A 1,000 kVA transformer at 480 volts and 5.75 percent impedance puts 20,918 amps on the bus, which is 5.2 percent under a 22 kA rating. The nameplate tolerance on that impedance is 7.5 percent. The band is wider than the margin, so the same unit can arrive at 22,615 amps and the label no longer covers it.
That gap is the whole article. Everything else follows from it.
You have probably seen the rule stated as a fact. A 1,000 kVA transformer at 480 volts needs 22 kA equipment, because 1,203 amps divided by 0.0575 is 20,918, and 22,000 is the next rating on the shelf. The arithmetic is correct. The conclusion is a rounding decision that has been repeated often enough to look like a code requirement, and it fails on a machine that is completely within specification.
Three things are true at once here, and they get mixed together in most of what is written on this subject. The fault current calculation is simple and almost nobody gets it wrong. The equipment rating system is a catalogue, not a rule. And the impedance on the nameplate is a guaranteed value with a tolerance, which means the number you calculate is one point inside a range you did not choose. The first two are widely understood. The third is where money gets lost.
The short version in five numbers
If you read nothing else, read this table. Five numbers decide whether a 22 kA rating holds on a 1,000 kVA unit at 480 volts, and only the first two are on the sales sheet.
| Question | Answer | Where it comes from |
|---|---|---|
| Full load current at 480 volts | 1,202.8 A | 1,000,000 divided by 1.732 times 480 |
| Fault current at nameplate impedance | 20,918 A | 1,202.8 divided by 0.0575 |
| Margin under a 22 kA rating | 5.17 percent | 22,000 divided by 20,918 |
| Allowed impedance tolerance | plus or minus 7.5 percent | IEEE C57.12.00, two winding units |
| Fault current at the low limit | 22,615 A | 1,202.8 divided by 0.053188 |
The third number and the fourth number are the ones to compare, and they are not the same size. That is the entire problem in one line.
Two of these numbers are stable. Full load current is fixed by kVA and voltage, and nothing on the production line can move it. Fault current at nameplate impedance is fixed by the first number and the nameplate. The other three are where the argument lives: the margin is arithmetic, the tolerance is a clause, and the fault current at the low limit is what the equipment actually sees if the unit comes in at the bottom of its allowed band.
Where 22 kA comes from, and where it does not
Twenty two thousand amps is not in the National Electrical Code. It is not in any standard as a threshold. It is a rung on a ladder that manufacturers build to, and the ladder is short.
Molded case circuit breakers are listed in a series that runs 10, 14, 18, 22, 25, 30, 35, 42, 50, 65, 100, 150 and 200 kA of interrupting rating. Assemblies are coarser. Switchboards and panelboards commonly come in 10, 22, 42, 65 and 100 kA. Notice that 22 appears on both lists and 25 appears on only one. That is not an accident of engineering. It is a reflection of what the market buys.
The rating that a distributor stocks at 480 volts is 22 kA, because it covers the largest single transformer that most commercial services use without jumping to the next price tier. If the ladder ran 20, 24 and 28, nobody would talk about a 22 kA rule. They would talk about a 20 kA rule, and the number would not look like a threshold. It would look like what it is: the nearest rung above the calculation.
What the code actually says
Three sections matter, and none of them names a number.
NEC 110.9 requires that equipment intended to interrupt current at fault levels have an interrupting rating not less than the current available at its line terminals. That is an obligation, not a value. It tells you the device must be at least as good as the fault, and leaves the fault to you.
NEC 110.10 goes further and asks for the whole circuit to be considered: the overcurrent devices, the total impedance, the equipment short circuit current ratings and the other characteristics of the circuit, selected and coordinated so that a fault is cleared without extensive damage. That sentence is why the tolerance matters. Total impedance is not the transformer’s nameplate number. It is the number the unit actually has.
NEC 110.24 requires service equipment, in other than dwelling units, to be field marked with the maximum available fault current and the date the calculation was performed. That is the section that puts a number on a sticker and makes it permanent. Everything downstream gets sized against that sticker.
So the code gives you a duty and a label. The number 22,000 comes from a catalogue. That distinction sounds academic until you are the one signing the short circuit study, at which point it becomes the difference between “I followed the rule” and “I calculated the fault current and selected equipment above it.”
Why the habit survives anyway
Because it usually works. On a 750 kVA unit at 5.75 percent the fault current is 15,689 amps, which is 29 percent under 22 kA at nominal and 23 percent under at the low impedance limit. On a 500 kVA unit it is 10,459, which is more than a factor of two of margin. The habit is right for most of the range and wrong at exactly one place, and that place is the size most people are buying.
