Why Outdoor LED Drivers Die After a Storm — and What 10/700 µs Means

|How to|11/06/2026|11.4 min|
Table of contents
Share Post

Every summer we get the same batch of emails, usually within a week of the first big thunderstorm: a car park in Italy, a billboard in Malaysia, a façade installation in Dubai. The drivers are dead, sometimes four out of twelve in a row, sometimes all of them on one circuit. The customer’s conclusion is that the product is not waterproof, or that the surge protection “didn’t work”, and both conclusions are usually wrong.

Two things are almost always true about these failures. First, the drivers were not hit by lightning. They were hit by the residual of a strike 200 metres away, arriving on a conductor nobody was protecting. Second, the driver datasheets said “surge: 10/700 µs, 6 kV” — a specification borrowed from telecom equipment that has nothing to do with a mains-fed outdoor luminaire.

Both of those facts are worth understanding before you specify the next outdoor job, because the difference between a driver that survives five summers and one that fails in the first is usually a €1.40 varistor and 40 mm of wiring.

The waveform on the datasheet is probably the wrong one

Surge testing comes down to one question: what shape is the pulse?

For mains-powered equipment in the EU, the relevant immunity standard is EN 61547, which points at IEC 61000-4-5. The standard waveform is a 1.2/50 µs open-circuit voltage combined with an 8/20 µs short-circuit current — the “combination wave”. Luminaires are typically tested at 1 kV line-to-line and 2 kV line-to-earth as a baseline, and outdoor or street-lighting specifications frequently demand 4 kV, 6 kV or 10 kV because the installation is exposed and the cost of a callout is high.

The 10/700 µs waveform belongs to a different world entirely: ITU-T K.21, surge testing for telecom subscriber lines. It models a slow, high-energy pulse arriving on a twisted pair from a distant strike. It is the right waveform to test a DSL modem. It is not the waveform that arrives at a street-light driver through a live conductor.

Both waveforms have their place, and testing to both is not harmful. But a datasheet that leads with “10/700 µs, 6 kV” and gives no 1.2/50–8/20 figure is telling you the product was tested against the wrong threat. When you see that, ask for the combination-wave result. If there isn’t one, you have found the reason the installation keeps failing.

Where the surge actually enters

Here is the part that changes how people design installations.

A direct strike on the luminaire is rare and usually destroys everything, including the evidence. What kills drivers in clusters is the indirect event: a strike to a nearby structure or to the soil, which raises the potential of the local earth by several kilovolts for a few tens of microseconds. Everything bonded to that earth rises with it. The neutral, the protective earth, the metal pole, the luminaire housing and the DC cable screen all move together — except the parts that are not bonded, or bonded through a long thin conductor, which see the difference.

There are three entry paths, and installers usually protect only the first:

  1. Mains L-N and L-E. The obvious one. Handled by a varistor or an SPD.
  1. N-E and the earth bond. Frequently ignored. A potential rise on the earth appears as a differential voltage between neutral and earth at the driver terminals.
  1. The DC side. This is the one that surprises people. A 15 m run of two-core cable to a luminaire is an antenna. An induced surge on that cable arrives at the driver’s output terminals, where there is no varistor at all, and takes out the output rectifier or the secondary-side capacitors.

Path 3 explains a pattern we see repeatedly: drivers that fail with a shorted output and an intact input section. The input protection did its job. The energy came in from the other end.

If you cannot protect the DC side — and on a potted IP67 driver you cannot retrofit protection inside — then the protection has to be at the luminaire, in a junction box, or you accept the risk and design for replacement. For long outdoor runs, a small TVS across the DC output at the luminaire end costs about €0.60 and is the single most effective addition most installations are missing.

