Acceptance & QC

Burn-in: what a 72-hour soak test can and cannot prove

Why early failures cluster, what a pre-installation soak test actually catches, why accelerated ageing cannot validate a lumen maintenance claim, and a practical protocol for buyers without a laboratory.

Written by Long-Term New Media Limited · Last reviewed August 2026

The argument for running fittings before installing them is simple: it is far cheaper to discover a dead unit on a bench than on a ceiling. The argument against is that people expect burn-in to prove things it cannot, and then treat a passed soak test as a lifetime guarantee.

Why early failures cluster

Electronic assemblies fail on a bathtub curve. There is an elevated failure rate at the start — infant mortality — then a long flat period of low random failure, then a rising wear-out phase.

Infant mortality is manufacturing, not design: a marginal solder joint, a capacitor damaged in handling, a connector not fully seated, a component from a bad reel. These fail early, under load, and they fail fast. That is precisely what a soak test is for.

Why accelerated ageing does not shortcut a lifetime claim

The tempting idea is to run fittings hot for a few hundred hours and use an Arrhenius relationship to convert that into thousands of hours at normal temperature. It works for some failure mechanisms and not for the one people want it for.

The problem is that a luminaire ages by several mechanisms at once, with different activation energies — LED phosphor degradation, encapsulant yellowing, electrolytic capacitor drying, solder fatigue — and they do not accelerate at the same rate. Push temperature high enough to get useful acceleration and you also change which mechanism dominates, so the failure you produce is not the failure that would have occurred in service.

This is exactly why the industry uses LM-80's long, slow measurement at controlled temperatures rather than a short hot test, and why TM-21 caps extrapolation at a multiple of the measured duration. If a supplier offers accelerated ageing as evidence of a lumen maintenance figure, that is not what the method supports — see what a 50,000 hour claim is allowed to mean.

A protocol for buyers without a laboratory

This is deliberately modest. It needs a power meter, a lux meter, a thermocouple and a corner of a room.

1. Sample across the delivery, as for incoming inspection — different cartons and pallet positions, not five units off the top.
2. Record the incoming baseline first. Input power, power factor, relative illuminance on the fixed jig, colour against the retained sample, flicker screening. Without a baseline the soak test proves only that the unit still works, not whether anything changed.
3. Run for 72 hours continuously at rated supply, in a mounting that represents the real installation rather than open air. Twenty-four hours catches the grossest failures; 72 is a reasonable compromise between confidence and schedule.
4. Include the switching cycles. If the installation will be switched frequently, add a timer cycling a subset — an hour on, ten minutes off. Thermal cycling stresses solder joints in a way continuous operation does not, and switching stresses drivers at inrush.
5. Measure the case temperature at the T_c point once stable, and compare against the datasheet limit. A unit running above its declared T_c in the real mounting is a finding regardless of whether it survives.
6. Repeat the baseline measurements at the end and compare. What you are looking for is not the absolute values but the deltas: output down more than a few per cent, power drifted, flicker appeared, colour moved.
7. Log everything, including the units that passed. The value compounds across deliveries — you build a picture of what normal looks like for this supplier, and an outlier batch becomes obvious.

Interpreting the result honestly

A clean 72-hour soak on a sample of units means: this delivery does not have a gross workmanship problem in the units tested. That is genuinely worth knowing and it is the whole claim.

It does not mean the batch will make its rated life, that colour will hold, or that the driver will survive the ceiling void it is going into. Those depend on LM-80 and TM-21 data for the package, a driver life figure at its actual case temperature, and a thermal design validated in situ — all of which are documents to be requested, not tests to be run in a corridor.

What to require of the supplier instead

Your soak test is a check on the delivery. The evidence of design adequacy has to come from the manufacturer:

1. Whether production burn-in is performed, for how long, on every unit or on a sample, and what the reject criteria are.
2. LM-80 data for the package with test duration and case temperatures, and the TM-21 projection with L and B values.
3. In-situ temperature measurement for the finished luminaire in the worst-case mounting.
4. Driver rated life at its actual case temperature, quoted separately from the LED figure.
5. Qualification test reports where the environment demands them — damp heat, thermal cycling, salt spray for coastal installations.