Lifetime is the single most consequential specification a luminaire designer reads off an LED datasheet — and the one most often quoted without evidence. This page publishes the measured lifetime data behind Queendom LED components: the 6,000-hour and 9,000-hour IESNA LM-80 lumen-maintenance runs, the TM-21 projections derived from them, 3,100-hour continuous aging results, and the seven-item reliability qualification matrix. Every figure below is traceable to a signed third-party or in-house test report available on request.
1. What “LED Lifetime” Actually Means
An LED does not burn out; it fades. Lifetime therefore is not a time-to-failure but a time-to-a-threshold. The industry standard threshold is L70 — the point at which maintained luminous flux falls to 70 % of its initial value. L70 was chosen because a 30 % loss is roughly the point at which a luminaire becomes visibly dimmer in a side-by-side comparison. Other thresholds (L90, L80, L50) are used for applications where colour or brightness stability matter more than raw output.
- L70 — lumen maintenance to 70 %. The default for general illumination.
- L90 — used for horticulture and imaging, where a 10 % PPF loss already shifts crop economics.
- TM-21 — the IES projection method that extrapolates a measured LM-80 curve to those thresholds.
- B50 / L50 — median life, where half a population falls below threshold.
Two numbers are routinely confused. Reported lifetime is the figure the certification body will defend as statistically supported by the test duration (usually a conservative floor such as “>36,000 hours”). Calculated lifetime is the raw TM-21 curve fit, which is often far higher. Queendom reports both, and this page shows both.
2. The Three Numbers That Decide Lifetime
Lifetime is set by three physical drivers acting together. Changing any one of them by 10 °C or 20 % of rated current moves the projected lifetime by tens of thousands of hours.
- Junction temperature (Tj) — the dominant driver. Every 10 °C of junction-temperature rise roughly halves the rate constant for phosphor and die degradation, per the Arrhenius relation.
- Drive current (IF) — higher current density accelerates both die and phosphor degradation and shifts the chromaticity coordinate.
- Package materials — silicone vs. epoxy encapsulant, ceramic vs. PPA substrate, and the thermal path resistance Rth J-A set how much of the electrical power actually becomes junction heat.
The design task is therefore a thermal one. Once the fixture engineer knows Tj under real operating conditions, LM-80 data at that case temperature can be read directly off the curve.
The B value shifts where the L70 crossing point falls even when the underlying degradation curve is identical. Two products can share identical L70 hours and still differ in how much of the population sits below the threshold.
3. Method: How Queendom Measures Lifetime
LM-80 is the IES-approved method for measuring lumen maintenance. Its requirements are strict, and they matter because a curve that does not meet them cannot legally feed a TM-21 projection. Samples must come from a minimum of three manufacturing lots, built from different wafer lots on non-consecutive days, and must be picked to represent the full parametric distribution. Each case-temperature set runs at least 25 units. Lumen maintenance is measured at 1,000-hour intervals with the case temperature held at the designated point, ambient airflow minimised, and relative humidity below 65 %.
| LM-80 requirement | Value used by Queendom | Why it matters |
|---|---|---|
| Sample lots | ≥ 3 lots, different wafer dates | Guards against a single lucky wafer |
| Sample size per Ts | 25 pcs | The minimum for a defensible TM-21 fit |
| Case temperature points | 55 / 85 / 105 °C (typical) | Lets the curve be interpolated to real Tj |
| Measurement interval | every 1,000 h | Produces ≥ 6 points for the exponential fit |
| Minimum duration | 6,000 h (LM-80-08); 9,000 h (LM-80-15 runs) | Longer runs give a tighter projection |
| Ambient humidity | < 65 % RH | Excludes moisture-driven artefacts |
| Measurement current | same as life-test current | Avoids current-density discontinuity |
TM-21 then fits an exponential plus a linear term to the measured points: Φ(t) = B·exp(−αt) + C. The fit is only accepted if the data are monotonically decreasing and pass the standard’s goodness checks; the projection itself is capped at six times the test duration. That cap is why a 6,000-hour run can never report a lifetime beyond 36,000 hours, no matter how flat the curve looks — the “>36,000 hours” figure is a reporting limit, not a measurement ceiling.