MOVs are consumables, not permanent protection

A metal oxide varistor works by becoming conductive above its clamping voltage. Every event pushes it slightly out of specification: the clamping voltage drifts down, the leakage current rises, and after enough events the leakage becomes self-heating. A varistor that has absorbed twenty small surges can reach 80 °C at nominal mains voltage, which accelerates the drift, which increases the leakage, which increases the temperature. The end state is a short circuit and, if there is no thermal fuse, a small fire inside a plastic enclosure.

Good driver designs use a varistor with a thermal fuse in series, bonded against the body so it opens before the disc reaches ignition temperature. Cheap designs use a bare 7 mm disc with no fuse. You cannot see the difference from the outside, and both look identical on a datasheet that says “surge protection: yes”.

Two practical consequences:

  • A driver that has survived several storms may already be at the end of its varistor’s life. If a site has had two events and the drivers still work, that is not proof of robustness; it is a reason to check them. Measuring leakage current across the varistor at nominal voltage will show whether it has drifted.
  • Varistor voltage rating matters more than disc size for survival. On a 230 V nominal supply with a +10% tolerance, the continuous voltage is 253 V. A 275 VAC-rated varistor is the minimum sensible choice; a 250 VAC part will run in continuous conduction and fail in months. For 277 V North American supplies, 275 V is marginal and 300 V or a 320 V part is correct — which is one of several reasons a European outdoor driver should not be dropped into a 277 V circuit unchanged.

Let-through voltage: why an external SPD can make things worse

Adding a DIN-rail SPD in the pole base sounds like the obvious fix, and sometimes it is. But an SPD does not clamp to zero; it clamps to a let-through voltage that depends on the device class and the current. A Type 2 SPD might clamp a 5 kA 8/20 surge at 1,200 V. If the driver’s own immunity is 2 kV, that is fine. If the driver has no tested immunity because nobody asked, you have added €40 of hardware and moved the failure by one component.

There is a second, more insidious problem: lead length. At 8/20 µs, the current is changing at roughly 10⁹ A/s. A 30 cm connecting lead has an inductance of about 0.3 µH, which produces a voltage drop of 300 V per 100 A of surge current. On a 5 kA event that is 15 kV across the leads — more than the SPD’s clamping voltage, and it appears directly across the equipment you were protecting. This is why SPD installation instructions insist on leads under 50 cm, why you should never coil the spare cable, and why a beautifully specified SPD wired with a 1.5 m loop protects nothing.

The rule is simple and usually broken: the SPD must be within 50 cm of the equipment, with the shortest possible straight leads, and the earth bond must be at least as thick as the phase conductor. If the SPD is in the pole base and the driver is 3 m up the pole, the SPD is protecting the pole base.

Water and surge are different problems, and IP67 does not solve either

While we are here, because half of the “surge” failures we are asked about turn out to be water:

An IP67 rating is a test of a new sample, submerged under defined conditions, with the gland and the lid as supplied. It says nothing about a unit installed with a cable gland torqued by hand, a lid gasket that has been opened twice, or a cable whose jacket diameter is outside the gland’s clamping range.

The three failure points, in order of frequency:

  1. Gland torque and jacket diameter. A gland rated for 6–12 mm cable, clamped onto a 5.5 mm jacket, will pass the factory test and leak in service. Tighten to the specified torque — usually 2.5 Nm for a plastic gland — and check the jacket diameter against the gland range, not against the cable’s nominal size.
  1. Condensation, not rain. A sealed box that heats to 60 °C in the sun and cools to 15 °C at night pulls in moist air through the cable’s own interstitial spaces if the cable end is not sealed. Over a year, that is a millilitre of water inside the enclosure. A breather membrane (ePTFE vent) equalises pressure without letting liquid water in, and costs about €0.80.
  1. Capillary action through the cable. Water travels along the stranded conductors from a poorly sealed joint at the luminaire end. This is why the “water inside the driver” returns almost always come from installations where the luminaire-end connection was made with tape.

Notice that surge damage and water damage look different on the bench. Surge failures are usually a shorted rectifier or a blown varistor with blackening. Water failures show corrosion on the secondary side, green copper, and a driver that worked for nine months and then didn’t. Opening the failed unit tells you which problem you actually have, and it is worth insisting that your supplier does it before they send a replacement.

