What Selective Coordination Means on a Service Call

Why this matters

The complaint that sounds like a nightmare is often the most informative thing you will hear all week: "one little machine shorted and the whole floor went dark." Nothing is wrong with the floor. What that sentence describes is two protective devices in series that both operated on the same event, which is a property of the pair rather than a defect in either one. A tech who recognises it stops hunting for a second fault, and stops replacing devices that were never wrong.

Coordination is a property of a pair, at a stated current

Selective coordination means that for a fault at a given point, the device nearest the fault clears it before the device upstream of it starts to open, so the outage is limited to the smallest possible piece of the system. Three things in that sentence do the work, and dropping any of them is where the confusion starts.

It is a property of two devices together. No device is "selective." A 20 A breaker is not coordinated or uncoordinated; a specific 20 A breaker under a specific 100 A breaker is, or is not, and swapping either half changes the answer.

It is evaluated at a current, not in general. The same pair can coordinate at one fault current and fail at another. That is the single fact most techs are missing, and it is the reason the same building can behave both ways on two different days.

It is about clearing time, not about the amp ratings. A pair whose handle numbers look sensibly stepped can still both open, and a pair with a small step can coordinate perfectly, depending entirely on where their time-current characteristics sit relative to each other at the current that actually flowed.

Why curve overlay stops working above the upstream instantaneous pickup

In the low-multiple region - a modest overload, say a couple of times rating - the two devices are being timed in seconds and minutes and their curves are genuinely separable. Downstream opens, upstream never gets warm, the system coordinates.

Then you cross the upstream device's instantaneous pickup. On a standard molded-case thermal-magnetic breaker that pickup lands somewhere in the region of ten times rating, though the real multiple is a property of that product's published curve and not a constant, and adjustable devices move it deliberately. Above it, the upstream device is no longer waiting on a thermal element. Both devices are now clearing in a fraction of a cycle, both curve bands carry tolerance, and which one wins is not something you can read off a log-log plot.

For that region, coordination has to come from the manufacturer's published tested selectivity tables for that specific pair of devices, produced by testing rather than by drawing, or from a deliberate design choice such as a fuse pair or a short-time-delay upstream device with the instantaneous function turned off. Reading two curves and declaring them separated above the upstream instantaneous pickup is a common and confident mistake.

Available fault current is what puts you in one region or the other

The current that flows in a bolted fault is set by the source and by everything in series between the source and the fault: the supply transformer's impedance, the length and size of the service and feeder conductors, and then the branch conductors. Impedance adds as you go, so available fault current falls with distance from the source. A fault at the lugs of a panel and a fault at the far end of a long branch circuit off that same panel are two very different currents.

That is why the same pair of devices behaves differently on two calls, and it is why any statement about coordination has to name the point in the system it applies to. The available fault current at each panel is normally on the short-circuit study or the arc-flash labelling; NEC Article 110 in the edition your authority having jurisdiction has adopted requires service equipment at other than dwelling units to be field marked with the available fault current and the date it was calculated, which makes that label the first thing to read rather than a number to estimate.

Where coordination is required and where it is a preference

Most ordinary commercial and residential distribution carries no general code requirement to coordinate. Where the code does require it, it names the system:

• Emergency systems, legally required standby systems, and critical operations power systems carry selective coordination requirements in NEC Articles 700, 701 and 708 respectively, with exceptions and with the requirement stated over a defined time range, in the edition your authority having jurisdiction has adopted.
• Feeders supplying more than one elevator drive carry a selective coordination requirement in NEC Article 620 of the adopted edition.

Everywhere else it is an owner's decision about how much of a building goes down for a fault in one machine, made at design time. That distinction matters on a service call because it tells you who owns the answer: on a required-coordination system, changing a device is a design change that goes back to the engineer of record, not a parts decision.

The system as a series path, with two places the same branch can fault

      utility transformer
              |
        service main, 400 A
              |
     feeder breaker in subpanel, 100 A
              |
        branch breaker, 20 A
              |
      +-------+----------------------+
      |                              |
  junction near the panel long run to far end
  (fault A, high current)      (fault B, lower current)

Both faults are on the same branch circuit, protected by the same pair of devices. The only thing that differs is how much conductor sits between the source and the fault, and that is enough to change which devices open.

The case: same pair, two events, opposite outcomes

A small commercial building. The complaint is that a failure in one packaged unit took out the entire subpanel, and the customer wants the feeder breaker replaced because "it is too sensitive."

