How to Decide Whether a Supply Can Start What You Are Adding

Why this matters

The equipment arrives, it gets connected, it runs, and the site's other machines start misbehaving in ways nobody can attribute. What went wrong is almost never capacity in the sense people check. The supply had plenty of room for the new load to run; it did not have room for the new load to start, and those are two separate tests against two separate numbers that can return opposite answers on the same site. This is the assessment sheet that makes them disagree on paper, before the crane is booked.

The sheet, and how to fill it

Eight fields. Each one is needed by the field after it, which is why they are in this order rather than in order of how hard they are to get. Fields 1 to 4 answer the running question. Fields 5 to 8 answer the starting question, and that is the one that usually fails.

1. The source, from its nameplate

Record nominal voltage, phase configuration, the source transformer's rating in kVA and its nameplate percent impedance. Percent impedance is the field people skip and it is the one the starting test cannot run without: it is the drop the transformer develops at its own rated current, and it is printed on the transformer, not derivable from anything else. Where the transformer is utility-owned and inaccessible, that number belongs to the utility and asking for it is a phone call, not an obstacle.

Compute the transformer's rated secondary current from its rating and voltage, because every later ratio is against that figure rather than against kVA.

2. The existing running load, measured at its own peak

Log current at the point of connection across a full duty cycle, in the peak season. A reading taken on a mild day in a building whose peak is set by cooling is not a measurement of the peak, it is a measurement of a different building. Installing a logger is energized work, permitted under 29 CFR 1910.333(a)(1) only where the employer can demonstrate that de-energizing introduces additional or increased hazards or is infeasible, which a load survey on an occupied building genuinely does. Clamp a single conductor without opening the enclosure further than the reading needs, work to the boundaries and PPE from NFPA 70E-2021, 130.5 and 130.7 in the edition your employer's program adopts, and use an instrument rated for the system.

What breaks if you skip it: a calculated load from a panel schedule reflects connected load, not demand, and it is the wrong number in both directions on most sites.

3. The new load's running current, from its nameplate

For a motor, note that two different currents exist and they are used for two different things. The full-load current from NEC Table 430.250 in the edition your authority having jurisdiction has adopted is what conductor and protection sizing uses; the nameplate current is what the overload is set from. Do not average them and do not substitute one for the other.

4. The running test

Add the measured peak from field 2 to the running current from field 3, and compare it against the source's rated current, the shared conductor's ampacity and the rating of every overcurrent device in the path. Conductor ampacity and its correction factors live in NEC Article 310, load calculation in Article 220 and motor circuits in Article 430, all in the adopted edition, and the authority having jurisdiction owns whether a given approach is accepted.

What breaks if you skip it: nothing subtle. A supply that fails this test fails visibly, on the first hot afternoon.

5. The new load's starting current, from the top of its band

For a motor, locked-rotor current comes from the nameplate code letter, which states a range of locked-rotor kVA per horsepower rather than a single value. Use the top of that band. You are sizing a margin against a disturbance, and the bottom of the band is the case you will not get on the day the machine is cold and fully loaded. For a transformer, a rectifier or a capacitive load, the starting or energizing current is a manufacturer figure and it can be a much larger multiple than a motor's.

6. The predicted dip at the point of connection

Two contributions add. The transformer's contribution is its nameplate percent impedance times the ratio of starting current to its rated current, which holds where the source impedance is dominated by transformer reactance and the starting current is heavily lagging. The shared conductor's contribution is its measured drop at a known current, scaled by the ratio of the start-condition current to that known current, and treated as an upper bound because a resistive run develops less drop from a lagging current than a straight multiplication suggests. The mechanism behind both is covered in the sibling reference on what inrush does to everything else on the supply.

Two details change the answer. The current through the shared conductor at the moment of start is the existing load plus the starting current, not the starting current alone, and using the sum is the defensible assumption, because the existing load's response depends on what it is: resistive and constant-impedance load draws less at reduced voltage, while motors holding constant torque and any constant-power load draw more. On a motor-heavy or drive-heavy panel the sum is a floor rather than a ceiling, a reason to widen the margin you demand. And only the elements between the source and the point of connection count: a branch carrying none of the starting current sees only what develops upstream of its tap.

7. The tolerance of the most sensitive thing already on the shared path

This is the field almost every survey omits, and it is the one that decides. Walk the shared path and record the lowest voltage tolerance on it: a backup supply's transfer threshold, a drive's undervoltage trip, a contactor coil's dropout voltage, a controller's supply holdup. These are data-sheet or settings values, so read them rather than estimate them.

What breaks if you skip it: you compute a dip and then have nothing to compare it against, so the assessment silently becomes an opinion about whether 7 percent sounds like a lot.

8. The starting test

Compare the predicted retained voltage from field 6 against the tightest threshold from field 7. Pass, fail, or too close to call, and "too close to call" is a legitimate result rather than a failure to decide.

