Protective relay coordination study field guide for data centers

What a coordination study is, and the one job it does

A protective device coordination study sets every breaker, fuse, and protective relay in a power system so that the device closest to a fault is the one that trips, and nothing above it opens. Get it right and a short on one rack drops that one branch breaker. The rest of the building never knows. Get it wrong and the same short trips the main, and a thousand servers go dark for a fault that should have cost you a single circuit.

That is the whole point of the study, and it is worth saying plainly because the name makes it sound like paperwork. Coordination is about which device trips and which one holds. The study looks at every fault the system can throw, at every bus, and assigns the settings so the protection clears the smallest possible slice of the system every time.

The output is not a number you carry in your head. It is a set of time-current curves, a settings table for every adjustable device, and an engineer's stamp that says these settings make the system selective. On a data center, where the building exists to never go dark, that selectivity is not a nicety. It is the difference between a contained fault and an outage.

Selective coordination, and where the code makes it mandatory

Selective coordination means that for any fault, only the overcurrent device immediately upstream of that fault operates, and every device above it stays closed. The terms get used loosely. Coordination in general can mean partial coordination, good up to some current and not beyond. Selective coordination is the strict version: full selectivity across the whole range of faults the system can produce, from a light overload up to the maximum bolted fault current.

For most commercial work, selective coordination is good engineering that the owner pays for to keep the lights on. For some systems it is a code requirement, not a preference. The NEC requires selective coordination for emergency systems under Article 700, for legally required standby systems under Article 701, and for critical operations power systems under Article 708. The requirement lives in the selective coordination sections of those articles, commonly numbered 700.32, 701.32, and 708.54 in recent editions, and the section numbers have moved across code cycles, so confirm them against the adopted edition.

There is a credentialing twist that catches people. Those articles require the selective coordination to be selected by a licensed professional engineer or another qualified person engaged primarily in the design or maintenance of electrical systems, and the selection has to be documented and available to the people who design, install, inspect, and operate the system. That is the code pointing straight at a stamped study. A data center that mixes any of these systems into its power chain inherits the requirement for those parts.

How do you read a time-current curve?

A time-current curve, the TCC, is a log-log plot with fault current on the horizontal axis and tripping time on the vertical axis. For any current you pick on the bottom, the curve tells you how long that device waits before it trips. High current is to the right, and the curve drops as you go right, because more current trips faster. A coordination study plots every device in a path on the same chart, from the branch device at the bottom of the path up to the source.

Selectivity on the chart is a visual rule. Read from the load side up to the source. The downstream device's curve has to sit fully below and to the left of every device above it, with no crossing anywhere in the current range the system can deliver. If two curves touch or cross at any current, the two devices can trip together at that current, and the upstream one might beat the downstream one to it. That is a loss of selectivity, and it shows up on the plot before it shows up as an outage.

The trap is the far right of the chart, the high-current end near the available fault current. Two breakers can look beautifully coordinated through the overload region and then collide in the instantaneous region where both go to their minimum time. The study has to check coordination out to the maximum fault current at that bus, not just through the load range. A curve set that coordinates at 2000 A and crosses at 40,000 A is not coordinated where it matters most.

The coordination time interval between devices

Curves that do not touch are not enough. There has to be a gap between them, a margin called the coordination time interval, or CTI. It is the vertical separation between the downstream device's curve and the upstream device's curve at any given current. For coordinating between protective relays, the CTI commonly runs about 0.3 to 0.4 seconds, and for some all-breaker schemes the practical interval is tighter. Confirm the value the engineer of record uses against the device types and the study basis.

The interval is not arbitrary padding. It has to cover the real delays that stack up while a fault is clearing: the downstream breaker's own interrupting time, the upstream relay's overtravel after the current goes away, and the timing tolerance of both devices. The downstream device has to fully clear the fault, and the upstream device has to recognize that the fault is gone and reset, all before the upstream device would otherwise trip. The CTI buys that sequence enough room to happen reliably.

Squeeze the interval too tight to fit more devices in series and you build a system that coordinates on paper and races in the field. Modern electronic trip units and microprocessor relays are faster and more repeatable than the old electromechanical gear, which lets the interval shrink, but the breaker's mechanical interrupting time does not shrink with it. The clearing time of the device, not the cleverness of the relay, sets the floor on how tight the interval can go.

