Short circuit, coordination and arc flash

Most of the time, a data center's electrical system carries steady, predictable load. The equipment that protects it, however, is designed for the rare moments when something goes badly wrong: a cable insulation failure, a dropped tool across busbars, a failed component inside a UPS. In those moments currents of tens of thousands of amperes can flow, enough to melt copper, burst enclosures and, if a person is nearby, cause fatal burns. Three linked studies make sure the system survives such events: the short-circuit study (how much current can flow), the coordination study (which device opens to clear it) and the arc flash study (how much energy a worker could be exposed to).

These studies are required for every data center and repeated whenever the system changes. This chapter explains what each one calculates, the simplified methods used for estimates, and how designers reduce fault and arc flash hazards without giving up reliability.

Short-circuit current

A short circuit is an unintended low-impedance path between conductors (a bolted three-phase fault, a line-to-line fault) or between a conductor and ground (a ground fault, the most common type). The current is limited only by the impedance of the sources and of the conductors between the source and the fault. The three-phase bolted fault usually gives the highest current, so it is used to check equipment ratings:

where \(Z_{\text{total}}\) is the per-phase impedance from the source to the fault. To calculate it, every impedance is converted to ohms at the same voltage (or to "per-unit" values on a common base), and the impedances in series are added:

• Utility source: from the available fault level at the service, \(Z_s = V^2/S_{\text{SC}}\).
• Transformer: \(Z_t = (V^2/S_{\text{rated}}) \times Z_{\text{pu}}\).
• Cables and busway: resistance and reactance per unit length from tables, times length, divided by the number of parallel sets.

Strictly, impedances add as complex numbers (resistance and reactance separately). Adding magnitudes gives a slightly lower current, which is acceptable for estimates; engineering software does the full calculation.

Other sources of fault current

The utility and transformers are not the only contributors. Running motors (chillers, pumps, fans) briefly act as generators during a fault and add current, typically several times their full-load current for a few cycles. Generators contribute far less than the utility: a generator's subtransient reactance is typically 10 to 20%, so a 2500 kVA generator delivers only about 5 to 10 times its rated current, and that decays within cycles. UPS inverters contribute least of all, often only 1.5 to 3 times rated current for a short time, because their electronics limit current to protect themselves. This creates a design challenge: a system protected for the high fault current of the utility must also clear faults reliably when running on generators or UPS, where fault currents are much lower and breakers may trip more slowly.

Equipment ratings

Every overcurrent device must have an interrupting rating at least equal to the fault current available at its terminals, and every assembly (switchboard, panelboard, busway, industrial control panel) must have a short-circuit current rating (SCCR) at least equal to the available fault current. A breaker asked to interrupt more than its rating can fail violently. Codes require the available fault current to be determined and, for service equipment, marked on the equipment, so that inspectors and future modifications can check ratings. A related technique, series rating, allows a downstream breaker with a lower rating to be used behind an upstream current-limiting device, but only in tested and listed combinations.

Coordination

When a fault occurs, the ideal is that only the protective device immediately upstream of the fault opens, and everything else stays on. This is selective coordination. If a fault on one rack circuit trips the main breaker of the whole critical switchboard, a small problem has become a large outage, which in a data center can mean thousands of servers going down for a failure that should have affected one rack.

Coordination is studied on time-current curves (TCCs), log-log plots of each device's operating time against current. For two devices in series to coordinate, the downstream device's curve must lie entirely below and to the left of the upstream device's curve, for every current up to the available fault current at the downstream device. Coordination is achieved by choosing device sizes and trip settings: long-time pickup and delay, short-time pickup and delay (which lets a large breaker wait for a smaller one to clear), and instantaneous settings.

The difficulty is at high currents, where all breakers trip instantaneously and their curves tend to merge. Designers solve this with breakers that have short-time withstand ratings (LV power circuit breakers can hold a fault for up to 30 cycles), carefully chosen instantaneous settings, current-limiting fuses with published selectivity ratios, and manufacturer-tested selectivity tables. Zone-selective interlocking (ZSI) adds communication: a downstream breaker that sees a fault tells the upstream breaker to wait; if the upstream breaker receives no such signal, the fault must be in its own zone and it trips without delay.