What to specify

A specification that will survive an actual thunderstorm looks like this:

  • Immunity: IEC 61000-4-5, 1.2/50–8/20 combination wave, 2 kV L-N, 4 kV L-E minimum. For exposed outdoor sites, 6 kV. Ask for the test report and the waveform, not a marketing number.
  • Varistor with integrated thermal fuse, rated 275 VAC minimum for 230 V systems, 320 VAC for 277 V systems, 14 mm disc or larger.
  • DC-side protection at the luminaire for runs over 10 m, or a driver specified with output TVS.
  • SPD within 50 cm of the driver, with a straight earth bond of at least 4 mm².
  • Gland torque and cable diameter range stated in the installation drawing, plus a breather vent on sealed enclosures.

For Class II (double-insulated) luminaires there is a further constraint: with no protective earth, surge current has no path to divert, and the varistor has to absorb the whole event between L and N. In that case the varistor rating and the series thermal fuse matter even more, and the 4 kV L-E figure is meaningless because there is no L-E path. Specifiers regularly write “4 kV L-E” onto a Class II product and then wonder why it fails. If the product has no earth, that test does not describe your installation; the L-N number is the one that counts.

The installation that got it right

A car park operator in northern Italy replaced 220 pole-top drivers after two seasons of failures. The original setup: 150 W IP67 drivers in the pole head, 4 m cable run to the luminaire, no SPD, no breather, glands hand-tightened. Failures clustered on the poles at the end of each row, which is the classic signature of an earth potential rise at the far end of a bonding run.

The replacement used the same power class, the 150W slim IP67 waterproof driver, with three changes: a Type 2 SPD in each pole head within 40 cm of the driver, a 6 mm² earth bond from the pole to the SPD and from the SPD to the driver, and an ePTFE breather on each enclosure. Two summers later, the return rate on those 220 units is zero. The driver did not change. The installation did.

That is the uncomfortable conclusion for anyone shopping purely on the driver’s surge rating: the number on the datasheet is at most half of what decides survival. The other half is bolted to the pole.

FAQ

Does “10/700 µs, 6 kV” on a datasheet mean the driver is well protected? It means it was tested against the telecom surge waveform. For a mains-fed outdoor luminaire you want the 1.2/50–8/20 combination wave result. Ask for both; a supplier who only has the first one has not tested for your installation.

Can a surge travel in on the DC output? Yes, and it is a common cause of failures that look like mains surge damage but leave the input section intact. Long outdoor DC runs pick up induced energy. Protect at the luminaire end or specify output-side TVS.

Do I need an SPD if the driver already has a varistor? For exposed sites, yes. The driver’s varistor is sized for the residual that gets past the installation’s protection, not for a direct-coupled 5 kA event. But the SPD must be within 50 cm with short straight leads, or its let-through voltage plus lead inductance exceeds the protection it was meant to provide.

Why did only some drivers on the circuit fail? Because surge is not evenly distributed. Poles at the end of a bonding run, poles nearest the strike, and poles with the longest DC cable see the highest residual. A cluster pattern points at earthing and bonding, not at a bad batch.

Is IP67 enough for a coastal installation? For water, mostly, if the glands are correct. For corrosion, no — salt mist attacks the housing screws and the cable jacket, and you should be asking for a salt-mist test result rather than an IP rating. Our IP20 vs IP67 enclosure design breakdown covers where the sealing actually fails.

Specifying for an exposed site? The 150W slim IP67 waterproof driver is available with a 6 kV combination-wave test report and a thermal-fused varistor, and we will send the test waveform rather than a number. If your site has already had a failure, send us the failed unit’s fault pattern — input or output side — and we will tell you which path the energy took, which is the only way to stop it happening again.

SANPU

Stay in the loop

Subscribe to our free newsletter.