Why do you need a short-circuit study first?

You cannot coordinate devices without knowing the fault current they will see, so the short-circuit study comes first, always. It calculates the maximum available fault current at every bus, switchboard, MCC, and panel in the system, built from the utility contribution, the transformer impedances, the conductor lengths, and any motor or generator contribution. Those fault-current values are what the coordination study plots its curves against, and they are where each device curve gets cut off on the chart.

The order is not negotiable, and it is the same short-circuit model that feeds the arc-flash study. The arc-flash study needs the bolted fault current to derive the arcing current and the incident energy at each piece of gear, exactly as the coordination study needs it to place the curves. One short-circuit model, three studies stacked on it: short circuit, then coordination, then arc flash. Change a transformer or the utility available fault current and all three move together.

Anyone offering a coordination study that did not start from a current short-circuit model is guessing at the one input that controls everything. The curves will look authoritative. The cutoffs will be fiction, because they were drawn against fault currents nobody calculated. On a data center with high available fault current from a stiff service and parallel sources, the short-circuit numbers are large and the margins are tight, which is exactly the condition where a sloppy fault-current input wrecks the coordination.

The devices and what each one can adjust

Three kinds of device do the protecting, and they differ in how much you can adjust. The inverse-time circuit breaker with an electronic trip unit is the workhorse of low-voltage distribution, and the adjustable ones carry LSIG functions: long-time, short-time, instantaneous, and ground fault. Each function has a pickup, the current where it starts to act, and most have a delay or band that shapes how long it waits. The adjustability is what lets you coordinate one breaker against another in series.

The fuse is the opposite. It has a fixed melting and clearing curve set by its type and rating, and there is nothing to adjust. You coordinate fuses by selecting ratings and types so their fixed curves stack, often using the manufacturer's selectivity ratio tables. A fuse is fast and current-limiting at high fault levels, which helps arc flash, but you give up the field adjustability a breaker gives you.

The protective relay is the precision instrument, used on medium-voltage gear, large mains, generators, transformers, and buses. Relays are identified by ANSI device numbers from IEEE C37.2, and the ones a coordination study lives on are the 50 instantaneous overcurrent and the 51 inverse-time overcurrent, with the 51G or 51N for ground. The 87 is differential protection, the 27 is undervoltage, and the 59 is overvoltage. A relay does not interrupt the fault itself. It senses the condition and tells the breaker to trip, which is why the relay setting and the breaker it commands are verified as a pair.

Setting pickup and delay on an LSIG trip unit

The LSIG trip unit is four protective functions in one box, and coordinating it means setting each one with a job in mind. Long-time protects the conductor and the equipment from sustained overload. You set the long-time pickup above the circuit's continuous load so normal operation does not trip it, and below the conductor and equipment damage limit so it clears before the wire cooks. The long-time delay shapes how long an overload can persist, and it is what you slide to coordinate against downstream long-time curves.

Short-time is the middle region, for faults large enough to be a short but where you want a deliberate delay so a downstream device can clear first. You set the short-time pickup and band so the upstream breaker waits out the downstream device, holding the coordination interval. Instantaneous is the fast region with no intentional delay, and it is the function that fights coordination hardest. An instantaneous setting on an upstream breaker that picks up at a current a downstream device also sees will trip the upstream breaker at the same instant, defeating selectivity.

That tension is the heart of low-voltage coordination. To coordinate a main against a feeder against a branch, you often turn the instantaneous off on the upstream devices and let short-time carry the coordination, so the downstream device always gets its window. The cost is that the upstream breaker now holds a fault longer, which raises the arc-flash energy at that bus. The ground function, the G in LSIG, is set the same way for ground faults, coordinated against the ground protection below it.

Pickup, delay, and the shape of the curve

Every overcurrent setting answers two questions: at what current does it start to act, and how long does it wait. The pickup is the current. Set it above the largest current the circuit sees in normal service, including the inrush and the continuous-load adder, so the device rides through normal operation, and below the level that damages the conductor or the equipment it protects. The window between those two is where the pickup lives, and on a tightly loaded feeder that window is narrow.

