose. It is named for William Thomson, Lord Kelvin, who invented the Kelvin bridge in 1861 to measure very low resistances.

The left-hand circuit of the figure is not a Kelvin circuit. Here, R2 and R4 represent test-lead resistances that the measurement instrument cannot distinguish from the DUT resistance. As a result, measurement errors are introduced, and for the values shown, VM1 will measure about 2 V higher than it should.

Single-Contact (left) and Kelvin (right) Measurement Circuits
Schematics created using Circuitlab.com

The right-hand circuit, in contrast, is a Kelvin circuit. Here, source (I2) and measurement (VM2) instruments each connect directly to the DUT, with each shared connection at the DUT representing a Kelvin connection as opposed to a single connection. In the figure, none of the lead resistances R6 through R9 contributes to the measurement VM2 makes. VM2’s leads will provide some resistance, but the current through the leads will be very small because of VM2’s extremely high input impedance.

Many standalone and benchtop instruments support Kelvin measurements, and if you are measuring a discrete component, you probably have plenty of room to get four leads to your DUT. The situation becomes more challenging, however, if you are dealing with an IC, particularly in a production-test environment.

As Jim Brandes of Multitest noted in a 26-page tutorial¹ on Kelvin measurements, a typical IC test system will include a DC measurement unit (DCMU) set up to force current and measure voltage. The DCMU applies its forcing current through an interface board and test contactor, whose combined resistances will introduce errors into the measurement the DCMU makes.

The situation can be improved by extending the source and measure paths independently across the interface board, but this approach still will exhibit inaccuracies due to voltage drops in the contactor. Brandes did note that lead resistances in the interface board could be accounted for (just as R2 and R4 could be accounted for in the figure by subtracting 2 V from the measured value, assuming they are constant).

But he added, “…the resistance of the contactor is not easily accounted for…because the contactor is a mechanical device, so its actual resistance changes slightly with each insertion. And a larger variation occurs over time, as the point of contact with the device wears and/or is contaminated during high-volume production test.”

The solution, he wrote, is to use a contactor that extends the four-wire connection all the way to the DUT’s lead, land pad, or solder ball. He explained that, over the years, approaches have been employed that bond force and sense contacts to one another with an insulator between them. That approach, however, requires a perfectly planar device presentation to the contactor to ensure that both the force and sense lines make contact. A better approach, he said, is to employ mechanically independent probes.

Brandes cited the experiences of one customer who evaluated a Kelvin contactor (Multitest’s Gemini) and found that, compared with the existing solution, RDSON measurements across a sample of devices dropped from 25 mW to 17 mW (a 32% improvement), and the standard deviation dropped from 35 mW to 350 µW (a 99% improvement). “With the Kelvin contact, the tester is seeing the resistance of the device, not the resistance of the contactor,” Brandes noted.

Kelvin measurements as applied to IC test do present challenges, Brandes explained: “The targets that need to be contacted with two electrically isolated, mechanically independent probes are very small and closely spaced (and getting smaller and closer).” He cautioned that Kelvin contactors and boards are more expensive than single-contact solutions and may require and more tester resources, potentially reducing the capacity for parallelism. Nevertheless, he added, Kelvin implementations—when indicated—can save money in the long run by minimizing maintenance requirements, avoiding multiple retests, and boosting yields.

He concluded, “There are many Kelvin contacts available from different vendors and using different technologies. The choice should be made carefully—as with many things in life, you often get what you pay for.”