That is the shape of every bad rule of thumb. It is calibrated on the easy cases and applied to the hard one.
The 5.17 percent that decides the question
Here is the calculation behind the third row of the table, because it is the one people skip.
22,000 / 20,918 = 1.0517The 22 kA rating is 5.17 percent above the calculated fault current. That is the entire margin. It is not a safety factor somebody chose. It is the leftover distance between a number you computed and a number that was on a shelf.
Now the other side. Percent impedance is a guaranteed value with a manufacturing tolerance, and for two winding transformers the standard figure is plus or minus 7.5 percent of the guaranteed value. A unit ordered at 5.75 percent is allowed to test anywhere from 5.31875 to 6.18125 percent.
5.75 x 0.925 = 5.31875
1,202.8 / 0.053188 = 22,615 ATwenty two thousand six hundred and fifteen amps against a 22,000 amp rating. The unit is inside its specification, the calculation was correct, the equipment was selected by the rule, and the equipment is undersized. Nothing in that chain is a mistake except the assumption that the nameplate is the number.
The tolerance is bigger in amps than in percent
There is a detail here that turns a small gap into a large one, and it comes from the fact that impedance sits in the denominator.
A 7.5 percent reduction in impedance does not produce a 7.5 percent increase in current. It produces an 8.1 percent increase, because the reciprocal of 0.925 is 1.0811.
1 / 0.925 = 1.0811So the comparison is not 7.5 against 5.17. It is 8.11 against 5.17, and the tolerance is 1.57 times the margin rather than 1.45 times. The shortfall is 2.94 percentage points of current, which on this unit is 615 amps.
Six hundred and fifteen amps is not a rounding error. It is larger than the full load current of a 500 kVA transformer secondary at 480 volts. It is the difference between a device that has been tested at the current it will see and a device that has been tested at a current 2.8 percent lower.
What happens above the rating
A breaker marked 22 kA has been tested to interrupt 22,000 amps symmetrical RMS at its marked voltage, under the conditions of its listing. Above that number, the listing says nothing. It does not say the breaker explodes. It says the manufacturer has not demonstrated that it will clear.
In practice a 22 kA breaker asked to clear 22,615 amps will often still clear, because the rating has margin built into the test and the actual fault is rarely a bolted three phase short at the terminals. But “often” is doing a lot of work in that sentence, and the failure mode when it does not clear is not a tripped breaker. It is a device that vents plasma into a panel that someone is standing in front of.
This is why 110.9 is written as an absolute. The comparison is not “probably fine.” It is “not less than.”
Two units, same nameplate, 615 amps apart
Picture two transformers on the same purchase order, from the same factory, built to the same drawing.
Both nameplates read 1,000 kVA, 480 V delta, 5.75 percent impedance. Both pass the factory impedance test. Both ship. One measures 5.72 percent and the other measures 5.32 percent, and both of those numbers are legal under the same clause.
The first one puts 21,028 amps on the bus. The second one puts 22,609. The difference is 1,581 amps, or 7.5 percent, and there is nothing on either nameplate that tells you which one you got. The nameplate says 5.75 on both, because that is the guaranteed value, not the measured one.
Now put a 22 kA switchboard on each. The first installation is correct by a comfortable margin that nobody will ever question. The second is 2.7 percent over the rating, and it will be over for the life of the building, every time there is a fault, and nobody will know until the fault happens.
That is the practical form of the problem. It is not that the rule is always wrong. It is that the rule produces a correct answer and an incorrect answer from identical paperwork, and the paperwork does not distinguish them.
The one document that would settle it
The factory impedance test report. Every unit gets one. It states the measured percent impedance at rated current and rated voltage, on a specific tap, at a specific temperature.
If that number is on your desk before the switchgear is released, the question is answered. If it is not, you are selecting equipment against a guaranteed value that has a 15 percent wide window, and you have chosen to assume the middle of the window.
Most projects release the gear before the transformer is built, which is why the assumption gets made. That is a scheduling fact, not an engineering one, and it is worth knowing which of the two you are doing.
973 and 1,015, and why 1,000 kVA is the flashpoint
The interesting thing about this problem is that it is almost perfectly localized. It is severe at one standard size and mild everywhere else, and the reason is arithmetic rather than anything about the transformer.
Ask a different question: at 480 volts, what kVA produces exactly 22,000 amps of fault current?
22,000 x 1.732 x 480 x 0.0575 / 1,000 = 1,051.7 kVAAt nameplate impedance, a 1,051.7 kVA unit lands exactly on 22 kA. So a 1,000 kVA unit has 5 percent of headroom, which is what the first table said.