The delay and the curve shape decide the timing. Inverse-time curves trip faster as the current climbs, and they come in shapes, commonly described as moderately inverse, very inverse, and extremely inverse, that change how steeply the time drops with current. Picking the right curve shape lets a downstream device with a steep curve clear a high fault fast while an upstream device with a flatter curve waits, which is how you hold the interval across the whole current range instead of just at one point.

The whole exercise is fitting each device's curve into the gap between the load below it and the source above it, with the conductor and equipment damage curves drawn on the same chart as ceilings. A setting that protects the conductor but loses coordination, or coordinates but lets the conductor exceed its damage curve, is not a finished setting. Both constraints have to hold at once.

How does coordination fight arc flash?

Coordination and arc-flash safety pull in opposite directions, and ignoring that is how a well-coordinated system ends up with dangerous incident energy. Incident energy is power multiplied by time, so the longer a breaker holds a fault before it clears, the more energy an arc throws at a worker. The short-time and long-time delays you add to make an upstream device wait for a downstream device, the very delays that buy selectivity, also extend the clearing time and push the arc-flash energy up at the upstream bus.

So the better you coordinate, the worse the arc flash tends to get at the mains and the larger upstream breakers, exactly where people have to work. You cannot fix it by just speeding the upstream device up, because that breaks the coordination you built. The resolution is to have it both ways in time, not at the same time. An energy-reducing maintenance switch lets a worker flip an upstream breaker into a no-intentional-delay mode while inside the arc-flash boundary, so a fault clears fast during the exposure, then the breaker returns to its coordinated settings afterward.

The NEC backs this. The arc-energy reduction article, commonly cited at NEC 240.87, requires a means to reduce clearing time on circuit breakers rated or settable at 1200 A or higher, and lists acceptable methods including the maintenance switch, zone-selective interlocking, differential relaying, an instantaneous trip, and active arc-flash mitigation. Note that some of those methods, the maintenance switch in particular, deliberately defeat coordination while engaged, which is fine because a worker is standing there and uptime is not the concern in that moment. Zone-selective interlocking is the one that cuts clearing time without giving up coordination, which is why it shows up so often on critical gear. The settings that drive coordination are the same settings that drive incident energy, so the coordination study and the arc-flash study are done off the same model and verified together in commissioning.

Differential protection (87) and zone selectivity

Differential protection, ANSI device 87, sidesteps the coordination-versus-speed tradeoff entirely for the equipment it protects. It works by comparing the current going into a zone against the current coming out. In normal operation, and even for a fault outside the zone, what goes in comes out and the relay sees no difference. For a fault inside the zone, the currents no longer balance, and the relay trips. Because it responds to where the fault is, not to how big the current is, it can trip fast without waiting on any coordination interval.

That is the value of it. A transformer differential, a bus differential, or a generator differential clears an internal fault in a few cycles, instantaneously by design, while still being perfectly selective, because it only ever sees its own zone. You get fast clearing and selectivity at the same time, which overcurrent coordination cannot give you. The catch is that differential only protects inside its zone. It is not a substitute for overcurrent coordination through the rest of the system, it is a fast, selective wrapper around the high-value assets.

On a data center this matters at the big iron. The main transformers, the main and tie bus sections, and the generators are exactly the equipment you want cleared instantly on an internal fault without a coordination delay raising the arc-flash energy. Putting differential on the transformer and the bus means an internal fault there does not have to ride through the long delays the overcurrent scheme uses for selectivity, which keeps both the clearing time and the incident energy down at the most critical, highest-energy buses in the lineup.

Coordinating ground-fault protection

Ground-fault coordination is its own layer on top of the phase coordination, and it gets skipped more than any other part of the study. Ground faults are usually lower current than phase faults, so a separate set of ground functions, the 50G and 51G on relays and the G in an LSIG trip unit, watches for current returning through the ground path. These have their own pickups and delays and have to coordinate among themselves the same way the phase devices do, so a ground fault on a branch trips the branch, not the main.

