Why Inexpensive Hazardous Location LEDs Fail in Year Three

The Driver Is What Fails

Start with the blunt fact. In hazardous-location LED luminaires, the dominant field failure mode is the driver, not the light-emitting diodes.

The reason is component physics. An LED die is a solid-state device with no wear-out mechanism short of thermal degradation of the package and phosphor. A driver is a switch-mode power supply containing electrolytic capacitors, magnetics, semiconductors, and solder joints — dozens of components, any one of which ends the fixture’s life when it fails.

Electrolytic capacitors set the ceiling. Their lifetime follows Arrhenius behaviour: operating life roughly halves for every 10°C rise in core temperature. A capacitor rated 10,000 hours at 105°C delivers perhaps 40,000 hours at 85°C and 160,000 at 65°C — or, running the arithmetic the other direction, a driver whose internal temperature sits 20°C above what the designer assumed has lost three-quarters of its life before it ever ships.

This is why the driver’s marked Tc point — the case temperature at which the manufacturer rates it — is the single most useful number on a hazloc luminaire’s supporting documentation, and the one least often supplied.

Now put that driver inside a sealed flameproof enclosure. There is no forced convection. Waste heat from the LED array rises. The driver is frequently mounted in the upper body of the housing, in the hottest air in the fixture. In a Permian summer or a Gulf Coast August, the enclosure’s internal air can sit 25–40°C above the outside ambient — which is itself well above the 25°C laboratory condition most component ratings assume.

Cheap fixtures fail in year three because their drivers were specified against ambient temperature rather than against internal enclosure temperature, and nobody measured the difference.

The LM-80 and TM-21 Gap

Here is where the datasheet actively misleads, and it’s worth being precise because the IES has been unusually direct about it.

IES LM-80 is a test method. It measures lumen maintenance of LED packages, arrays, and modules — not luminaires — over a minimum of 6,000 hours, sampled at 1,000-hour intervals, at a minimum of three case temperatures, typically 55°C, 85°C, and a third of the manufacturer’s choosing.

IES TM-21 is the projection method that extrapolates LM-80 data into an L70 figure — the hours until output falls to 70% of initial. It carries a hard limit: projections may not exceed six times the test duration. Ten thousand hours of test data supports a 60,000-hour claim and no more.

The IES position statement PS-10-18, IES Position on LED Product Lifetime Prediction, says this plainly. Marketing claims beyond the 6× limit are not supported by IES standards. A 100,000-hour L70 claim built on 10,000 hours of LM-80 data is not a conservative estimate; it is outside the method. And critically, the IES states that it is inaccurate to make lifetime claims about an entire luminaire using TM-21, because the method applies only to the LED components inside it.

Four gaps follow, and every one of them lands on the buyer:

1. LM-80 tests the LED package in a controlled oven. It does not test your driver, your gaskets, your optics, your solder joints, or your enclosure.
2. L70 is a dimming metric, not a survival metric. It describes how a surviving LED fades. It says nothing about the probability that the fixture is still energized.
3. The case temperatures may not resemble yours. An L70 figure quoted at 55°C case temperature is not applicable to a driver-heated sealed enclosure at 55°C ambient, where internal temperatures are considerably higher.
4. Luminaire-level methods exist and are rarely quoted. IES LM-84 and TM-28 extend lumen-maintenance testing and projection to complete luminaires, accounting for driver and thermal system effects. A supplier who publishes LM-84/TM-28 data has tested the product you’re buying. A supplier who publishes only LM-80/TM-21 has tested a component someone else made.

Ask for LM-84 and TM-28. The response is diagnostic on its own.

Chemical Attack: H₂S and Chlorides

Sour service is a solved problem for piping metallurgy and a largely unsolved one for lighting electronics.

The relevant framework is ANSI/ISA-71.04, which classifies environmental corrosivity for electronic equipment using copper and silver coupons exposed for 30 days and measured in angstroms of film growth:

Most of a gas plant, a sour crude unit, a wastewater headworks, or a lift station — whatever its area classification — sits in G2 or G3. Consumer-grade and light-industrial electronics are designed for G1.

The mechanism that catches people out is creep corrosion. Hydrogen sulfide attacks silver and copper, forming sulfide films that migrate across a printed circuit board’s surface — including across areas with no electrical function — until they bridge adjacent conductors and short them. The board looks superficially intact. It is electrically dead.