The sheet, filled in

A three-phase 208Y/120 V service. The customer wants a 10 hp three-phase motor added to an existing panel.

Running test. 120 plus 30.8 is 150.8 A against 208 A rated, which is about 72 percent of the transformer's rating. Conductor ampacity and the overcurrent devices in the path also clear it. The running test passes, and it passes comfortably enough that anybody checking only this field would sign the job off.

Starting test. The transformer contributes 4 percent times 175 divided by 208, which is 3.36 percent. The shared feeder carries 120 plus 175, which is 295 A at the moment of start, so the ratio against the measured condition is 295 divided by 120, which is 2.46, and the feeder's contribution is 1.5 percent times 2.46, which is 3.69 percent as an upper bound. Together that is 7.05 percent, leaving about 93 percent of nominal retained at the panel bus.

The tightest threshold on the shared path is 90 percent retained. Predicted 93 percent clears it by about 3 points. Pass, with the margin stated rather than implied.

Now change field 7 and watch the result change without any of the electrical arithmetic moving. Suppose the same panel also feeds a drive whose undervoltage trip is set at 8 percent below nominal, which is 92 percent retained. The predicted 93 percent clears that by 1 point, and 1 point is inside the uncertainty of a calculation whose feeder term was explicitly an upper bound and whose starting current came from the top of a code-letter band. The correct output of the sheet there is not "pass". It is too close to call, and the resolution is to measure a dip on that panel during an existing large start and calibrate the feeder term against reality before committing.

Change field 2 instead. Had the existing peak been logged in the shoulder season at 70 A rather than 120 A, the feeder drop measured at that same moment would have read about 0.875 percent rather than 1.5. The drop and the current it was measured at are one pair and neither travels alone: keep the 1.5 percent and scale it by a 70 A ratio and you are mixing two measurement bases. Done properly it computes 0.875 times 245 over 70, or 3.06 percent from the feeder and 6.42 percent total, predicting 93.6 percent retained against a reality of 93.0. So the mistimed logger did not make the answer conservative, it made it optimistic, and by more the larger the site's seasonal swing. That is why field 2 specifies the season.

The failure mode this sheet exists to prevent is a survey that reports a percentage of transformer capacity and stops. Seventy-two percent loaded sounds like room. It is a true statement about running load and it contains no information at all about whether the next start will drop a controller.

When the sheet returns a no

Four remedies, and each belongs to a different party, which is why naming the owner matters more than naming the fix.

Reduce the starting current. A different starting method changes what the line sees, and it changes what the motor sees differently depending on whether the method transforms voltage or merely drops it. That trade has consequences for acceleration time and winding heating, and the sibling reference on what a starting method changes about the supply carries the arithmetic. The motor manufacturer owns whether a given method is permitted on that machine.

Reduce the shared impedance. A larger or separate feeder from the source removes the shared element rather than reducing the current through it, and it is the only remedy that also improves running voltage. Conductor sizing belongs to the adopted NEC edition and the authority having jurisdiction.

Raise the source. A larger transformer or a service upgrade is the utility's and the designer's call, and it has a lead time measured in months on many systems, which is why field 1 gets filled at the quoting stage rather than at the install.

Move the victim. Where one sensitive load sets the threshold and everything else on the path has wide tolerance, re-feeding that one load from upstream of the shared element is often the smallest true fix. It stops being the right answer the moment there are several victims.

When the sheet returns "too close to call"

Say so on the quote and price the measurement that resolves it. Capture a start the equipment makes on its own schedule rather than commanding one: every motor and compressor has a starts-per-hour limit, hermetic compressors carry anti-short-cycle timing, and a deliberate start on a supply you have just called marginal can drop the exact load field 7 flagged. If nothing starts inside the window, extend it. A one-day recorder on the panel during an existing large start gives you a measured dip and a measured current at the same instant, which calibrates the feeder term for that specific building and turns every later assessment on that site into arithmetic instead of an estimate.

What that recording must capture to be worth anything: the current at the source or the shared feeder and the voltage at the panel bus, sampled fast enough to resolve a start, timestamped together. A voltage log without the matching current log tells you a dip happened and not what caused it, which is the same position you were in before.

References

• 29 CFR 1910.333(a)(1) energized-work gate and (b)(2) electrical lockout; 29 CFR 1910.147(a)(1)(ii)(C) excludes electric utilization installations; 29 CFR 1926.417 construction counterpart
• NFPA 70E-2021, 130.5 and 130.7 for boundaries and PPE and 120.5 for live-dead-live verification, in the edition adopted by your employer's electrical safety program
• NEC Articles 220, 310 and 430, and Table 430.250, in the edition adopted by your authority having jurisdiction
• Transformer nameplate percent impedance and rating; motor nameplate code letter and manufacturer data
• See related: What Inrush Does to Everything Else on the Supply; What a Starting Method Changes About the Supply