The code forces ground-fault protection onto the largest services. The NEC requires ground-fault protection of equipment, commonly cited at 230.95, on solidly grounded wye services over 150 volts to ground but not over 1000 volts phase to phase, for each service disconnect rated 1000 A or more. In practice that is the 480Y/277 V service main on most large buildings and data centers. The maximum setting is capped, commonly at 1200 A pickup with a time delay limited at the higher fault currents, so confirm the pickup and delay limits against the adopted edition.

Here is where ground-fault coordination bites in the field. If the main has ground-fault protection and the feeders below it do not, a ground fault anywhere downstream trips the main, because the main is the only device watching for it. That takes the whole service down for a fault that should have been local. Coordinating ground fault means adding ground sensing at the downstream levels and setting the delays so the lowest device clears first, which is required on these systems anyway and is one of the most common holes an inspector or a commissioning agent finds.

What the study delivers

A finished coordination study is a document set, and knowing what is in it tells you whether you got a study or a stack of curves. It opens with the system one-line diagram, marked up with the device identifiers, ratings, and the available fault current at each bus from the short-circuit study. The short-circuit results themselves are tabulated by bus. Then come the TCC plots, one per coordination path, showing every device in series on the same chart with the conductor and equipment damage curves drawn in.

The part that turns into field work is the settings table. For every adjustable device it lists the recommended settings: the long-time, short-time, instantaneous, and ground pickups and delays for breakers, the pickup and time dial and curve for relays, and the rating and type for fuses. That table is what gets loaded into the gear. The study closes with the engineer's narrative, the recommendations where full coordination could not be achieved, and the stamp.

The stamp is not decoration. For the emergency, legally required standby, and critical operations systems where the NEC requires selective coordination, the code requires the selection to be made by a licensed professional engineer or qualified person and documented. The stamped study is that documentation. A study with curves but no stamped settings table is missing the only part that changes what the breaker actually does, and a study no one ever loaded into the gear is a binder, not a protected system.

Loading the settings into the gear and verifying them

The coordination study is worthless until its settings are in the gear, and the most common power QA failure in the industry is gear left on factory defaults. Every adjustable breaker trip unit and every protective relay ships with default settings that have nothing to do with the system it was installed in. The commissioning step is to load the stamped study values into each device, then verify, device by device, that the as-left settings actually match the study. This is a slice of the broader electrical commissioning and power QA scope, and it belongs in the same record set.

The verification is not a glance at the dial. You read back the setting in the trip unit or the relay, confirm it against the settings table from the study, and record the as-left value on a settings sheet that becomes a turnover document. A breaker on its default short-time setting instead of its coordinated one will either trip the whole bus on a downstream fault, because it is too fast, or fail to clear, because it is too slow. Either way the selectivity the study paid for is gone, and nobody knows until a fault proves it.

Loading the number is only half the job. The device also has to actually trip at the setting it was given, which the trip-unit and relay tests confirm. The as-left settings sheet, tied to each device, is one of the most important documents in the turnover package, because it is the proof that the stamped study made it into the building. The next engineer who modifies the system starts from that sheet. Verify the gear, not the binder.

Secondary injection: proving the device trips at its setting

Loading a setting and proving the device honors it are two different things, and the second is what secondary injection testing is for. The test injects a simulated current or voltage straight into the relay or electronic trip unit's inputs, downstream of the current transformers, and confirms the device picks up and times out where its setting says it should. It proves the protection logic without energizing the primary system or stressing the breaker the way primary injection does.

A proper test does more than confirm a single pickup. The technician injects current and ramps it up to find the actual pickup, then checks the timing at several points along the curve, and confirms the trip output operates the breaker, including the target and seal-in. Relay and trip-unit accuracy is typically specified in the few-percent range, so a 6.0 A pickup setting that actually picks up at 5.8 A is within tolerance, while one that picks up at 7 A is a finding. This is the acceptance test, run to the NETA Acceptance Testing Specifications, ANSI/NETA ATS, by an independent test agency.

Skip it and you have a settings sheet that says the right thing and a breaker that may do anything. The study assumes the device clears at a specific time, the arc-flash labels assume the same, and the only thing that proves the assumption is a test that made the device operate at its setting. A relay that was set correctly but never injection-tested is an unverified link in the protection chain, and it is exactly the gap a thorough commissioning is supposed to close.