RoHS made this worse, and the industry doesn’t talk about it much. Lead-free PCB surface finishes — immersion silver, ENIG, OSP, immersion tin — are more vulnerable to sulfide attack than the leaded finishes they replaced. A driver built to consumer standards, with an immersion-silver finish, in a G3 environment, has a service life measured in months to a couple of years regardless of how good its thermal design is.

Conformal coating is the standard mitigation, and it’s worth being honest about its limits. Coating protects what it covers. It does not reliably cover board edges, connector contacts, the underside of tall components, or areas masked for thermal contact — and those are exactly where sulfide creep initiates. A “conformal coated” claim without a stated coating class, thickness, and masking scheme tells you very little. Potting is more effective and brings its own trade-offs in thermal transfer and repairability.

Offshore and Coastal: Salt Fog

Chloride attack works differently and needs testing to match.

The distinction that matters in a specification: ASTM B117 salt spray is a static, continuous exposure test, widely quoted and weakly predictive. IEC 60068-2-52, Test Kb, is a cyclic salt mist test — alternating salt exposure with humidity and drying phases — which reproduces the wet/dry cycling that actually drives marine corrosion. A supplier quoting cyclic salt mist hours has tested against a realistic mechanism. One quoting only B117 hours may not have.

The specific failure modes offshore and on the coast:

• Galvanic couples. Stainless fasteners threaded into aluminum bosses, dissimilar-metal mounting brackets, and steel conduit into aluminum entries all corrode preferentially at the aluminum, which is your enclosure.
• Alloy copper content. Marine-service aluminum housings should use low-copper alloys — conventionally below 0.4% copper. Higher-copper alloys pit rapidly in chloride environments regardless of coating quality.
• Coating systems. ISO 12944 defines corrosivity categories through C5 and CX; offshore is CX. NORSOK M-501 is the reference many operators specify. A powder coat qualified for C3 industrial service will not survive a splash zone.
• Chloride plus sulfide together. Where both are present — and on many offshore platforms they are — attack is faster than either alone. ISA 71.04’s severity levels are based on combined reactivity for exactly this reason.

Vibration on Compressor and Pump Skids

A fixture bolted to a reciprocating compressor skid, a pump baseplate, or a structure carrying rotating equipment lives in a different mechanical world than one on a building wall.

What breaks, in order of frequency:

Solder joint fatigue on heavy components. Electrolytic capacitors and inductors are the tallest and heaviest parts on a driver board. Under vibration they act as levers on their solder joints. Cracks propagate over months, producing intermittent operation before outright failure — which is why “it works when you tap it” is a diagnosis rather than a mystery.

Connector fretting. Micro-motion at connector contacts wears through plating and builds insulating oxide films. Board-to-board and wire-to-board connectors are common failure points; soldered or crimped terminations are markedly better.

Resonance, not amplitude. This is the part most specifications miss. A fixture doesn’t fail because vibration is severe; it fails because the mounting arrangement has a natural frequency near an excitation frequency of the machine. A bracket that is perfectly adequate at 1× running speed can amplify at 2× or 4×. Reciprocating compressors are especially punishing because they excite multiple harmonics simultaneously.

Specify testing, not adjectives: IEC 60068-2-6 (sinusoidal), IEC 60068-2-64 (random), or ANSI C136.31 vibration classes, with the g level and frequency range stated. “Vibration resistant” is not a specification.

The Ingress Problem Nobody Designs For

Everyone specifies an IP rating. Almost nobody accounts for what a sealed enclosure does over a daily thermal cycle.

IEC 60529 defines the IP code — IP66 for dust-tight and powerful water jets, IP67 for temporary immersion, IP68 for continuous. These are static tests. Your fixture is not static.

Here’s what actually happens. The fixture heats during operation and internal air expands, raising internal pressure slightly. It cools overnight, internal pressure drops, and the enclosure inhales through whatever path is available — gasket interfaces, thread paths, cable entries. In humid air, it inhales moisture. That moisture condenses on the coolest interior surface as the fixture cools further, and on many designs the coolest interior surface is the driver board.

Repeat 365 times a year and a fixture that passed IP66 on the day it shipped accumulates liquid water inside it. This is why fixtures fail after the first full seasonal cycle rather than immediately, and why coastal and offshore installations fail faster: the inhaled air is salt-laden.

Mitigations that actually work: certified breather-drain devices appropriate to the protection concept, which allow pressure equalization while draining condensate; gasket materials chosen for the real temperature range (silicone and fluorosilicone hold up where EPDM takes a compression set); and mounting orientation that puts the drain at the low point rather than wherever the bracket happened to allow.