Now apply the tolerance, because the unit is allowed to be at the low limit.
22,000 x 1.732 x 480 x 0.053188 / 1,000 = 972.8 kVAAt the low impedance limit, the largest unit that stays under 22 kA is 972.8 kVA. Not 1,051.7. Nine hundred seventy three.
A 1,000 kVA transformer sits between those two numbers. It is above 973 and below 1,052, which means it passes the check at nameplate impedance and fails it at the low limit. No other standard size does that. The 750 does not, because 750 is below 973 by a wide margin. The 1,500 does not, because 1,500 is above 1,052 by a wide margin and needs 35 kA regardless.
That is the answer to a question that comes up constantly in procurement meetings: why does everyone argue about 1,000 kVA? Because 1,000 kVA is the only standard size where the tolerance decides the outcome. On every other size the answer is the same whether you use the nameplate or the low limit.
What it takes to make 22 kA hold
If you want the rating to hold at the low impedance limit, the nameplate impedance has to be high enough that the low limit is still above 5.4673 percent, which is the impedance that produces exactly 22,000 amps on this unit.
1,202.8 / 22,000 = 0.054673
0.054673 / 0.925 = 0.059106Five point nine one percent. Round it to what a factory will actually quote and you are ordering 6.0 percent instead of 5.75.
At 6.0 percent the numbers become 20,047 amps at nameplate and 21,672 at the low limit. The rating holds, with 1.5 percent of margin. That is thin, but it is a real margin instead of a negative one.
| Nameplate impedance | kVA at 22 kA, nameplate | kVA at 22 kA, low limit |
|---|---|---|
| 5.75 percent | 1,051.7 | 972.8 |
| 6.0 percent | 1,097.4 | 1,015.1 |
| 6.5 percent | 1,188.8 | 1,099.6 |
The third row is worth a look. At 6.5 percent, the low limit still allows 1,099.6 kVA, so a 1,000 kVA unit is 100 kVA clear of the line rather than 15. That extra quarter point of impedance buys a margin that survives a motor contribution, which is the subject of the next section.
The four corrections, ranked by size
The fault current at a transformer’s secondary terminals is a starting point, not an answer. Four things move it, and they are not equally knowable at design time. This is the part that most short circuit studies handle correctly and most rules of thumb ignore completely.
| Correction | Direction | Size on this unit | Do you know it when you release the gear |
|---|---|---|---|
| Impedance tolerance | adds | 1,697 A, 8.11 percent | You know the clause, not the unit |
| Source impedance | subtracts | 1,360 A at 250 MVA, 6.50 percent | Usually not |
| 100 ft of three sets of 500 kcmil | subtracts | 2,586 A, 12.37 percent | Yes, from the one line |
| Motor contribution | adds | 1,268 A for 317 A of motor FLC | Yes, from the motor list |
Add them up in the wrong order and you get a number that has nothing to do with the installation. Add them up in the right order and you get a number that straddles 22 kA depending on which panel you are standing at.
Source impedance, the one nobody knows
The infinite bus method assumes the utility has zero impedance, so the transformer’s own impedance is the only thing limiting the fault. That is the standard first pass and it is described everywhere as conservative.
Take a 250 MVA source, which is a stiff urban distribution bus. On a 1 MVA transformer base its impedance is 0.004 per unit.
1 / 250 = 0.004 pu
1,202.8 / (0.0575 + 0.004) = 19,558 AThe source takes 1,360 amps off the number, which is 6.5 percent. So the infinite bus assumption overstates the fault current, which is the definition of conservative, and everyone is happy.
Except that the same study used 5.75 percent, and the unit is allowed to be at 5.32. Combined with the 250 MVA source:
1,202.8 / (0.053188 + 0.004) = 21,032 ATwenty one thousand and thirty two. Still under 22,000, so in this particular case the two errors nearly cancel and the equipment is fine. That is luck, not method.
The infinite bus method is not conservative
Here is the uncomfortable part. The method’s two assumptions push in opposite directions, and they are not the same size.
The source assumption is conservative by 6.5 percent at 250 MVA. The impedance assumption is optimistic by 8.11 percent. The optimism is larger than the conservatism, so the “worst case” method is not a worst case. It is a slightly optimistic case with a conservative reputation.
Make the bus stiffer and it gets worse. At 1,000 MVA:
1 / 1,000 = 0.001 pu
1,202.8 / (0.053188 + 0.001) = 22,197 AThe infinite bus method with nameplate impedance gives 20,918. The true worst case at a 1,000 MVA bus with a low limit transformer is 22,197. The method understates the fault current by 1,279 amps, or 6.1 percent, on a unit whose entire margin was 5.17 percent.