Coordinating with the utility upstream

The coordination does not stop at the service. The utility has its own protection above your main, a relay or a fuse on the primary, and your most upstream device has to coordinate with it. The goal is unchanged: a fault inside your building should trip your main, not the utility's device, so a fault on your gear does not take out the utility feeder and, with it, your neighbors. If your main does not clear before the utility's device times out, the utility opens and you have a larger outage than the fault deserved.

The utility usually dictates the terms. Many utilities publish required or maximum settings for the customer's main protection, and some require a relay coordination review or impose a specific time-current characteristic on the service so the customer device stays selective with theirs. Those requirements are not optional, and they come from the utility, not the engineer of record, so they have to be gathered early. The study coordinates down from the utility's curve, treating it as the top of the chart.

The medium-voltage primary protection is where this lives on a large data center service. Confirm the utility's protection data and any setting constraints before the study is finalized, because a study that coordinates beautifully internally and ignores the utility curve can still drop the whole service on the first feeder fault. The utility curve is one of the inputs you cannot assume.

When does a coordination study go stale?

A coordination study is valid only for the as-built system it was done on. The settings are tuned to specific fault currents, specific transformer impedances, and a specific device lineup. Change any of those and the study no longer describes the system in the building. The most dangerous part is that the gear keeps running on the old settings, which were correct for a system that no longer exists.

The changes that invalidate a study are the ones crews make and forget to flag. A new or replaced transformer changes the available fault current at every bus below it, which moves where the curves get cut off and can break coordination that used to hold. A utility upgrade raises the fault current at the service. A breaker swap or a trip-unit replacement can change the available functions or the defaults. Added feeders and new loads reshape the curves. Each one is a reason to re-run the affected part of the study, not the whole thing necessarily, but the parts the change touched.

This is the same trigger logic that drives the arc-flash review, and for the same reason: the short-circuit model underneath both moved. The arc-flash study has a documented outer review interval, and a system change resets the clock for both studies. When a transformer or a source changes, the coordination study and the arc-flash study are both stale until someone re-runs the affected buses and reloads the settings. A label and a settings sheet that describe last year's system are worse than none, because they tell a worker and an operator the protection is sorted when it is not.

The data center case: 2N, selectivity, and maintainability

Data centers raise the stakes on coordination because the building is built to never drop the critical load, and they complicate it because of the topology. A 2N system has two complete, independent power paths to the load, often with tie breakers between them, and the coordination study has to hold selectivity across the normal configuration and the maintenance and failure configurations, when a tie is closed or one path is down. A setting that coordinates with both paths running can lose selectivity when the system reconfigures, so the study covers the operating modes, not just the normal one.

Selectivity is the whole reason the topology pays off. The redundancy on the one-line only delivers if a fault on one branch stays on that branch instead of cascading up and taking out a path, which is precisely what the coordination study enforces. Pair that with the differential protection on the mains, transformers, and bus to clear internal faults fast, and the energy-reducing maintenance switches on the large breakers so technicians can work the gear safely, and you get a plant that is both selective and maintainable.

Maintainability is the part that gets shortchanged. People have to open and work this gear while the building runs, which is why the maintenance switches and the as-left settings verification are not optional extras on a data center. The coordination and arc-flash work is part of the commissioning turnover, and the critical check is that the as-left settings in every device match the stamped study, because a single breaker on defaults breaks the selectivity for the whole path it sits on and changes the arc-flash energy at the same time.

What to document

The study and the as-left settings are only worth what the record can prove later. The settings sheet is what the next engineer reads before touching the system, and the as-left values are the proof the stamped study made it into the gear. Capture each device and its function settings so a reviewer can confirm the gear still matches the study and a future change can be traced through it.

Record the device identifier and type, the protective function, the pickup, the delay or band, what the device coordinates with above and below it, and whether the as-left setting was verified and injection-tested. Tie it back to the bus and the available fault current the curve was drawn against, so a change in fault current flags every device it affects.

