Power_
Breaker Coordination & Selective Tripping in GPU Power Systems
Details the sequence of breaker selection, TCC curve review, and field verification required to achieve selective tripping between rack PDUs and upstream switchgear so that a single PSU fault isolates to one tray or node rather than tripping an entire rack or row.
Key facts
- IEEE Std 242 provides the recommended practice for protection and coordination studies that data-center engineers use to set long-time, short-time, and instantaneous pickup values on molded-case and power circuit breakers.
- NEC Article 240 requires that overcurrent devices be coordinated so the device nearest the fault clears first, and this applies directly to the distribution feeding GPU racks.
- GPU power supplies present inrush currents many times steady-state for the first half-cycle, which must be accounted for when setting instantaneous trip thresholds on branch breakers.
- Factory-terminated power whips and pre-wired rack PDUs shift field work to verification of breaker settings and insulation resistance rather than conductor terminations.
- An upstream breaker with short-time delay set longer than the downstream device clearing time prevents loss of an entire rack row during a single-node fault.
- Leviathan Systems performs coordination studies on the as-built single-line before energizing any rack so that settings match the installed breaker frames and trip units.
- Periodic infrared scans and secondary injection tests on breakers detect contact wear or setpoint drift before a fault occurs.
Zone-Selective Coordination Between Rack PDUs and Row Switchgear
In GPU racks the final overcurrent device is the branch breaker inside the rack PDU feeding individual power supplies. The next upstream device is the molded-case breaker in the row distribution cabinet. Selective coordination requires the downstream breaker to clear a bolted fault before the upstream device begins timing.
Time-current curves must be overlaid so the short-time delay band of the upstream breaker sits above the total clearing curve of the downstream breaker. This separation is verified with the actual trip-unit models installed, not generic curves. When the curves overlap, a fault on one tray can propagate and drop multiple nodes.
Leviathan Systems obtains the exact trip-unit part numbers from the PDU manufacturer before finalizing settings so the published curves match the hardware in the rack.
Setting Long-Time, Short-Time, and Instantaneous Elements
Long-time pickup is set just above the continuous load of the GPU tray plus margin for measurement tolerance. Short-time pickup is placed above inrush but below the available fault current at the rack. Instantaneous is disabled or set high on the upstream device to allow the downstream breaker to act first.
Many modern electronic trip units allow an I²t slope on the short-time region; this shape must be matched between upstream and downstream units so the bands do not cross. Field crews confirm each setting with the trip-unit display or a handheld programmer before the rack is powered.
Changes to any setting require re-plotting the curves and updating the coordination study document kept with the rack documentation package.
Accounting for GPU Power-Supply Inrush and Hold-Up Behavior
GPU power supplies draw a large magnetizing current when the input contactor closes. This current can exceed the instantaneous threshold of a sensitive branch breaker if the threshold was set only for steady-state load. Coordination studies therefore include the measured inrush waveform captured during factory test of the supply.
Hold-up capacitors inside the supply keep the GPU running after input voltage collapses for the interval specified by the supply manufacturer. This interval gives the downstream breaker time to open before the upstream device sees the fault. If the upstream short-time delay is shorter than this hold-up time, the entire rack may reboot even though the fault is cleared locally.
Engineers therefore request the hold-up specification from the power-supply vendor and incorporate it into the minimum short-time delay setting.
Common Field Failure Modes and How They Are Detected
The most frequent coordination failure occurs when a replacement breaker with a different trip-unit rating is installed without updating the study. The new unit may have a lower instantaneous threshold that overlaps the downstream curve. Infrared scans performed during commissioning catch temperature rise on the line-side lugs that indicates a partial trip or contact erosion that will shift the actual clearing time.
Another common error is leaving the short-time I²t function enabled on one device and disabled on the other; the curves then cross at an intermediate current level. Secondary injection testing with a calibrated test set verifies the actual pickup and timing points against the published curves before the rack is released to production.
Leviathan Systems requires a signed coordination study and test report before any rack is energized so these mismatches are caught while the breakers are still accessible.
Verification Sequence During Rack Commissioning
After mechanical installation, each rack PDU breaker is exercised and its trip settings are read with the manufacturer programmer. The values are compared to the approved coordination study. Any deviation is corrected and the study is re-issued before proceeding.
Insulation-resistance tests are performed on the power whips between the row cabinet and the rack PDU using a megger at the voltage specified in the test procedure. Readings below the minimum require rework. Once insulation is verified, primary current injection is applied at the row breaker to confirm it does not trip within the time window allocated to the downstream device.
Final sign-off includes a one-line diagram annotated with the actual settings and the date of the last injection test.
Standards referenced: IEEE Std 242 · NFPA 70 Article 240
Frequently asked_
How often should breaker settings be re-verified after initial commissioning?
Settings are re-checked whenever a breaker is replaced or the rack power distribution is modified. Annual secondary injection testing is performed on the upstream row breakers. Infrared scans are repeated quarterly on loaded racks to detect contact degradation that could alter clearing times.
What documentation must accompany a coordinated GPU rack installation?
The package contains the approved coordination study with overlaid TCC curves, the as-left trip-unit settings for every breaker, the primary injection test results, and the insulation-resistance readings. Leviathan Systems also supplies a marked-up single-line showing the zone boundaries.
Can instantaneous trip be enabled on the rack PDU breakers?
Instantaneous is normally enabled on the final branch breakers because no downstream device exists. It is disabled or set above available fault current on the upstream row breakers to preserve selectivity. The exact threshold is taken from the coordination study.
Who supplies the coordination study for a new rack deployment?
The electrical engineer of record produces the initial study. Leviathan Systems reviews it against the actual installed equipment, updates settings if the PDU or breaker models differ from the design, and re-issues the study before energization.