An LED lamp rarely stops working. It gets dimmer. Lumen maintenance describes that slow decline: the ratio between output at a stated number of operating hours and the output measured in the first hundred hours. Once that ratio crosses a stated threshold, the product has reached the end of its rated life, even though the driver still energises the board. Most procurement language in this market is built around one figure, B50 L70, and it is worth understanding what the two numbers mean before signing.
The Shape of the Curve
Output starts slightly above its nominal value at the end of the first hundred hours, then declines as the phosphor converts less efficiently and the encapsulant darkens with temperature. The curve is steep in the first 1000 hours, flattens between 1000 and 10,000 hours, and turns downward again as solder joints fatigue. A useful mental model is three phases: a small initial drop, a long near-linear slope, and an end-of-life knee. The knee is where thermal resistance rises and the junction temperature climbs, which accelerates the very process that caused the rise in the first place.
B50, B10, L70, L90
The B number is a percentile of a population. B50 L70 means half of a batch of units still delivers at least 70 percent of initial light at 25,000 hours. B10 L70 is stricter: nine out of ten units are still above 70 percent at that hour. The L number is the maintenance threshold itself, written as a percentage of initial output. The two are often read together with an hour figure, and the hour figure is the part that changes fastest between product classes.
| Term | Meaning | Typical use |
|---|---|---|
| Initial lumens | Output after the first 100 hours | Marketing headline only |
| L70 | Output falls to 70 percent of initial | Most common end-of-life point |
| L80 | Output falls to 80 percent of initial | Street and facade lighting |
| L90 | Output falls to 90 percent of initial | Museum, retail, hospitality |
| B50 | Half the population is still above | Default statistical claim |
| B10 | Nine tenths are still above | High-reliability specification |
Reading LM-80 and TM-21
LM-80 is the test method. A sample of devices runs on a chamber at one or more case temperatures for at least 6000 hours, and luminous flux is tracked over time. TM-21 is the reporting extrapolation method that turns those measured hours into a projected maintenance curve. Three rules follow from that structure. First, the extrapolated figure can never exceed the tested duration, so a claim of 60,000 hours from a 6000-hour test is allowed but a claim beyond it is not. Second, a single temperature cannot describe a fixture that will see -20°C outdoors and 55°C indoors. Third, the sample size and the spread between samples matter as much as the average.
What Shortens the Curve
Current density is the main lever. Running a chip at 70 percent of its rated current typically cuts junction temperature by 10°C to 20°C and extends the L70 point by thousands of hours. Thermal path is the second: a poor solder joint, a thin dielectric, or a driver that pushes 20 mA more than the datasheet allows all shift the same curve downward. Encapsulant quality is the third. Yellowing from heat, humidity, or sulfur in the air removes blue flux before it can be converted, so the package loses output without any drop in electrical power.
Key Takeaways
Is B50 L70 the same as product life? It is the hour at which half the population reaches 70 percent output. The last unit usually fails much later.
Why do two fixtures with the same chip age differently? The driver current and the thermal path decide junction temperature, and junction temperature decides the slope.
What should a specification ask for? A rated hour, a maintenance threshold, a test standard reference, and an operating temperature range, not just a wattage.