4. Measured Result 1: SuperFlux Piranha 0.2 W (LM-80, 6,000 h)
A 0.2 W through-hole piranha package (part SU**W309DK, 3000 K nominal CCT) was run to 6,000 hours at three case temperatures by an accredited third-party laboratory. 75 units were tested, 25 per temperature set, with zero failures at every data point.
| Data set | Case temp Ts | Units | Lumen maintenance @ 6,000 h | Δu′v′ @ 6,000 h | Calculated L70 | Reported L70 |
|---|---|---|---|---|---|---|
| Set 1 | 54.4 °C | 25 | 97.47 % | 0.0016 | 65,000 h | > 36,000 h |
| Set 2 | 84.3 °C | 25 | 96.81 % | 0.0020 | 58,000 h | > 36,000 h |
| Set 3 | 104.5 °C | 25 | 96.30 % | 0.0014 | 51,000 h | > 36,000 h |
The separation between curves is the physics made visible. Moving the case from 54 °C to 105 °C — a 50 °C swing — costs only about 1.2 percentage points of lumen maintenance at 6,000 hours, but it pulls the projected L70 from 65,000 hours down to 51,000 hours. That is a 22 % reduction in projected life from thermal design alone.
5. Measured Result 2: HP3535 Ceramic 3 W (LM-80, 9,000 h)
For high-power packages Queendom runs the longer LM-80-15 protocol. An HP3535 ceramic substrate emitter (2700 K, CRI 80, 3 W, driven at 1,000 mA) was aged to 9,000 hours at 85 °C and 105 °C, 25 units per set, again with zero failures.
| Data set | Case temp | Units | Lumen maintenance @ 9,000 h | Δu′v′ @ 9,000 h | α | Reported L70 |
|---|---|---|---|---|---|---|
| Set 1 | > 83 °C | 25 | 98.23 % | 0.0010 | 2.575 × 10−6 | > 54,000 h |
| Set 2 | > 103 °C | 25 | 97.91 % | 0.0011 | 2.808 × 10−6 | > 54,000 h |
| Hours | 0 | 1,000 | 2,000 | 3,000 | 4,000 | 5,000 | 6,000 | 7,000 | 8,000 | 9,000 |
|---|---|---|---|---|---|---|---|---|---|---|
| Set 1 (85 °C) | 100.00 % | 100.15 % | 99.93 % | 99.72 % | 99.50 % | 99.26 % | 99.00 % | 98.75 % | 98.49 % | 98.23 % |
| Set 2 (105 °C) | 100.00 % | 100.10 % | 99.83 % | 99.56 % | 99.29 % | 99.02 % | 98.74 % | 98.47 % | 98.18 % | 97.91 % |
The nearly linear shape of both curves is the signature of a healthy package. A concave curve that bends downward early — a “knee” — indicates phosphor degradation or encapsulant yellowing rather than gradual die wear, and it invalidates a simple exponential fit. Neither HP3535 set shows a knee through 9,000 hours.
6. Measured Result 3: Continuous Aging at Room Temperature
LM-80 is an accelerated, elevated-temperature test. To verify that the projections hold under ordinary operating conditions, Queendom runs independent continuous-aging campaigns at ambient temperature with no thermal acceleration.
A 3535 ceramic warm-white emitter was driven at 700 mA for 3,100 hours of continuous operation at 27–30 °C and 40–45 % RH. Luminous flux was sampled every three days via a 10-hour pulse measurement. No unit failed to light at any point.
| Metric | Minimum | Maximum | Average |
|---|---|---|---|
| Luminous flux decay @ 3,100 h | 2.21 % | 4.93 % | 3.41 % |
| Colour temperature decay @ 3,100 h | 4.43 % | 5.12 % | 5.05 % |
Three thousand one hundred hours is roughly eight months of continuous operation, or about two years of a typical 12-hour-per-day duty cycle. The 3.41 % average flux decay over that period extrapolates to a comfortably sub-L70 loss over a decade of service. Note that the chromaticity drift of 5.05 % exceeds the photometric decay of 3.41 % — the package shifts colour slightly faster than it dims. For applications where colour consistency is critical, this is the number to watch, not the lumen figure.
7. Reliability Qualification Matrix
Lifetime testing answers “how fast does it fade?”. Reliability testing answers the different question “does it survive the environment at all?”. Every Queendom component family passes a seven-item qualification sequence before release.
