Why 20 LED Drivers Trip a 16 A Breaker: Inrush Current, MCB Curves and the Maths Nobody Does

|How to|10/28/2026|12 min|
Table of contents
Share Post

A panel builder in Poland sent us a photo last winter: a row of twenty 60 W drivers, all brand new, all dead on the same circuit. A 16 A miniature circuit breaker had tripped at switch-on, the maintenance team reset it, it tripped again, and by the fourth reset two drivers had lost their input rectifiers. The electrician’s conclusion was that our drivers were faulty. The real cause was that nobody had done the inrush calculation, because the arithmetic most people do says twenty units is fine.

It isn’t fine. And the reason it isn’t fine has almost nothing to do with the load those drivers present once they are running.

This article is the calculation we now send to panel builders before they wire anything. It comes with measured numbers from our own bench, the breaker curves that actually matter, and the one failure mode that only shows up after a power cut — which is why so many cabinets pass the factory test and fail on site three weeks later.

The arithmetic that gets people into trouble

Take a 60 W constant voltage driver on a 230 V supply. Input current at full load is around 0.43 A once you account for a power factor of about 0.6. Divide 16 A by 0.43 and you get 37 drivers. Apply the standard 80% derating for continuous load and you get 29. Apply the rule that says never load a breaker above 80% of its rating and you might land on 24 to be safe.

So the electrician puts 20 on a B16 breaker and expects a comfortable margin. At 20 × 0.43 A, the running current is 8.6 A, which is 54% of the breaker rating. Nothing about that load profile should trip anything.

What the calculation ignores is the first 900 microseconds after the contactor closes.

What inrush current actually looks like

Every switch-mode driver has a bridge rectifier feeding a bulk capacitor. At the instant of switch-on, that capacitor is empty, so it looks like a short circuit. The only thing limiting the current is the resistance of the wiring, the impedance of the supply, and whatever inrush limiting the driver has on its input — usually a negative temperature coefficient thermistor, sometimes a fixed resistor with a relay bypass.

On our 60 W models, the cold-start inrush peak is typically 25 to 32 A and it decays with a time constant of roughly 0.6 to 1.2 ms. On the 400 W PS400-H1V, it is 55 to 70 A. These are not unusual figures; a 400 W supply with 85–88% efficiency and a 175–240 V input range has a bulk capacitor in the hundreds of microfarads, and that capacitor has to be filled.

Twenty drivers switching at once means twenty capacitors charging at once. If the contactor closes at a random point on the mains waveform, several of them will close near the voltage peak, and those are the ones that draw the full peak.

The naive calculation — 20 × 30 A = 600 A — is also wrong, but for the opposite reason. It overstates the problem. Here is why.

The half-cycle integration that everyone forgets

A miniature circuit breaker does not trip on instantaneous peak current. Its magnetic element responds to the current that persists over roughly half a mains cycle, because the trip coil has to build enough flux to move the latch. A 30 A spike that lasts 800 µs is not equivalent to 30 A for 10 ms. It is equivalent to something closer to 8 to 12 A for that half cycle.

This is why 20 drivers sometimes work and 20 drivers sometimes trip the same breaker, depending on the closing angle of the contactor and the ambient temperature of the thermistors. It is also why the field reports are contradictory: two identical cabinets, one works, one doesn’t.

But it also means the margin is razor thin. A B-curve breaker trips magnetically at 3 to 5 times its rating. On a B16 that is 48 to 80 A. Once you are relying on the difference between a 0.8 ms spike and a 10 ms window, you are no longer engineering, you are gambling on the phase angle.

Typical measured values on our bench, 230 V / 50 Hz, 25 °C:

Driver Rated power Input current at full load Cold inrush peak Inrush duration (to 10%)
60 W CV 60 W 0.43 A 25–32 A 0.6–1.0 ms
100 W CV 100 W 0.72 A 30–38 A 0.7–1.2 ms
150 W CV 150 W 1.09 A 38–45 A 0.8–1.4 ms
250 W CV 250 W 1.81 A 45–55 A 0.9–1.5 ms
400 W CV 400 W 3.60 A 55–70 A 1.0–1.8 ms

Two things in that table are worth pausing on. The 400 W unit draws 3.6 A at 230 V, not the 1.74 A you would get by dividing 400 by 230 — because the power factor is 0.6, and the input current is set by apparent power, not real power. And the inrush peak barely changes between 60 W and 400 W, which is the counterintuitive part: a 400 W driver is not six times harder to start than a 60 W driver. It is about twice as hard. The bulk capacitor size does not scale linearly with output power, because the holdup time requirement doesn’t either.

