Inside the LED: what you are actually specifying
Blue die plus phosphor, and everything that follows from it — COB against mid-power, junction temperature, and why a lumen figure without a drive current means nothing.
Chip xanh lam cộng lớp phosphor, và mọi hệ quả kéo theo — COB so với mid-power, nhiệt độ mối nối, và vì sao con số lumen không kèm dòng điện là vô nghĩa.
Almost every white LED in architecture is the same trick: a die that emits blue at around 450 nm, covered in a phosphor that absorbs some of that blue and re-emits it broadly across yellow, green and red. What reaches the eye is the leftover blue plus the phosphor’s re-emission, and the ratio between the two is what you buy when you specify a colour temperature.
Once you know that, most of the datasheet stops being arbitrary.
Why the spectrum has that shape
The blue spike is the die showing through. The broad hump is the phosphor. The dip between them — usually around 480 nm — is where neither contributes much, and it is a direct consequence of the method rather than a flaw in a particular product.
That dip matters twice over. It sits close to the melanopsin peak at 490 nm, so two sources at the same CCT can differ in circadian effect more than their colour temperature suggests. And a thin red tail is why R9 is so often the weakest number on the sheet: adding red phosphor costs efficacy, so it is the first thing value-engineered out.
More phosphor means warmer light and lower efficacy. A 2700 K product is typically 10–20% less efficient than the same platform at 4000 K, for exactly this reason. That is not a defect to argue about with a supplier — it is physics, and it belongs in the energy calculation from the start.
The three packages
Mid-power. Small plastic-bodied emitters, dozens or hundreds on a board. Cheap, efficient and the basis of almost all panels, battens and linear fittings. Each emitter is a separate small source, so a clear lens over an array gives you a grid of bright dots — which is why these fittings nearly always need a diffuser, and why they look terrible when the diffuser is removed to gain output.
High-power. Single ceramic-based emitters, roughly 1–3 W each, each with its own optic. Chosen when you need genuine beam control: spotlights, projectors, exterior. More expensive per lumen, but you can actually put the light where you want it.
COB (chip-on-board). Many dies bonded directly to a substrate under one shared phosphor layer, presenting a single small bright surface. The closest thing to a point source, so it takes a reflector cleanly and gives crisp shadows — good for accent work, and correspondingly harsh if that is not what you wanted.
The package choice is really a choice about source size, and source size determines what optics can do. A small source can be collimated into a narrow beam; a large one cannot, no matter what reflector you put in front of it.
Binning, and the number that actually matters
Manufacturing spreads output across flux, forward voltage and chromaticity, so emitters are sorted into bins. Flux binning affects consistency of level; chromaticity binning affects whether a run of fittings looks like one product.
The specification number is SDCM — MacAdam steps. Ask for ≤3 SDCM on anything where fittings are seen together, and ask for it maintained, not just initial. A tight initial bin that drifts apart over two years is a warranty argument you will lose without that word in the specification.
Heat is the whole story
An LED converts roughly a third of its input to light and the rest to heat, and unlike a lamp it cannot radiate that heat away — it must be conducted out through the board and the heat sink.
Junction temperature (Tj) is the temperature at the die itself, and it governs everything:
- Output falls as Tj rises. A hot LED is a dim LED, immediately and reversibly.
- Colour shifts, usually toward green in phosphor-converted white.
- Lifetime collapses. The relationship is roughly exponential; 10 °C hotter can halve it.
Which is why a lumen figure quoted without a drive current and a case temperature tells you nothing. The same emitter driven at 350 mA and 700 mA gives very different output, efficacy and life, and manufacturers quote whichever combination flatters the product.
What to ask for: LM-80 test data with a TM-21 projection, at the drive current and case temperature the fitting actually runs at. Not the emitter manufacturer’s bench figures. If a supplier cannot produce this, they do not know their own product’s lifetime.
Reading a lumen claim
Three different numbers get called “lumens”, and suppliers move between them freely:
| What it is | Roughly | |
|---|---|---|
| Emitter lumens | The bare LED on a bench at 25 °C | 100% |
| Module lumens | On its board, at working temperature | 85–92% |
| Luminaire lumens | Out of the fitting, after optics, lens and driver losses | 65–85% |
Only the third is real. Ask for delivered luminaire lumens and luminaire efficacy in lm/W, tested per LM-79, and treat any chip-level efficacy claim as marketing.
What this changes about specifying
Write the drive current into the schedule. It is the single fact that connects the lumen claim, the efficacy, the lifetime and the warranty. Without it every other number is unanchored.
Ask how the fitting sheds heat. A slim, sealed, plastic-bodied downlight in an insulated ceiling void is a thermal problem regardless of the brand on it. This is where cheap fittings actually fail — not in the LED, but in the path the heat was supposed to take.
Treat efficacy and colour quality as a trade you are making deliberately. Ra 90 with a good R9 costs roughly 10–15% of efficacy against Ra 80. That is a decision about the building, not a technical detail to be delegated to the contractor.
Assume the driver dies first. It is the shortest-lived component in the assembly, and whether it can be replaced without dismantling the fitting is a maintenance decision made at concept stage.