One more, which connects to a broader point: every time the enclosure is opened, the gasket clock resets — and not favourably. A gasket that has taken a compression set does not re-seal to its original performance. Opening a fixture for a lamp change or a driver swap is a sealing event that needs the gasket inspected and, often, replaced. That’s part of why minimizing joint disturbance has real value, and part of why the installed cost of touching a fixture is so much larger than the parts cost.

Why Year Three Specifically

The timing isn’t coincidence. Several curves converge:

• Thermal wear-out. A driver running 20–25°C hotter than designed exhausts a nominal 50,000-hour capacitor life in roughly 12,000–20,000 hours. At continuous operation, that’s 18 to 30 months.
• Corrosion incubation. Sulfide creep and chloride pitting need time to bridge or perforate. Two to three years is a typical incubation in G2–G3 service.
• Seasonal cycling. Moisture accumulation from enclosure breathing needs several full annual cycles to reach failure quantities.
• Warranty structure. Many warranties run five years, but they are typically parts-only. The fixture that dies in year three costs you nothing in parts and several thousand dollars in access, permits, and labour. The warranty pays for the cheapest component of the failure.

That last point is the one to bring to a procurement conversation. A five-year warranty on a product that fails at three years transfers almost none of the actual cost.

What to Ask For

A supplier who has engineered for this environment can answer all of these. One who hasn’t will answer none of them, and the pattern of non-answers is more informative than any single specification. Ten parallel questions on the certification side work the same way.

1. Driver Tc point, and measured internal enclosure temperature at your specified maximum ambient — not calculated, measured.
2. Capacitor type, rated life, and rated temperature. Ask whether the design uses electrolytics at all; some drivers avoid them.
3. LM-84 and TM-28 data for the complete luminaire, not just LM-80/TM-21 for the LED package.
4. Conformal coating class, thickness, and masking scheme — or potting details.
5. ISA 71.04 severity level the product is designed for, with supporting mixed-flowing-gas test data.
6. Cyclic salt mist hours per IEC 60068-2-52, not just ASTM B117.
7. Vibration qualification with standard, g level, frequency range, and axis count.
8. Gasket material and rated temperature range, and whether a certified breather-drain is fitted.
9. Field failure data — actual return rates by year, by environment. Suppliers who track this will share it.
10. What the warranty actually covers: parts only, or parts and labour, and whether access cost is included.

Being Straight About Our Own Limits

[SECTION TO BE COMPLETED BY THE PUBLISHER] — this is the one part of the article that cannot be written from published standards. It should state plainly where your own product’s boundaries are. Suggested structure, to be filled with your real figures:

• The maximum ambient at which you warrant full output, and what happens above it — thermal foldback, reduced life, or not recommended.
• The ISA 71.04 severity level you design and test to, and the environments where you would tell a customer to specify something else or add environmental controls.
• Your vibration qualification level, and the skid applications that exceed it.
• Where your gasket and breather arrangement has known limits — orientation constraints, or a required inspection interval.
• Your actual field return rate, if you track it, and what the dominant returned-part failure mode is.
• Applications you decline. Naming two or three situations where you are not the right answer does more for credibility than any specification sheet.

Everything above this section is verifiable against published standards. This section is the one that determines whether a reader believes the rest of it.

Bottom Line

Inexpensive hazardous-location LED fixtures don’t fail because LEDs are unreliable. They fail because the driver was rated against the wrong temperature, the board finish was chosen for a G1 environment and installed in a G3 one, the mounting was never resonance-checked, and the enclosure spent three years quietly breathing humid air onto the power supply.

None of that appears in an L70 number, because L70 was never designed to describe it.

Every one of those answers belongs in the asset register at closeout, not in a quote folder. The specification questions that predict year-three survival are about driver thermal margin, board-level corrosion protection, mechanical qualification, and moisture management — and every one of them can be answered with test data by a supplier who has done the work.

Ask for the data. The silence is the answer.

References: IES PS-10-18 on LED lifetime prediction · Illuminating Engineering Society · ANSI/IES RP-7-21 · IEC 60529 (IP Code) · IEC IP ratings overview · IEC 60068-2-52 cyclic salt mist · IEC 60079-0 · IEC 60079-1 flameproof · IEC 60079-17 inspection · IEC 60079-19 repair · IEC 60079 series · ISA · ASTM · ISO · UL 844 · IECEx · NFPA 70 / NEC · OSHA 1910.307