That is the finding worth carrying away. Not that the method is wrong, but that its error budget is smaller than the tolerance it ignores, so it cannot be relied on to be conservative for a single unit. For a system study with many sources and many branches, the errors wash out. For one transformer and one switchboard, they do not.
Conductors, the one that works
Feeder impedance is the correction that reliably helps, and it is the one you control.
A 1,000 kVA unit at 480 volts needs about 1,203 amps, which is three sets of 500 kcmil copper. For 100 feet in PVC, using the resistance and reactance from Chapter 9 Table 9, the per unit impedance works out to 0.00811 on the transformer base.
0.2304 ohm base = 480 squared / 1,000,000
Zc = 0.00811 pu for 100 ft of three sets of 500 kcmil
1,202.8 / (0.0575 + 0.00811) = 18,332 ATwelve point four percent off, which is larger than the 8.1 percent the tolerance adds. A 22 kA switchboard 100 feet from the transformer is not marginal. It is comfortable.
This is why the answer to “is 22 kA enough” is a question about geometry rather than about the transformer. At the secondary terminals, a low limit unit is over the rating. At the first panel down the feeder, it is 17 percent under. Both statements are about the same transformer and the same rating.
| Distance from the transformer | Fault current | Against 22 kA |
|---|---|---|
| At the terminals | 22,615 A | 2.8 percent over |
| 25 ft | 21,783 A | 1.0 percent under |
| 50 ft | 21,012 A | 4.5 percent under |
| 100 ft | 19,621 A | 10.8 percent under |
| 200 ft | 17,328 A | 21.2 percent under |
That table uses the low impedance limit throughout, so it is the pessimistic version of a geometry question. Fifty feet is where the rating starts to hold, and fifty feet is a very short run. It is also a very common run, which is why real installations with 22 kA gear usually do not fail. They are saved by a number nobody put in the calculation.
Motors, the one that always adds
Motor contribution is the correction that people remember and then forget to apply.
When a fault occurs, motors that are spinning keep generating into it until their field collapses. For low voltage induction motors the usual first cycle estimate is four times the motor full load current, added directly to the transformer’s contribution.
NEC Table 430.250 gives 240 amps for a 200 hp 480 volt motor and 77 amps for a 60 hp. A plant with one of each, both running, contributes 1,268 amps.
(240 + 77) x 4 = 1,268 A
20,918 + 1,268 = 22,186 ATwenty two thousand one hundred and eighty six, on a transformer at nameplate impedance with an infinite bus. Over the rating by 0.8 percent, with no tolerance applied at all.
The threshold is worth remembering. The gap between 20,918 and 22,000 is 1,082 amps, which is 270 amps of motor full load current at the four times factor. Two hundred and seventy amps is one 200 hp motor and a bit. On any plant with real process load, that is not a large motor list.
| Motor load at 480 V | Full load current | Contribution at 4x | Fault current | Against 22 kA |
|---|---|---|---|---|
| None | 0 A | 0 A | 20,918 A | 5.2 percent under |
| One 150 hp | 180 A | 720 A | 21,638 A | 1.7 percent under |
| One 200 hp plus one 60 hp | 317 A | 1,268 A | 22,186 A | 0.8 percent over |
| One 200 hp plus one 150 hp | 420 A | 1,680 A | 22,598 A | 2.7 percent over |
So the fault current path that ends at a switchboard feeding a plant with two large motors is already over 22 kA before the impedance tolerance is considered. The rule does not merely fail at the edge of the specification. It fails in the middle of the specification.
Tolerance, the one already in the box
The fourth correction is the one that requires no assumption at all, because it is a property of the thing you are buying.
It is also the only one that is not visible on any drawing. Source impedance is a utility question, conductors are on the one line, motors are on the equipment schedule. The tolerance is in a standard, and it applies whether or not anyone read it.
Put all four together and the result depends on which of them you include:
| Scenario | Impedance | Source | Motors | Fault current | Against 22 kA |
|---|---|---|---|---|---|
| Sales sheet | 5.75 percent | Infinite | None | 20,918 A | 5.2 percent under |
| Low limit, terminals | 5.319 percent | Infinite | None | 22,615 A | 2.8 percent over |
| Low limit, 250 MVA source | 5.319 percent | 250 MVA | None | 21,032 A | 4.4 percent under |
| Low limit, 250 MVA, motors | 5.319 percent | 250 MVA | 317 A FLC | 22,300 A | 1.4 percent over |
| Low limit, 250 MVA, 100 ft, motors | 5.319 percent | 250 MVA | 317 A FLC | 19,688 A | 10.5 percent under |
Five rows, one transformer, one rating. Three of them pass and two of them fail. The row you are in depends on where the switchboard sits and what is running when the fault happens.
