Why 20 percent headroom: load ratio, losses, tc and life in one worked 100 W table, plus the five conditions that tell you when 10 percent is enough and when you need 30.
"Leave twenty percent" has been repeated in this industry for years, rarely with a derivation. It is a chain you can compute: load ratio, then losses, then temperature rise, then tc, then life. Once you run the numbers, headroom stops looking like caution and starts looking like a purchase - of life and of stability.
The one-sentence rule: buy life with headroom first, then adjust for the installation
Normal ventilated installation, not 24 hour duty: leave 20 percent (80 percent load)
All five favourable conditions met: 10 percent is fine
Any one unfavourable condition hit: you need 30 percent
The criterion is not the power percentage, it is the gap between measured tc and the tc ceiling in the spec
Headroom side one: how load ratio becomes tc
The driver dissipates power, and all of it becomes heat.
P_loss = P_out x (1/eff - 1)
DeltaT = P_loss x Rth (case to ambient thermal resistance, degrees C/W)
tc = T_ambient + DeltaT
P_out actual output power (W)
eff efficiency (from the spec; 0.88 if unknown)
Rth case to ambient thermal resistance (degrees C/W),
about 2.5 ventilated, 4 to 6 in a sealed cavity
For a 100 W constant-voltage driver at eff 0.88, Rth 2.5, ambient 40 degrees C (these are worked-example values; substitute your own spec and installation):
Load ratio
Output
Loss
Rise
tc
Relative life (10 degrees C rule)
100%
100 W
13.6 W
34.1 degrees C
74.1
1.0x
90%
90 W
12.3 W
30.7 degrees C
70.7
1.3x
80%
80 W
10.9 W
27.3 degrees C
67.3
1.6x
70%
70 W
9.5 W
23.9 degrees C
63.9
2.0x
60%
60 W
8.2 W
20.5 degrees C
60.5
2.6x
The row that matters is the gap between 80 percent and 100 percent: only 6.8 degrees C apart, about 1.6 times the life. Two metres less strip on the same driver buys you sixty percent more life, not wasted capacity. The 10 degrees C rule is an empirical observation about electrolytic capacitor ageing (life roughly doubles for every 10 degrees C the case is cooled). It is an engineering estimate, not a commitment in any standard.
Each step down in load ratio cuts losses linearly, drops tc, and roughly doubles life by the 10 degrees C rule
Headroom side two: hot-state current climbs on its own
This side is often missed, and on constant voltage it runs the wrong way - strip current rises as the strip warms up.
LED Vf falls with junction temperature, on the order of minus 2 mV per degree C per LED; junction moving from 25 to 85 degrees C (delta 60) drops a single Vf by about 0.12 V
Constant-voltage strip is resistor limited, so a lower Vf means more voltage across the resistor: hot-state current typically runs 5 to 15 percent above cold (use the strip maker's data or your own measurement)
A run you sized at 80 percent can therefore settle at 84 to 92 percent
That 5 to 15 percent is the other half of what headroom has to absorb. On constant current this side is favourable: falling Vf slightly reduces output power and the driver runs easier, at the cost of light output falling with it.
The third side is the input: the same driver dissipates slightly more when the supply runs persistently high (a 220 V system that actually sits at 240 V), and ageing electrolytic capacitors raise ESR and self heating. Neither is easy to compute, so keep about 10 percent of the margin free to absorb them.
When 10 percent is enough (all five)
Burning 4 hours a day or less (showcases, shop windows, trading-hours lighting)
Ambient 25 degrees C or below, ventilated, no direct sun
Driver mounted on a metal surface or the luminaire body, so real Rth is well below 2.5
Normally dimmed to 70 percent or less, because dimming is headroom in itself
One driver per circuit, not several packed side by side
All five, and 10 percent is acceptable - but tc must still be measured and must sit at least 10 degrees C under the spec ceiling.
When you need 30 percent (one is enough)
Sealed cavity: no convection, Rth reaches 4 to 6, tc rises 15 to 25 degrees C at the same load
Ambient 45 degrees C or above: inside light boxes, kitchen ceilings, sunlit metal enclosures, commonly measured at 50 to 65
24 hour duty: car parks, emergency routes, cold rooms - no cooling window, heat accumulates
Several units packed together: the middle one is heated from both sides, another 5 to 10 degrees C; keep 50 mm or more between units
Persistently high or unstable supply: check losses once at the worst-case input voltage
Site checklist (do these and you will know)
Touch the case after 2 hours: comfortable to hold (about 60 degrees C or less) means margin is fine; too hot to hold (above 65) means more margin or a bigger unit
Shoot the tc marked point with an IR gun, not just any face; compare against the spec ceiling and require 10 degrees C or more of gap
Clamp the current: actual load ratio equals measured output power divided by rated power; confirm it is inside 80 percent (or your target)
Re-measure hot: check current again after 30 minutes; 80 percent cold should not climb past 90 percent hot
Estimate life by the 10 degrees C rule: 10 degrees C under the rating roughly doubles life; engineering estimate, not a warranty basis
Constant-voltage units K01, S01 and H01 and constant-current unit R01 all follow this chain. The difference is side two: constant voltage must also absorb the 5 to 15 percent hot-state current climb, constant current does not. Power and current conversion is done in the LED driver sizing calculator; how heat consumes life and where the five heat sources sit is in lifetime, tc temperature and thermal; going the other way, deriving strip metres from an 80 percent load, is in how many lights per driver.
High ambient, sealed cavity, 24 hour duty - should the margin grow or can it shrink? Send us the driver model, rated power, actual load, cavity type and ambient temperature and we will compute the headroom and the tc you should accept. Contact →