Common mistakes

• Starting a coordination study without a current short-circuit study, so the curves are drawn against fault currents nobody calculated.
• Curves that touch or cross, usually at the high-current end near the available fault current, where both devices go to minimum time.
• Leaving an instantaneous trip on an upstream breaker that defeats the downstream device's coordination.
• Squeezing the coordination time interval too tight to fit devices in series, so the system coordinates on paper and races in the field.
• Coordinating the phase devices and skipping ground-fault coordination, so a downstream ground fault trips the service main.
• Stacking long delays for selectivity without a maintenance switch, so the arc-flash energy spikes at the upstream gear.
• Loading the settings but never injection-testing the device to confirm it trips where the setting says.
• Leaving a breaker or relay on factory defaults so the as-left settings never match the stamped study.
• Treating last year's study as valid after a transformer, source, or utility change moved the fault current.
• Coordinating the internal system and ignoring the utility's required upstream settings, so the first feeder fault drops the whole service.

Field checklist

Standards and references

The coordination study draws on IEEE protection practice. IEEE 242, the Buff Book, the recommended practice for protection and coordination of industrial and commercial power systems, is the foundational reference for how the study is performed and how devices are coordinated. IEEE 1015, the Blue Book, the recommended practice for applying low-voltage circuit breakers, covers the breaker side. The device identifiers come from IEEE C37.2, the standard for electrical power system device function numbers, which is where the 50, 51, 87, 27, 59, and the G and N ground suffixes are defined.

The NEC, NFPA 70, is where the mandate lives. Selective coordination is required for emergency systems in Article 700, legally required standby systems in Article 701, and critical operations power systems in Article 708, in their selective coordination sections commonly numbered 700.32, 701.32, and 708.54 in recent editions. The arc-energy reduction requirement at 240.87 mandates a means to cut clearing time on breakers rated or settable at 1200 A or higher, and ground-fault protection of equipment is required at 230.95 on solidly grounded wye services rated 1000 A or more over 150 volts to ground. The article and section numbers have shifted across code cycles, so confirm them against the edition the jurisdiction has adopted and any local amendments before citing them on a submittal.

The arc-flash study that shares the coordination study's model follows IEEE 1584 for arc-flash hazard calculations, with the worker-side rules in NFPA 70E. The field testing that verifies the settings follows the NETA Acceptance Testing Specifications, ANSI/NETA ATS, performed by an independent test agency. The engineer of record owns the design intent and the stamped settings, and the code requires the selective coordination selection to be made and documented by a licensed professional engineer or qualified person. The project documents and the adopted code editions control; the standards give the framework.

Units, terms, and acronyms

Coordination work carries a vocabulary that reads differently across a study report, a settings sheet, and the gear, and the same idea shows up under more than one name. The terms below are the ones that travel across the whole subject.

Selective coordination

Full selectivity across the whole fault range, so only the overcurrent device immediately upstream of a fault operates and every device above it holds

TCC (time-current curve)

The log-log plot of tripping time against fault current; the coordination study plots every device in a path on one chart

Coordination time interval (CTI)

The minimum time margin between two device curves, commonly about 0.3 to 0.4 seconds for relays, covering interrupting time, overtravel, and tolerance

LSIG

Long-time, short-time, instantaneous, and ground-fault functions of an electronic breaker trip unit, each with its own pickup and delay

Pickup and delay

The current at which a function starts to act and the time it waits; the two settings that fit a curve between the load below and the source above

ANSI device numbers

The IEEE C37.2 identifiers for relay functions: 50 instantaneous overcurrent, 51 inverse-time overcurrent, 87 differential, 27 undervoltage, 59 overvoltage, with G or N for ground

Differential (87)

Zone protection that compares current in against current out and trips fast for an internal fault without a coordination delay

Energy-reducing maintenance switch (NEC 240.87)

A means to put a breaker into no-intentional-delay mode during work so the arc clears fast, resolving the conflict between coordination delay and arc-flash energy

Secondary injection

Injecting simulated current into a relay or trip unit's inputs to confirm it picks up and times out at its setting, without energizing the primary system

Related tools

Calculators and readiness checks for this work

Compare your options

FAQ

People also ask

• How long does data center electrical commissioning take?
• What do I check when switchgear is delivered?
• What is selective coordination?
• How does a data center get power?
• What is busway?