Event, six months earlier. A cord failure at the far end of a long branch run. The 20 A branch breaker opened. The 100 A feeder breaker held. Nobody thought about it, because that is what everyone expects to happen.

Event, now. A short at a junction box a few steps from the subpanel, on the same branch circuit. The 20 A branch breaker opened and so did the 100 A feeder breaker, and everything else on that subpanel went out with it.

The first hypothesis on the ticket - the feeder device is weak or oversensitive - has evidence pointing the other way, and it is worth saying why. A weak thermal element would show up as trips under normal load over minutes, not as an instantaneous operation during a fault. Nothing about this system trips in normal service. The device has operated exactly once, during a genuine short circuit, which is the duty it exists for.

Now put the two events against the general rules above.

• The short-circuit study on the drawings gives roughly 8,500 A available at the subpanel bus. That is a figure read off the study, not one to estimate on site.
• The 20 A branch device has an instantaneous pickup in the region of ten times rating, so around 200 A. The 100 A feeder device, same family, around 1,000 A. Both of those are order-of-magnitude figures from the product curves, and the curves are the authority.
• Fault A, at the junction near the panel: conductor impedance between bus and fault is short, so the current is close to the bus figure. At roughly 8,500 A, the branch device sees about 42 times its instantaneous pickup and the feeder device sees about 8 times its own. Both are deep in their instantaneous regions, both clear in a fraction of a cycle, and which opens first is a race the curves do not decide. Both opened. That is the expected behaviour of an uncoordinated pair above the upstream instantaneous pickup, not a defect.
• Fault B, at the far end of the long run: the branch conductors add enough impedance to bring the fault current down by roughly an order of magnitude, into the low hundreds of amps. That is comfortably above the branch device's roughly 200 A instantaneous pickup and well below the feeder device's roughly 1,000 A. Only the branch device could operate, and only it did.

Same pair, same building, opposite outcomes, and the variable is fault location. Nothing about the hardware changed between the two events.

What this makes the correct answer. The system is not coordinated at the subpanel bus for high-current faults, and it never was. If the owner needs it to be - and on this system no code article requires it - that is an engineered change: a tested selective pair from the manufacturer's tables, or an upstream device with a short-time-delay function and the instantaneous turned off, sized against the withstand rating of everything it now protects for longer. Both of those are design decisions with fault-current and equipment-withstand implications, and neither is a truck-stock swap.

The failure mode if the tech takes the ticket at face value: the feeder device gets replaced with a same-amperage device from a different family. The next high-current fault behaves the same way or worse, the outage repeats, and the shop has now also put an untested combination in a panel where the original pair at least had a known behaviour.

What a service tech actually owes here

• Do not change a device's frame, family, trip type or instantaneous setting on a system where coordination was engineered. On an emergency, legally required standby, critical operations or multiple-elevator feeder, that change goes back to the engineer of record because the coordination is a code-required property of the design under the adopted NEC articles above.
• Record which devices opened, in order if anyone saw it, and where the fault was. That pair of facts is what makes a later study possible and it is free while you are standing there.
• Read the available fault current label before quoting anything. It is the number that decides which region you are in.
• Say the honest thing to the customer: the extra outage is a design property they may or may not want to pay to change, not a broken part. Anyone who sells them a breaker to fix it is selling a part that will not change the outcome.

How to verify you got this right

• Check that your explanation predicts both events, not just the one on the ticket. If your story explains why both devices opened this time but implies they should also have both opened six months ago, it is wrong. A coordination explanation has to survive the case where the system behaved.
• Confirm the operated devices were in series, not in parallel branches. If two devices on unrelated circuits opened, this is not a coordination question at all and you are looking at a common-mode event upstream.
• Look for a heat signature or damage on the upstream device. A device that operated instantaneously on a genuine fault has done its duty; one showing contact damage, discolouration or a scorched enclosure has taken an interrupting event worth inspecting, which is the subject of the article on interrupting ratings.

References

• NEC Articles 110, 240, 620, 700, 701 and 708, in the edition your authority having jurisdiction has adopted
• Manufacturer-published time-current curves and tested selective coordination tables for the specific device pair installed
• The project short-circuit study and the field-applied available fault current marking on the service equipment
• See related: What an Interrupting Rating Is and Why It Is Not a Trip Setting; Why a Breaker That Trips Is Usually Not the Fault