LED Light Engine Buying Guide: COB vs SMD vs Filament, and the Binning Specs That Decide Color Consistency

The Light Engine Is the Fixture

Strip away the aluminum housing, the diffuser, the paint job, and what's left is a light engine: a board of LED dies, a phosphor layer, and a way to get heat out. That engine decides the lumen output, the color, the beam, and most of the lifetime. The housing just keeps the dust off.

So when two suppliers quote the same 12W downlight and one is 40% cheaper, the difference is almost always in the engine. Cheaper dies, a looser bin, a smaller heat sink. The fixture still lights up on day one. Six months later, it's pink, dim, and dead.

COB, SMD, and Filament: The Actual Difference

COB, chip-on-board, packs a dense grid of dies onto one substrate and covers them with a single phosphor layer. The result is one smooth light source, no visible dots, soft shadows. That's why you see COB in downlights, track heads, and anything where the beam has to look clean. It costs more per lumen, but the light quality is better and it's easier to control optically.

SMD, surface-mounted device, scatters individual diodes across a board. Each die has its own little dome, and there are gaps between them. For linear lights, panels, and strips, that's fine, the diffuser hides the dots. SMD wins on cost per lumen and flexibility, which is why it dominates high-volume fixtures.

Filament engines string tiny dies along a transparent strip so the whole thing glows like a wire. It's a look, not a performance choice. Filament bulbs trade efficiency and sometimes durability for the incandescent aesthetic. Buy them for restaurants and decorative pendants, not for the warehouse.

Notice none of these is "better" in the abstract. A COB in a warehouse high bay is wasted money. An SMD in a museum spotlight looks cheap. Match the engine to the beam you need, and the rest is cost.

The Binning Problem Nobody Puts in the Brochure

Two LEDs from the same production line are never identical. So manufacturers sort them after the fact into bins, grouped by color temperature, flux, and forward voltage. The width of that bin is what separates a fixture that looks consistent from one that doesn't.

The number to ask for is the MacAdam ellipse step. A 3-step ellipse is tight; you can line up a hundred fixtures and not see a difference. Five steps is looser but often acceptable for general lighting. Seven steps means visible patchwork on the ceiling. And the classic supplier trick is quoting a CCT like "3000K" while shipping bins spread from 2700K to 3300K, because all of them technically sit inside that label.

Also demand the R9 value alongside CRI Ra. Ra averages the first eight test colors and ignores deep red. A fixture can score Ra 90 while rendering red meat, red paint, and skin tones badly, because its R9 is 20. For retail and hospitality, R9 below 50 is a visible defect the customer notices before you do.

Reading the Thermal Section Like a Buyer

Everything about LED lifetime comes back to temperature. The junction temperature, the hottest point inside the die, decides how fast the phosphor degrades and how much light you lose. Suppliers quote 50,000 hours, but that number only means something when it's tied to a test condition.

Ask for the LM-80 report, which measures lumen maintenance at a stated case temperature, and the TM-21 projection that turns it into an L70 lifetime. The honest ones test at 85 degrees C, because that's where a real fixture runs. An engine rated at 55 degrees C on a bench is a fantasy number if your housing lets the junction run hotter.

The heat sink is part of the engine spec, not an afterthought. A 50,000-hour engine mounted on a heat sink sized for half its heat load will still fail early. When you compare quotes, compare the full thermal path, not just the LED brand.

Common Questions from Buyers