How Do Led Lights Work

LED Lighting Explained Without the Fluff

An LED is a light-emitting diode. It's a semiconductor device that converts electrical current directly into photons. No filaments, no gas, no mercury vapor. That's basically it. But there's enough nuance in how they actually function that people get it wrong constantly. When you run current through the p-n junction of an LED's semiconductor material, electrons drop from the conduction band into holes in the valence band. That energy difference releases as a photon. The wavelength, and therefore the color, is determined by the band gap energy of the semiconductor material. Gallium nitride gives you blue. Indium gallium phosphide gives you red or amber. Mix them with a phosphor coating and you get white.

How Do Led Lights Work

The phosphor conversion method is what almost all "white" LEDs use. You've got a blue LED die, usually aluminum gallium indium phosphide based, coated in a yellow-emitting phosphor—typically cerium-doped yttrium aluminum garnet, or YAG:Ce. The blue light excites the phosphor, which re-emits some of that energy as broader yellow-green wavelengths. The mix of residual blue and down-converted yellow light reaches your eye as white. Different phosphor blends give different color temperatures. Add a red-emitting phosphor and you push the CCT down to something warmer like 2700K. Take more of it out and you're into 5000K+ territory. Here's the thing most people don't realize: LED light output degrades over time, and it's not linear. Manufacturers will quote L70 ratings—the point at which the fixture outputs 70% of its initial lumens. But that doesn't mean the light suddenly dies at that threshold. It just means you've hit the industry standard for "end of useful life." An LED can still be producing light at 50% output after hundreds of thousands of hours, it just won't be very bright anymore. And if the thermal management is poor, you'll hit that L70 point much faster than the datasheet claims. I once sourced a batch of 12W replacement LED bulbs from a budget manufacturer. The lumen output on the spec sheet looked competitive. I pulled three from the lot and ran them in an integrating sphere setup. Two of them were dead within 600 hours—actual failure, not just dimming. The third was still on but at roughly 40% of rated output after 2000 hours. The issue was cheap epoxy underfill that couldn't handle the thermal cycling. The solder joints on the driver board cracked, and the current regulation drifted. I ended up switching to a different supplier who used underfill compound with a higher Tg rating and matched copper PCB traces. Cost went up maybe 18%, but failure rates dropped to under 1% over 5000 hours. Not worth chasing the cheapest option on this one.

What Actually Determines Efficiency

Wall-plug efficiency in LEDs is measured as lumens per watt, but that number is nearly useless without context. It depends heavily on drive current. An LED pushed at its maximum rated current will be less efficient than the same LED run at half current. The extra heat generated at high drive levels increases non-radiative recombination—phonon losses instead of photon output. This is why dimming an LED often makes it more efficient per watt, even though total light output drops. A typical LED might peak at 150 lm/W at 350mA and drop to around 110 lm/W at 700mA. The driver circuit matters enormously too. Cheap constant-voltage drivers that just drop resistance across a series resistor are brutally wasteful. A proper constant-current driver with power factor correction will actually deliver more usable light per watt drawn from the mains because it maintains optimal current regardless of line voltage fluctuations. I've seen installations where swapping a $4 resistor-based driver for a switching constant-current unit cut the total energy draw by 30% with the same light output. Not because the LED changed, but because it was finally being driven correctly. Thermal management is where most LED failures happen, and it's also where manufacturers cut corners. The junction temperature of an LED die is the single most important factor in longevity and performance. Every 10°C rise above the rated junction temperature can halve the expected lifetime. The aluminum PCB substrate in a good LED panel might cost $2 to $3 more than the FR4 version, but it drops thermal resistance from around 8°C/W down to 3°C/W. That difference determines whether your fixture lasts 20,000 hours or 50,000.

Common Pitfalls When Specifying LEDs

Color consistency across multiple fixtures is a real headache if you don't understand binning. LED manufacturers sort their dies into color bins based on chromaticity coordinates and forward voltage. Two LEDs from the same batch labeled "3000K" can have noticeably different color if they come from different bins. MACI—Maximum Coefficient of Variation of Illuminance—values matter here. A good manufacturer guarantees a 3-SDCM bin spread, which is imperceptible to the human eye. Cheap products might mix 5 or 6 SDCM spreads, and you'll walk into a room and see obvious color mismatches between fixtures even though they're all rated the same temperature. Another issue nobody warns you about: capacitive coupling in dimming circuits. If you're retrofitting LEDs into an existing switched circuit with a nearby motor load—like an exhaust fan or HVAC system—you can get ghosting. The LED will glow faintly when it should be off. This happens because the LED's input capacitance picks up stray electromagnetic fields through the wiring. The fix is usually a small bleed resistor across the LED terminals, something like 100K ohms at 1 watt. I've also seen people solve it by moving the switched line to the hot conductor only and keeping neutral continuous, but that's not always practical in existing installations. UV and blue light exposure is another area where marketing outpaces science. The LED blue light hazard standard (IEC/EN 62471) exists for a reason. Some high-CRI white LEDs have a significant spike in the 440-450nm range from the blue pump diode. For residential use this is generally fine, but in environments where people spend prolonged periods directly under these fixtures—operating rooms, inspection stations, workshop benches—the cumulative exposure can be a concern. I always check the photobiological safety rating before specifying LEDs for any space where someone would be working directly underneath them for hours.

When LEDs Just Don't Work

There are legitimate cases where LEDs are the wrong tool. High-intensity discharge lamps still outperform them in terms of raw lumen output per fixture in applications like stadium lighting, high-bay industrial spaces over 30 feet, and agricultural grow operations. An LED array producing 200,000 lumens will cost significantly more upfront than a 1000W metal halide setup producing roughly the same, and the heat sinking required for that kind of output becomes a serious engineering problem. Emergency egress lighting is another area where incandescent and HID still have a place. LEDs have a turn-on time measured in milliseconds, which sounds great, but some jurisdictions require a specific warm-up period for verification, and older building codes written for instant-start fixtures haven't always caught up. More importantly, a failing LED driver can cause partial illumination—enough to confuse someone reading an exit sign in a panic. Incandescent bulbs either work or they don't. With LEDs, you can get that weird flickering phase where the driver is dying but the light still comes on. I've seen this in fire-rated exit sign replacements where the LED version passed inspection initially but failed two years later with only marginal output. If you need UV or infrared illumination—curing resins, plant pathology work, night vision illumination—standard white LEDs aren't going to cut it. You need specially engineered narrow-band LEDs or laser diodes for that. The band gap material has to be completely different. Gallium arsenide for IR, aluminum indium phosphide for deep UV. They're not cheap and they have very different thermal requirements than visible-spectrum devices.