The NEC requires selective coordination for emergency systems (Article 700) and legally required standby systems (Article 701), among others. For the critical IT load of a data center, which is usually an optional standby system, it is normally an owner's requirement written into the design criteria, because the cost of an unnecessary outage is so high.

Arc flash

An arcing fault is a fault through ionized air rather than a solid connection. The arc reaches temperatures of thousands of degrees, vaporizes copper (which expands enormously), and releases intense heat and light, a pressure wave, molten metal and shrapnel. A worker in front of energized equipment when an arc flash occurs can suffer severe burns at a distance of a meter or more, even without touching anything. Arc flash injuries are among the most serious hazards in electrical work, and data centers, with their high fault currents and frequent maintenance on live systems, take them very seriously.

Incident energy

The hazard is measured by the incident energy: the thermal energy per unit area at a specified working distance (typically 455 mm, about 18 in, for LV equipment), expressed in calories per square centimeter (cal/cm²). An exposure of about 1.2 cal/cm² is the threshold for the onset of a second-degree burn on bare skin. The arc flash boundary is the distance from the equipment at which incident energy falls to 1.2 cal/cm²; inside it, arc-rated protective clothing is required.

Incident energy is calculated with the empirical model of IEEE 1584 (2018 edition), which depends on the system voltage, the arcing current (lower than the bolted fault current), the gap between conductors, the enclosure size and electrode configuration, the working distance and, most importantly, the arcing time: how long the arc lasts before an upstream device clears it. Incident energy is roughly proportional to arcing time, which gives designers their most powerful lever.

PPE categories and labels

Workers who must be within the arc flash boundary wear arc-rated clothing and equipment. Standards define four PPE categories by the minimum arc rating of the clothing:

Above 40 cal/cm², the blast and pressure effects make energized work unacceptable regardless of clothing, and the equipment must be de-energized. Every piece of equipment that might be worked on while energized carries an arc flash label stating at least the nominal voltage, the arc flash boundary and either the incident energy at the working distance or the required PPE category, together with the date of the study. Labels must be updated when the system changes.

In the US, the workplace electrical safety standard is NFPA 70E, Standard for Electrical Safety in the Workplace, used by employers to meet the general requirements of OSHA. The NEC (Section 110.16) requires arc flash warning labels on equipment likely to be worked on while energized, with more detailed information required on service equipment and other large equipment, including the date of the assessment.

Reducing the hazard

The safest approach, required by the standards, is to de-energize equipment and establish an electrically safe work condition (lockout, tagout, testing for absence of voltage) before working on it. Energized work is allowed only when de-energizing would create a greater hazard or is infeasible, and it requires a documented justification, an energized work permit and appropriate PPE. In a 2N data center, the redundant path usually makes de-energizing one side feasible, which is one of the strongest operational arguments for that architecture.

Where equipment must be worked on live, designers reduce incident energy with:

• Energy-reducing maintenance switching: a switch that temporarily sets a breaker to trip instantaneously, with an indicator, while work is in progress.
• Zone-selective interlocking and bus differential protection, which clear faults within the protected zone without coordination delays.
• Current-limiting fuses, which clear high currents in less than half a cycle.
• Arc-resistant switchgear, remote racking of breakers and remote operation, which put distance between the worker and the arc.
• Lower fault current, through transformer impedance and system design (Lesson 3).

The NEC requires an arc energy reduction method on breakers rated or adjustable to 1200 A or more (Section 240.87) and on fused equipment of the same size (Section 240.67), and NFPA 70E places elimination and engineering controls ahead of PPE.

Maintenance

Protective devices only work as calculated if they are maintained. A breaker that has not operated in years can trip slowly, which lengthens arcing time and raises incident energy far above the label value. Periodic testing of breakers and relays, infrared scanning of connections, and cleaning and torque checks are part of every data center maintenance program, and the condition of the equipment is an input to the arc flash study.