Tests such as device efficiency and battery-drain analysis measure voltage and current to compute power and other electrical parameters. The accuracy of these tests is directly dependent on the accuracy of the constituent measurements, often targeting a total error of 0.1 percent or better. At a glance, current measurement might seem like a straightforward application of Ohm's Law, but measuring large or highly dynamic currents can be deceptively challenging due to range and accuracy limitations in real-world equipment.
Digital multimeters (DMMs) can offer excellent current measurement accuracy, but many DMMs can’t measure more than 10A with their internal ammeter ranges. Clip-on current probes typically offer modest accuracies in the neighborhood of 0.5 percent to 1 percent. Though convenient to install, they suffer from poor short-term drift and must be manually re-zeroed periodically. As a result, for measuring current from 10s to 100s of amps, engineers must often build custom solutions using shunt resistors.
Unfortunately, when you build custom measurement equipment, potential sources of error are abundant, and you can’t be certain of the final accuracy without investing substantial effort to externally validate the test results.
Figure 1: A shunt resistor with Kelvin connections.
Accounting for imperfection
A precision shunt might be specified with a 0.5 percent or 0.1 percent tolerance on the nominal resistance. Without further characterization, even an expensive 0.1 percent part would consume the entire error budget before including any other error terms. Because the absolute resistance value can’t be adjusted and varies with temperature, calibration by the end-user isn’t possible, so you must settle for characterization. However, an ordinary DMM can’t directly characterize a milliohm-level shunt due to lack of resolution at low resistance values.
One way to characterize an unknown shunt is to wire it in series with a pre-characterized shunt and force current through the string with a programmable power supply. Using the known shunt to measure the current, you can measure the voltage across the unknown shunt and compute the resistance. You must wait for the shunts to reach thermal equilibrium to capture the changes in resistance resulting from temperature dependencies. This process is time consuming, and you must repeat it at regular current intervals up to the maximum anticipated current to characterize the increasing effects of self-heating.
With milliohm shunt values, you can’t ignore resistance in the package leads. For a 10-mΩ shunt, a mere 10 µΩ of extra resistance in the leads would result in a 0.1 percent error. As shown in Figure 1, a purpose-made shunt resistor with a 4-wire Kelvin connection to the resistive element is essential to prevent lead resistance from adding to the stated shunt value and affecting the measurement.