That is not a computation problem. It is a specification problem. The answer does not exist until you decide which location and which operating state you are designing for, and the 22 kA rule is a way of not deciding.
The 232 MVA crossover
There is one number that tells you which correction wins, and it is worth knowing even though almost nobody can supply it at design time.
The tolerance adds 8.11 percent. Source impedance subtracts a percentage that depends on how stiff the bus is. They exactly cancel at one source strength.
Zs such that 0.053188 + Zs = 0.0575
Zs = 0.004312 pu
1 / 0.004312 = 232 MVATwo hundred and thirty two megavolt amperes. If the utility bus behind your transformer is stiffer than that, the tolerance wins and the fault current is above what you calculated at nameplate. If it is softer, the source wins and you are below.
A 250 MVA bus is marginally stiffer, so on a low limit unit you get 21,032 rather than 20,918. A 100 MVA bus is much softer and you get 17,819 at nameplate impedance, which is why rural services with long primary feeders almost never have a fault current problem.
The point of the number is not that you should calculate it. The point is that its value is unknown, it is on the same order as the tolerance, and it flips the answer. A method that assumes it is infinite and a method that assumes it is 250 MVA differ by more than the margin, which means the “conservative” assumption and the “realistic” assumption produce different equipment classes.
If your utility will tell you the available fault current at the service point, that number replaces the assumption and the problem collapses to arithmetic. Many will, on request, for a specific address. It is worth the phone call before the switchgear is released.
The second cliff, at 42 kA
The strongest evidence that 22 kA is a habit rather than a rule is that the same habit fails again one rung up, in exactly the same way.
A 2,000 kVA unit at 480 volts and 5.75 percent produces 41,837 amps. The next standard assembly rating is 42 kA. The margin is 163 amps, which is 0.39 percent.
2,405.6 / 0.0575 = 41,837 A
42,000 / 41,837 = 1.0039One hundred and sixty three amps. That is less than the difference between one transformer and the next one off the line. At the low impedance limit the same unit produces 45,229 amps, which is 7.7 percent over.
| kVA | Impedance | At nameplate | At low limit | 42 kA verdict |
|---|---|---|---|---|
| 1,500 | 6.0 percent | 30,070 A | 32,508 A | 23 percent under, holds |
| 2,000 | 5.75 percent | 41,837 A | 45,229 A | 0.4 percent under, fails at the limit |
| 2,000 | 6.5 percent | 37,010 A | 40,010 A | 4.7 percent under, holds |
| 2,500 | 6.0 percent | 50,117 A | 54,181 A | 19 percent over, needs 65 kA |
To make 42 kA hold at the low limit on a 2,000 kVA unit you need a nameplate impedance of 6.19 percent, which is not a stock value on most dry type lines and is a special order on many liquid filled ones. So the practical answer at 2,000 kVA is 65 kA, and the practical answer at 1,000 kVA is 25 or 35 kA once motors are on the bus.
That is the real shape of the problem. The rule works at the sizes where the margin is large and fails at the sizes where the margin is thin, and the sizes where the margin is thin are the sizes people actually buy. There is no size where the rule is both useful and safe.
Peak is not RMS, and the difference has its own arithmetic
Everything above is symmetrical RMS current. That is the number in the rating, the number in the calculation and the number on the label. It is also not the number the bus feels.
A fault has a DC offset for the first few cycles. The offset decays with a time constant set by the ratio of reactance to resistance, and it adds to the AC peak during the first half cycle. The peak current is:
I_peak = I_rms x 1.414 x (1 + e^(-pi / (X/R)))At X/R of 6, the term in brackets is 1.592, so the peak is 2.25 times the RMS. At X/R of 12 it is 2.50, and at X/R of 20 it is 2.62.
| X/R | Peak multiplier on RMS | Peak on 20,918 A |
|---|---|---|
| 4 | 2.06 | 43,071 A |
| 6 | 2.25 | 47,108 A |
| 10 | 2.45 | 51,191 A |
| 12 | 2.50 | 52,352 A |
| 20 | 2.62 | 54,866 A |
A 22 kA assembly is tested on a circuit with a defined X/R, and the low voltage test basis corresponds to a power factor of about 15 percent, which is an X/R near 6.6 and a peak of about 2.3 times the RMS. So the momentary or closing rating of a 22 kA assembly is roughly 50,600 amps peak, and that is the number the bus bars and the breaker mechanism actually have to survive.