The hot-restart trap

Here is the failure mode that costs people the most money, and the one that no datasheet mentions prominently.

An NTC thermistor works by being hot. At 25 °C it might present 5 Ω, which limits inrush to 46 A on a 230 V peak of 325 V. After a few seconds of running, it self-heats to 120 °C or more and drops to well under 1 Ω. That’s the point — you don’t want 5 Ω in series during normal operation, because it would dissipate 1 W and cook itself.

Now the power fails. Or someone throws the isolator to change a lamp. Or the building’s transfer switch operates during a generator test. The contactor re-closes after 3 seconds. The thermistors are still at 110 °C. Their resistance is maybe 0.8 Ω instead of 5 Ω. Inrush is now five times higher than the cold-start figure on the datasheet.

This is why a cabinet that passed a factory run-test at 9 a.m. can trip the breaker at 6 p.m. after a brief outage. It is also why the “we tested it, it works” defence from a supplier is worthless unless the test included a hot restart.

If you want to reproduce it: energise the circuit for ten minutes, kill the supply for five seconds, re-energise. If it holds, the design has real margin. If it trips, the cabinet has a latent problem that will surface at the worst possible moment.

Where the breaker curve actually sits

Most European panels use B-curve breakers on lighting circuits, because B-curve is cheap and the loads are usually resistive-ish. That is the wrong choice here.

Breaker Thermal trip 1.13–1.45 × In Magnetic trip Effective ceiling for 60 W drivers (measured)
B16 18–23 A 48–80 A 8–10 units
C16 18–23 A 80–160 A 14–16 units
D16 18–23 A 160–320 A 20–24 units
C20 23–29 A 100–200 A 18–20 units

The thermal column is why you cannot just keep adding drivers: at 20 units × 0.43 A you are at 8.6 A, which is fine, but at 35 units you are at 15 A, which is 94% of a 16 A breaker and will trip on a warm afternoon in a closed cabinet. The magnetic column is why you cannot just buy a bigger breaker to solve the inrush problem: going from B16 to D16 solves the inrush but leaves you with a breaker that will not clear a genuine fault on a 1.5 mm² cable.

That trade-off — protection versus nuisance tripping — is the actual engineering problem. Upsizing the breaker is almost always the wrong fix, because it fixes the symptom and removes the protection. There is one exception, covered below.

Leakage current: the second thing that trips

There is a second failure mode that gets blamed on inrush and isn’t.

Every driver has Y-capacitors from line and neutral to earth for EMI suppression. Each one leaks a small current to earth, typically 0.25 to 0.75 mA depending on the capacitor size and the switching frequency content. Twenty drivers at 0.5 mA is 10 mA of standing leakage on a 30 mA RCD. Add cable capacitance on a long run — 100 m of four-core cable can contribute another 2 to 5 mA — and you are at 15 mA before anything has even switched on.

A 30 mA RCD is designed to trip somewhere between 15 and 30 mA, and it must not trip below 15 mA. If your standing leakage is already at 14 mA, the next event that adds a milliamp will take the circuit out, and it will look like an intermittent fault that no one can find.

This is why we ask panel builders for the driver count per RCD, not just per breaker. The rule of thumb we use: no more than 25 drivers per 30 mA RCD, and if the run is over 80 m, treat it as 15.

Four fixes, ranked by what they cost

1. Split the circuit. Two circuits of 10 drivers each instead of one of 20. Costs one extra breaker and 3 m of cable. Solves both the inrush and the leakage problem at the same time. This is the correct answer 70% of the time.

2. Stagger the start. A contactor with a 0.5 s delay, or a sequencer relay, or just two contactors on a timer. Costs €30 and eliminates the simultaneity that causes the trip. If the cabinet has a PLC, use it. This is the second-best answer and it scales: with staggering, the limit goes back to the thermal calculation, which is 35 drivers on a C16.

