High-Current Battery Testing: The Joint Fails Before the 8,000A Supply

Battery test · 5 min read · 15 cited facts

On paper, current is the solved half of battery test. The Procyon PTS 2100-20 delivers up to 8,000A, the CTS4000 cycler's channels span 0A to 3000A, and the Vega PSI 9000 source racks reach 8670A. Yet high-current test programmes go wrong, and when they do the instrument is rarely the reason. Somewhere between the output terminals and the cell there is a bolted joint, and that joint — not the supply — is the component under the most stress in the room.

The circuit does not end at the instrument

At high current the test circuit is a series chain: output stage, cable or busbar, every bolted interface, the contactor, the DUT terminals, and back. The datasheet describes the first element only. Each joint in the chain has a contact resistance set by torque, contact area and surface condition; current through that resistance makes heat, heat loosens and oxidises the interface, and the resistance rises. The chain is not static — it ages in the direction of failure, and it ages fastest at exactly the currents the programme was bought for.

Contact resistance sits inside your measurement

This is not only a reliability problem; it is a measurement problem. The resistances being measured on the DUT side are of the same order as the resistances of the connections doing the measuring — LV 148 board-net work, for instance, is defined against a source impedance of 10–100 mΩ (VDA 320), which is the scale a single ageing joint can reach on its own. The instruments are precise: the Procyon holds DC current accuracy better than 0.2% and DC voltage better than 0.1%, and the CTS4000 holds ± 0.05% FS on current and voltage. An uncharacterised, drifting resistance in series with the DUT spends that precision on measuring the state of your bolts. The catalogue itself concedes the point: the PSI 9000 carries a programmable internal resistance setting specifically to simulate wiring and connector impedance — a feature that exists because tens of milliohms visibly change results.

• Busbar cross-section is a thermal decision — sized for temperature rise across the full sequence, not for surviving one pulse
• Every bolted interface is a resistor whose value follows torque, plating and surface preparation, and it reads differently after a year of thermal cycling
• Joint resistance drifts with age, so a trend log of connection resistance is maintenance data, not paranoia
• A supply upgrade puts more current through the same joint — more heat at precisely the point that was already failing

Sense leads decide what the instrument reports

Remote sense exists because the voltage at the output terminals and the voltage at the DUT are different numbers, and at high current the difference is not small. Sense leads belong on the DUT terminal — Kelvin-connected, mechanically independent of the current path. Landed on the busbar instead, they hand the instrument your cabling drop as if it were cell behaviour, and every accuracy figure in the previous section becomes decoration. Dynamic work raises the stakes further: the AN236 loads run dynamic loading at frequencies up to 25kHz, and a profile that fast through a long, inductive current path arrives at the DUT reshaped. Short, paired, properly landed conductors are part of the test design, not part of the installation.

The floor under "zero volts"

High current also eats the bottom of the voltage window. The AN236 series is rated from 0-150V in its 150 V class, but the manual sets a minimum operating voltage of 1.8V at rated current (1.8V@200A) — a load needs voltage across it to conduct, and rated current raises that floor. Every milliohm of cable and joint between the cell and the instrument subtracts from what the instrument sees. A deeply discharged cell behind a resistive run can leave the load below its operating floor while the cell still has territory the profile was supposed to cover — a test that dies of connection design, with the instrument blameless.