| # | Test | Condition | Duration / cycles | Samples | Accept criteria | Result |
|---|---|---|---|---|---|---|
| 1 | High-temperature / high-humidity | 85 °C, 85 % RH, IF = 5 mA | 1,000 h | 22 | ΔVF ≤ 10 %, ΔIV ≤ 30 % | PASS |
| 2 | Operating life | Ta = 25 °C, IF = 20 mA | 1,000 h | 22 | ΔVF ≤ 10 %, ΔIV ≤ 30 % | PASS |
| 3 | Reflow soldering | 260 °C peak | 3 passes | 22 | No cracking, ΔVF ≤ 10 % | PASS |
| 4 | Thermal shock | −40 °C / 5 min ↔ 100 °C / 5 min | 300 cycles | 22 | ΔIV ≤ 30 % | PASS |
| 5 | Low-temperature storage | −40 °C | 1,000 h | 22 | ΔVF ≤ 10 % | PASS |
| 6 | High-temperature storage | 100 °C | 1,000 h | 22 | ΔVF ≤ 10 % | PASS |
| 7 | Temperature cycling | −40 °C / 30 min ↔ 100 °C / 30 min | 200 cycles | 22 | ΔVF ≤ 10 % | PASS |
The two worst cases are the high-temperature/high-humidity test (−13.49 %) and the operating-life test (−15.44 %). Both are drift figures, not failures: the acceptance window is ±30 % on intensity, so the margin is roughly twofold. Forward voltage moved less than 2 % in every case, and no unit failed to illuminate.
8. Working Example: Reading a Lifetime Claim
A designer considering a 3 W ceramic emitter drives it at 700 mA in a fixture whose thermal simulation predicts a case temperature of 88 °C. Which lifetime number applies?
- Step 1 — pick the temperature bracket. 88 °C sits between the 85 °C and 105 °C LM-80 sets. Use the 85 °C data set for a conservative estimate, or interpolate.
- Step 2 — check the current. The LM-80 data were taken at 1,000 mA, well above the 700 mA operating current. Because degradation slows at lower current density, the 85 °C curve is a conservative bound.
- Step 3 — read the curve at the operating hours. At 9,000 hours the 85 °C set shows 98.23 % maintenance; extrapolating the TM-21 fit yields L70 > 54,000 hours.
- Step 4 — de-rate for the driver and optics. Real fixtures add driver inefficiency, optical absorption and dirt depreciation. A 20–30 % system-level de-rate on the LED-level figure is standard practice.
- Step 5 — quote the reported, not the calculated, number. The defensible public claim is “> 54,000 hours L70”, not the raw curve-fit result.
9. Why Lifetime Claims Differ Between Suppliers
Two suppliers can quote L70 figures that differ by a factor of three for nominally identical emitters. The causes are systematic, not mysterious:
- Test duration. A 6,000-hour run is capped at 36,000 hours reported L70 by the TM-21 rule; a 9,000-hour run can reach 54,000. Longer tests legitimately unlock larger numbers.
- Case temperature. A supplier quoting a projection based on 55 °C data while the fixture actually runs at 90 °C is overstating life. Always ask which Ts the number came from.
- Reported vs. calculated. Some datasheets silently quote the raw curve fit without the reporting cap.
- Sample provenance. A projection from fewer than three manufacturing lots may simply reflect a good wafer.
- Current derating. Extrapolating a curve taken at 60 mA to a 350 mA application is invalid without a current-derating correction.
Queendom publishes the case temperature, the drive current, the test duration, the sample count and the fit coefficients for every lifetime number quoted, so that the figure can be checked rather than trusted.
10. Reliability Metrics: FIT, MTTF and Confidence
Lifetime describes gradual degradation. Failure rate describes sudden failures. The two are reported on different scales and both belong on a complete reliability datasheet.
- FIT (Failures In Time) — failures per 109 device-hours. A sample of 500 devices run 10,000 hours with two failures accumulates 4,983,000 device-hours and a failure rate of about 400 FIT.
- EFR (Early Failure Rate) — ppm of failures in the first 300–1,000 hours; the “infant mortality” portion of the bathtub curve, and the part that falls inside a typical warranty.
- LFR (Long-term Failure Rate) — the constant-rate portion; this is what underlies useful-life estimates.
- MTTF = 1 / λ — mean time to failure for non-repairable devices such as LEDs.
- Confidence interval — derived from the χ2 distribution. A standard 50-device, 2,000-hour qualification test can only verify failure rates around 9,300 FIT at 60 % confidence; tighter claims require longer tests, larger samples, or accelerated stress.
Accelerated stress testing uses the Arrhenius relation to translate a short elevated-temperature test into a long field-life prediction. The acceleration factor is AF = exp[(Ea/k)·(1/T1 − 1/T2)], with Boltzmann’s constant k = 8.63 × 10−5 eV/K. For optoelectronic devices the industry-standard activation energy is Ea = 0.8 eV — higher than the 0.6–0.7 eV used for silicon ICs, reflecting the additional phosphor and encapsulant degradation paths.