3. Use C-curve breakers. Moving from B16 to C16 roughly doubles the magnetic threshold for about €2 extra per pole. It does not change the thermal limit, so it pairs well with fix 1 or 2. Do not go to D-curve on a lighting circuit unless the supply impedance is genuinely high and you have calculated the fault loop.

4. Specify a driver with active inrush limiting. A relay that bypasses the NTC after 200 ms gives you a cold inrush of 30 A and a hot inrush of 30 A, because the limiting resistor is always cold when it does its job. It costs more — roughly 4% on a 150 W unit — and it is the only fix that solves the hot-restart problem at the component level rather than at the panel level.

What does not work: adding a bigger breaker (removes protection), adding an inrush current limiter module in series with twenty drivers (the module has its own thermal limits and will cook), or hoping the building’s supply impedance is high enough to limit the peak (it varies by site and by time of day).

What to ask before you buy

If you are specifying drivers for a panel and you do not want to run this calculation yourself, ask your supplier four questions. The answers will tell you whether they have actually measured anything.

  1. What is the cold-start inrush peak and the duration to 10%? If they can only give you a peak number with no duration, they are quoting a simulation, not a measurement.
  1. What is the inrush on a hot restart, five seconds after switch-off? Most suppliers will not have this number. The good ones will.
  1. What is the earth leakage current per unit, at nominal voltage and at 110%? This one decides your RCD count.
  1. What is the input current at 100% load and at 50% load? Input current is not linear with load on a driver with a passive front end, and the 50% figure is what you need for thermal sizing in a sealed cabinet.

At SANPU we publish the input current and the inrush figures on the datasheet for the 400W IP20 indoor driver and the 250W 24V indoor power supply because panel builders kept asking for them and we got tired of emailing PDFs. If you need the full waveform, we can send the oscilloscope capture — we use a HIOKI MR6000 with a 10 MHz current probe for this, and the capture includes the closing angle, which matters more than most people realise.

One more thing about that Polish cabinet

The two dead drivers were not killed by the inrush. They were killed by the fourth reset. Each re-energisation pushed another surge through the rectifier while the bulk capacitor was partially charged and the thermistor was hot, and the bridge ran at its absolute maximum forward surge current four times in ninety seconds. The breaker was doing its job. The panel was asking it to do the wrong job.

Twenty drivers on a B16 is a design that works until it doesn’t. Ten on a C16, with a 0.5 s stagger, is a design that works in February in an unheated plant room with the contactor closing at the worst possible phase angle. The difference in bill of materials is about €40. The difference in field failures is the difference between a warranty claim and a repeat order.

FAQ

Can I put 20 LED drivers on a 16 A breaker? Only if they are 60 W or smaller and the breaker is C-curve, and even then you are relying on the phase angle at switch-on. Measured, a C16 holds 14 to 16 units of 60 W drivers on a cold start, fewer on a hot restart. Split the circuit instead.

Why does my breaker trip only after a power cut? Because the NTC thermistors are still hot and no longer limiting inrush. A five-second outage is the worst case: the capacitors have discharged enough to demand full charging current, but the thermistors have not cooled. That combination gives you five times the datasheet inrush.

Is it safe to use a D-curve breaker to stop the tripping? It stops the tripping and it also stops the breaker from clearing a short circuit on a small cable. D-curve magnetic thresholds start at 10 × In, which on a 16 A device is 160 A — too high for a 1.5 mm² final circuit. Fix the inrush instead.

Does a soft-start module in the panel work? Yes, if it is rated for the total inrush energy and not just the current. Most panel-mounted limiters are specified for a single driver’s I²t and will fail open after a few hundred operations on a twenty-driver circuit.

What about 230 V vs 120 V? At 120 V the inrush peak is roughly half, but the input current for the same wattage is roughly double, so the thermal limit on a 16 A breaker drops by half. The two effects very nearly cancel out, which is why the same cabinet design behaves similarly on both supplies — and why nobody notices the problem until they ship to a 230 V market.

Need the inrush waveform for a specific model before you wire the panel? We keep oscilloscope captures for every catalogue model and can send them with the closing angle and the hot-restart figure. Ask for it when you request a quote on the 400W IP20 indoor driver or browse the constant voltage indoor range for the 15 W to 600 W models.

SANPU

Stay in the loop

Subscribe to our free newsletter.