Look at what that means for the 1,000 kVA unit at nameplate impedance. At X/R of 6 the peak is 47,108, which is inside 50,600. At X/R of 10 it is 51,191, which is outside, and the RMS is still 5 percent under the rating. So the equipment passes the check everyone performs and fails one that almost nobody performs.
Where this bites hardest
Higher kVA means higher X/R, because the winding has more inductance relative to its resistance. A 1,000 kVA dry type sits around 4 to 6. A 2,000 kVA liquid filled unit can be 12 or higher.
Take the 2,000 kVA case from the previous section. At nameplate impedance the RMS is 41,837, which is under the 42 kA rating by 0.39 percent. At X/R of 12 the peak is 2.50 times that, or 104,706 amps. The momentary rating that goes with a 42 kA assembly is 42,000 times 2.3, or 96,600.
41,837 x 2.5027 = 104,706 A peak
42,000 x 2.3 = 96,600 A peak ratingThe RMS check passes. The peak check fails by 8.4 percent. Same transformer, same rating, two different answers depending on which number you compare.
This is why a short circuit study asks for X/R and a rule of thumb does not. The study is not being pedantic. It is checking a second rating that exists for a reason, and the reason is that bus bars bend and breaker contacts weld under the peak, not under the RMS.
Series rating, and why it is not the escape hatch
The usual answer to all of this is series rating. Put a current limiting fuse or a current limiting breaker upstream, and the combination is rated for a higher available fault current than the downstream device could handle alone. A 22 kA panel behind the right fuse can be applied on a 100 kA service. It is a real, listed, code recognized solution, and it is used constantly.
It has three limits that matter here.
It is a property of a tested pair, not of a device. The combination has to appear in the manufacturer’s tabulation for that specific breaker and that specific fuse, at that voltage. You cannot calculate it, you cannot mix brands, and you cannot assume that because one 22 kA breaker series rates with one fuse, the next model does too.
The code puts conditions on it. Series ratings are permitted only for tested combinations, and the allowance is restricted where motor load is significant, because the fuse has to clear before the downstream breaker sees the peak. The restriction is expressed as a percentage of the lower device’s interrupting rating.
1 percent of 22,000 = 220 A of motor full load currentTwo hundred and twenty amps. That is one 200 hp motor at 480 volts, roughly, and nothing else. A plant with two large motors and a couple of pumps has already used up the allowance, which means the series rating is not available on the very installation where the fault current is highest.
And it does nothing for the peak. A series rated combination is rated for the RMS it can interrupt. The momentary withstand of the assembly is still its own number, and the upstream fuse does not make the bus bars thicker.
So series rating is a legitimate way to keep a 22 kA panel on a stiff service. It is not a way to make the tolerance question go away, and it is not available once there is real motor load. On a plant with process equipment, the answer is usually to buy the higher rating.
The same transformer on 208 volts
One more thing that the rule hides. The 22 kA number is entirely voltage specific, and it is easy to carry it to a voltage where it does not apply.
The fault current formula divides by voltage twice, once in the full load current and once more through the impedance base. So the same kVA at a lower secondary voltage produces proportionally more fault current, and the relationship is the inverse of the voltage ratio.
| Secondary | Full load current | Fault at 5.75 percent | Fault at low limit | Equipment class |
|---|---|---|---|---|
| 600 V | 962.3 A | 16,735 A | 18,092 A | 22 kA, 18 percent margin |
| 480 V | 1,202.8 A | 20,918 A | 22,615 A | 22 kA, on the line |
| 240 V | 2,405.6 A | 41,837 A | 45,229 A | 42 kA at nameplate, 65 kA at the limit |
| 208 V | 2,775.7 A | 48,273 A | 52,187 A | 65 kA |
Same transformer, same impedance, same kVA. Four secondary voltages, three equipment classes, and a decision that goes from comfortable to impossible without anything on the transformer changing.
The 208 volt row is the one that catches people, because 208Y/120 is a common commercial secondary and 1,000 kVA is a common size for a building. A 1,000 kVA unit at 208 volts produces more than twice the fault current of the same unit at 480, which is why large 208 volt services are expensive in a way that is not obvious from the transformer price.
If you take one thing from this table, take the direction. Going down in voltage raises the fault current by the ratio of the voltages, and it raises it faster than the kVA falls when you pick a smaller unit. That is why the 22 kA rule, which is calibrated at 480, is wrong at 208 in the unsafe direction.