10b. Accelerated Life Testing and the Physics Behind It
Waiting 54,000 hours to confirm a claim is not commercially possible, so the industry accelerates. Raise a stress factor, measure how much faster degradation proceeds, and extrapolate back to service conditions using a physical model.
| Acceleration model | Stress factor | Applies to | Typical acceleration |
|---|---|---|---|
| Arrhenius | Junction temperature | Phosphor aging, defect growth, encapsulant yellowing | Doubling per 10-15 °C rise |
| Eyring | Temperature plus humidity | Moisture-driven corrosion, silver migration | Strongly super-linear in RH |
| Coffin-Manson | Thermal cycle amplitude | Solder-joint and wire-bond fatigue | Power law on ΔT |
| Peck | Humidity at bias | Bias-driven moisture failures | Exponent on RH and 1/T |
| Nordtest / Norris-Landzberg | Cycle frequency plus ΔT | Field-representative cycling | Corrects for dwell time |
11. Factors That Shorten Real-World Lifetime
Laboratory lifetime is an upper bound. The gap between the datasheet and the field is closed by these mechanisms, in rough order of impact:
- Thermal path degradation. Dried-out thermal interface material, or a solder void under the die pad, raises Tj over time and accelerates everything downstream.
- Drive-current overshoot. A constant-current driver with poor transient response can push peak current 20 % above nominal at each switching edge.
- Encapsulant yellowing. Silicone outperforms epoxy by a wide margin; high-flux and short-wavelength (UV, blue) emitters stress the encapsulant hardest.
- Sulphur and chlorine exposure. Common in rubber gaskets and in industrial or agricultural air; causes silver-migration darkening of the reflector cup.
- Moisture ingress. In humid or condensing environments, moisture reaching the die or phosphor causes delamination and flux loss.
- Mechanical overstress at assembly. Exceeding the epoxy glass-transition temperature during soldering can crack the package or break bond wires — see the mounting guidance in the Installation Guides section.
- Optical contamination. Dust and outgassed volatiles on the lens reduce delivered flux without affecting the LED itself.
12. Applying This Data to a Design
The workflow that turns these reports into a defensible design decision:
- Establish the actual Tj under worst-case ambient, using the fixture’s thermal simulation and the LED’s Rth J-A.
- Select the LM-80 data set whose case temperature brackets that Tj. Interpolate only between measured points.
- Confirm the test drive current is at or above your operating current; if below, request a derating curve.
- Read L70 from the TM-21 fit and apply a system-level de-rate of 20–30 % for driver, optics and dirt.
- Cross-check against the reliability matrix for the environmental stressors specific to the installation.
- For horticulture, imaging or colour-critical work, repeat the exercise against L90 rather than L70.
13. Frequently Asked Questions
What is the difference between L70, L80 and L90?
All three are lumen-maintenance thresholds: L70 means the point at which output has fallen to 70 % of initial. L70 is the default for general lighting; L90 is used where colour and PPF stability matter more than absolute brightness, such as horticulture and machine vision.
Why does the datasheet say “> 36,000 hours” when the calculated figure is 65,000?
TM-21 caps the projection at six times the LM-80 test duration. A 6,000-hour test therefore cannot report more than 36,000 hours regardless of how flat the curve is. The calculated value is disclosed separately as the raw fit result.
Can I compare L70 figures from two different suppliers directly?
Only if both were derived at the same case temperature, the same drive current and the same test duration. A 55 °C figure and a 105 °C figure are not comparable, and a figure without a stated Ts should be treated as unusable.
Does LM-80 test the luminaire or the LED?
The LED package or module only. Luminaire-level performance is covered by LM-79 for photometry and by separate system-level lumen-maintenance guidance that adds driver and optical losses.
What case temperatures does Queendom test at?
The standard set is 55 °C, 85 °C and 105 °C. The three-point spread allows the measured curve to be interpolated to a wide range of real fixture temperatures.
How long does an LM-80 run take?
A 6,000-hour run is roughly nine months of continuous operation; a 9,000-hour run is about thirteen months. That is why lifetime data are always published some time after the part is released.
Is a lower L70 always worse?
No. A vendor quoting a low but well-documented L70 at a realistic high case temperature may be more trustworthy than one quoting a high L70 derived from an unrealistically cool test. Always read the test conditions alongside the number.
14. Requesting Test Reports
The full reports behind every figure on this page — LM-80 data sets with per-unit measurements, TM-21 fits with coefficients, the 3,100-hour aging log, and the seven-item reliability reports — are available under NDA or as summary documents depending on the program. Contact Queendom with the specific part number and the case temperature of interest, and we will supply the matching data set along with the laboratory calibration records.