What goes on the purchase order
The output of all of this is not a calculation. It is a set of sentences that go into the specification, and each one closes a hole that the rule leaves open.
| Write this | Because | What it costs |
|---|---|---|
| Impedance 6.0 percent minimum at the rated tap, guaranteed | moves the low limit above the 22 kA line | about 0.1 points of voltage regulation |
| Measured impedance from the factory test report, issued before shipment | the nameplate is a guarantee, the test is a fact | nothing |
| X/R on the same report | the peak is a separate rating | nothing |
| Motor list and the connected kVA to the short circuit study | 317 A of motor FLC is 1,268 A of fault | nothing |
| Available fault current from the utility, at the service point, in writing | replaces the infinite bus assumption | a phone call |
The first row is the one that changes the hardware. Going from 5.75 to 6.0 percent impedance is a small change to the design and a small change to the price, and it moves the low limit from 5.32 to 5.55 percent, which moves the fault current from 22,615 to 21,672.
What that quarter point costs you
Higher impedance is not free. It shows up in voltage regulation, which is the change in secondary voltage between no load and full load.
For a lagging power factor, regulation is approximately the resistance times the power factor plus the reactance times the sine of the angle, expressed as a percentage. On a 1,000 kVA unit with about 1.2 percent resistance, going from 5.75 to 6.5 percent impedance raises the reactance from 5.62 to 6.39 percent and moves full load regulation at 0.9 power factor from roughly 3.5 to 3.9 percent.
| Impedance | Reactance | Regulation at 0.9 pf | Fault current at low limit |
|---|---|---|---|
| 5.75 percent | 5.62 percent | 3.5 percent | 22,615 A |
| 6.0 percent | 5.88 percent | 3.7 percent | 21,672 A |
| 6.5 percent | 6.39 percent | 3.9 percent | 20,001 A |
| 7.5 percent | 7.40 percent | 4.3 percent | 17,338 A |
Four tenths of a point of regulation, in exchange for 11.6 percent less fault current. That is a good trade on almost any project, and it is the reason a 6.5 percent unit is often the right answer at 1,000 kVA even though 6.0 is enough on paper.
The trade gets bad above about 7.5 percent, where the regulation starts to interfere with motor starting and with the voltage window that the utilization equipment expects. That is the upper bound, and it is set by the load, not by the fault current.
What to write on the label
NEC 110.24 requires the service equipment to be marked with the maximum available fault current and the date of the calculation. That sticker is the number every future electrician, engineer and inspector will use. It outlives the project, the contractor and often the building owner.
So which number goes on it?
The temptation is 20,918, because that is what the calculation produced and it is a nice round story. The problem is that the sticker is a promise about the worst case, and 20,918 is the middle of the range, not the top.
Write the low limit figure, and write the assumption next to it. Something like 22,600 amps available, based on minimum guaranteed impedance at the low tolerance limit, infinite bus, no motor contribution. That is a longer sticker and a more honest one, and it tells the next person exactly what they are working with.
The alternative, a sticker that says 20,900 when the unit can deliver 22,615, is worse than no sticker, because it carries the authority of a field marking while being 8 percent low. Every panel added downstream of it gets sized against a number that the transformer is allowed to exceed.
And update it when the utility changes
110.24 requires the marking to be updated when the available fault current changes. The most common cause is a utility transformer replacement. A utility that swaps a 500 kVA transformer for a 1,000 kVA one, or upgrades a feeder, can raise the available fault current at your service without touching anything inside your building.
The sticker is the only place that change becomes visible. If it is not updated, the building keeps its old equipment ratings and its old assumptions, and the fault current quietly goes up.
Three questions that settle it
You do not need a study to know whether 22 kA is right on a 1,000 kVA unit. You need three answers.
Where is the first overcurrent device? If it is at the transformer secondary terminals, you are in the worst of the five scenarios. If it is 100 feet down a feeder, you are in the best one. Most installations are somewhere in between, and the distance is on the one line.
Is there more than 270 amps of connected motor full load current on the bus? That is the threshold at which motor contribution alone crosses 22 kA at nameplate impedance. If the answer is yes, the tolerance question is already moot, because you are over before it applies.
Can you get the factory impedance test before the switchgear is released? If yes, use the measured value and the whole tolerance discussion disappears. If no, you are assuming the middle of a 15 percent window and you should assume the bottom instead.
Three yes or no answers, and the equipment class follows. That is a shorter conversation than a short circuit study and it gets you to the same place for a single transformer.
The rule is a rounding, and the rounding is the risk
Twenty two thousand amps is a real rating on real equipment. It is what a distributor stocks, what a switchboard costs, and what a specification usually says. None of that is wrong.
What is wrong is treating it as a threshold that a calculation either meets or does not, when the calculation has a 15 percent wide input and the threshold has a 5 percent margin. The comparison is being made between two numbers of different quality, and the number being treated as exact is the one that is not.
The fix does not require a study. It requires four sentences in a specification and one phone call, and each one closes a hole that a rule of thumb cannot see.
FAQ
What is the 22 kA rule for transformers?
It is a rounding habit, not a code requirement. On a 1,000 kVA transformer at 480 volts and 5.75 percent impedance, the fault current is 20,918 amps, and the next standard equipment rating above that is 22 kA. The rule says to buy 22 kA gear. The arithmetic is right and the margin is 5.17 percent, which is smaller than the 7.5 percent impedance tolerance the unit is allowed.
Is 22 kA enough for a 1,000 kVA transformer at 480 volts?
At nameplate impedance and at the transformer terminals, yes by 5.2 percent. At the low impedance limit, no. A unit that tests at 5.32 percent impedance produces 22,615 amps, which is 2.8 percent above the rating. In practice the answer also depends on distance to the first panel and on motor load, so 25 or 35 kA is the safer purchase at this size.
How do I calculate fault current from transformer impedance?
Divide the full load current by the per unit impedance. For three phase, full load current is kVA times 1,000 divided by 1.732 times line to line voltage. For a 1,000 kVA unit at 480 volts, that is 1,202.8 amps, and dividing by 0.0575 gives 20,918 amps. This is the infinite bus value and it ignores source impedance, conductor impedance and motor contribution.
What is the impedance tolerance on a transformer?
Plus or minus 7.5 percent of the guaranteed value for two winding transformers, and plus or minus 10 percent for three winding units and autotransformers. A unit ordered at 5.75 percent may measure anywhere from 5.31875 to 6.18125 percent and still be within specification. IEC units use a different band, typically 10 percent below 10 percent impedance.
Why does a lower impedance cause higher fault current?
Because impedance is the only thing between the source and the fault, so it sits in the denominator of the calculation. Halving the impedance doubles the fault current. The relationship is not linear in the other direction either, because the reciprocal of 0.925 is 1.0811, so a 7.5 percent drop in impedance produces an 8.1 percent rise in current.
Is the infinite bus method conservative?
Not reliably, and that is the point of this page. It is conservative about the source, which it assumes to be infinitely strong, but it is optimistic about the transformer, because it uses the nameplate impedance instead of the low tolerance limit. The optimism is 8.1 percent and the conservatism is 6.5 percent at a 250 MVA bus, so the net error is on the unsafe side.
Does 22 kA include motor contribution?
No. The transformer calculation gives you the transformer’s contribution only. Running motors add roughly four times their full load current during the first cycle. On a 1,000 kVA unit the gap to 22 kA is 1,082 amps, which is 270 amps of motor full load current, about one 200 hp motor. Above that, motor contribution alone puts you over.
What is the difference between 22 kA and 22 kAIC?
Nothing on a breaker. Both mean an interrupting rating of 22,000 amps symmetrical RMS at the voltage marked on the device. The formal term in the listing is interrupting rating, and kAIC is the trade shorthand for it. Neither term says anything about the peak current, which is a separate rating calculated from the X/R ratio.
Can a series rated panel solve this?
Sometimes, with limits. A series rating is a property of a tested pair of devices, listed in the manufacturer’s tabulation, not a calculation you can perform. The allowance is also restricted where motor load is significant, commonly at 1 percent of the lower device’s interrupting rating, which is 220 amps of motor full load current for a 22 kA device.
Does conductor length reduce fault current enough to matter?
Yes, more than the tolerance adds. Three sets of 500 kcmil copper for 100 feet adds about 0.0081 per unit impedance on a 1,000 kVA base, which drops the fault current from 20,918 to 18,332 amps, a 12.4 percent reduction. That is larger than the 8.1 percent the tolerance adds, which is why 22 kA gear often works in practice despite being marginal at the terminals.
What impedance should I specify for 1,000 kVA at 480 volts?
Six percent minimum, guaranteed, if you want 22 kA to hold at the low tolerance limit. The low limit of 6.0 percent is 5.55 percent, which produces 21,672 amps, so the rating holds with 1.5 percent of margin. If there is meaningful motor load, 6.5 percent is the number that survives both the tolerance and the motors.
Why does the same transformer need 65 kA at 208 volts?
Because fault current scales inversely with voltage while kVA stays fixed. A 1,000 kVA unit produces 20,918 amps at 480 volts and 48,273 amps at 208 volts, with the same impedance. The 208 volt service needs 65 kA equipment while the 480 volt service needs 22 kA, and nothing on the transformer is different.
Free: Transformer Specification Checklist
The exact RFQ sheet our engineers use — send it and get a quoted proposal within 24h.
Download checklist
