Same wattage, different chip, different result. A tour through SMD package geometry, the binning system that sorts chips at the fab, and how to spec a chip that performs the same in lot one and lot fifty.
Not even close. Two “identical” LEDs can vary 25% in lumens and 200K in colour. The chip and its bin code decide which one you actually got.
Four ideas to carry into the next specification conversation you have.
2835 = small, cheap, ≤ 0.5 W. 3030 = mid-power workhorse. 5050 = three chips in one. COB = high-density, smooth optical source.
Flux (lumens), CCT (white-point), and CRI. A '6-step MacAdam ellipse' is loose; a '2-step' is tight. Lots without binning shift over time.
Two chips of the same bin can have Vf differing by ±5%. The driver must accommodate. A constant-current driver does; a constant-voltage one fights it.
Write 'Lumileds LM301H / Samsung LM301B / Osram Oslon', not '0.2 W LED'. Names lock the binning regime.
Duv is the distance off the blackbody curve. Two 4000K chips, one green and one magenta, both pass a CCT check. Spec Duv near zero on any uniform run.
Before the binning and the packages, there is chemistry. The material stack is the first reason two same-watt chips give different light.
Every white LED starts as a blue one. Thin layers of indium gallium nitride (InGaN) are grown on a sapphire or silicon-carbide wafer by MOCVD epitaxy. When current crosses the InGaN quantum wells they emit blue light near 450 nm. The indium fraction sets the wavelength, more indium shifts it toward green, and crystal quality decides how much current becomes light. Two "identical" 1 W chips with different InGaN quality give different lumens at the same watt.
There is no white junction. White is made by coating the blue chip with a phosphor, classically cerium-doped yttrium aluminium garnet (YAG:Ce), which absorbs part of the blue and re-emits a broad yellow. Leaked blue plus that yellow reads as white. The phosphor recipe is the chemistry that sets CCT and CRI: more phosphor, plus added red nitride or fluoride phosphors (CASN, KSF), pulls the white warmer and lifts CRI. Same blue chip, different phosphor blend, a completely different white.
The deep-red and amber emitters in RGB and tunable fixtures are not InGaN at all, they are aluminium indium gallium phosphide (AlInGaP). It loses output with heat faster than the blue InGaN chips beside it, which is why a poorly matched RGBW module drifts in colour as it warms up. Knowing which chemistry sits behind each colour is how you predict how a fixture will look in year five.
Typical numbers, real values depend on bin and current. Cite the manufacturer datasheet for design.
| Package | Typical drive · efficacy | Use case |
|---|---|---|
| 2835 | 60 mA · ≈ 180 lm/W | Linear bars · panels |
| 3030 | 120 mA · ≈ 170 lm/W | Streetlight modules · highbay |
| 5050 (3-in-1) | 60 mA × 3 · 150 lm/W | RGB · tunable white |
| 3535 (high-power) | 350 mA · ≈ 160 lm/W | Spot · stadium |
| COB (chip-on-board) | 0.5-2 A · ≈ 150 lm/W | Downlight · gallery |
| CSP (chip-scale package) | 1-3 A · ≈ 200 lm/W | Compact spot · flash |
Every chip is measured after production, then dropped into a bucket on three axes. The bin code printed on the datasheet is the name of that bucket, and it is the single most useful line on the page.
A MacAdam ellipse describes how far a chip's measured white point may sit from the nominal CCT before the eye starts to notice. A 2-step match is invisible in practice. A 3-step is the sensible floor for anything a client will see as a continuous run of light. A 5-step is the outer edge of what anyone should call “matched”, and it is visible on a white wall or a plaster ceiling. Write the step count into the specification, because the chip maker will happily ship the loosest bin you did not exclude.
Within one flux bin, lumen output varies by roughly ±5%. Across adjacent bins, ±15% is normal. That is not a manufacturing fault, it is how the yield of a wafer gets sold: the tight, bright bins cost more, the wide bins clear at a discount. If a fixture is priced against one bin and built with the neighbouring one, the lux calculation quietly loses its margin, and nobody finds out until commissioning.
The third axis is colour rendering. A single chip model is typically offered at CRI 70, 80, 90 and sometimes 95, and the higher grades cost efficacy: a CRI 90 chip usually delivers 10 to 20% fewer lumens per watt than the CRI 80 version of the same die, because more of the blue is being converted by red phosphor. Specifying “CRI 90” without adjusting the lumen target is one of the most common ways a design fails its own lux calculation.
CCT moves along the blackbody curve, warm to cool. Duv is the distance off that curve, toward green or magenta. It is the number that explains why two 4000K fixtures on the same ceiling do not match.
Correlated colour temperature tells you where a light source sits along the blackbody locus. It says nothing about how far off the locus the source has landed. Duv is that second coordinate: positive Duv pulls the white toward green, negative Duv toward magenta or pink. Two chips can both certify as 4000K and still be visibly different, because CCT alone cannot describe the difference.
Phosphor thickness and blue-pump wavelength both vary slightly across a production run. A little more phosphor, or a pump die a few nanometres off, and the resulting white point drifts off the locus even though the CCT reading stays inside its bin. This is why a green cast can appear in one production lot of an otherwise perfectly specified fixture.
For any uniform installation, ask for Duv within ±0.003 of the locus and a 3-step MacAdam bin or tighter. On a facade, a lobby, or a long corridor run, that single line prevents the most expensive kind of complaint: a client who can see the difference between two fixtures but cannot name it.
An LED is a current device. Forward voltage drifts chip to chip, so a constant-current driver is what holds the light steady. The watt label on the box is a by-product of the two numbers that actually matter.
Forward-voltage variation between two chips inside the same bin. A constant-current driver absorbs it. A constant-voltage supply turns it into a brightness difference.
Lumen variation between adjacent flux bins of the same chip model. This is the gap that quietly eats a lux calculation when the bin was never named.
Wattage is current multiplied by a forward voltage that drifts. It is the one number on the box, and the least useful of the three.
Five lines you should never sign off on without checking.
Major makers (Lumileds, Samsung, Osram, Cree, Nichia) print a bin code like "LM301H J5", the J5 names the flux × CCT × CRI bucket. Two chips with the same model and different bin codes will look measurably different on the wall.
A MacAdam ellipse describes how far a chip's white point can sit from the nominal CCT. 2-step is invisible to the eye; 5-step is the limit of "matched" for a uniform installation. Always spec 3-step or tighter for premium projects.
Within a single bin, lumen output can vary ±5%. Across adjacent bins, ±15% is normal. If your driver is set for one bin but the chip-maker ships an adjacent one, your lux drops without anyone noticing, until commissioning.
Vf bins (e.g., V1, V2, V3) describe the voltage at the rated current. A multi-chip module that mixes Vf bins burns the lower-Vf chips harder. Reputable assemblers sort to a single Vf bin per fixture.
An LED is a current device. A 1 W chip driven at 350 mA gives different lumens than a 1 W chip driven at 700 mA, even though wattage matches. The datasheet's drive-current curve is the truth, wattage alone lies.
Every fixture we sell lists the chip family and bin code in the BOM. No anonymous LEDs. No silent lot-to-lot shift.
The difference between a line item that can be checked and one that cannot is about fifteen words.
LED, 0.2 W, white
Nothing here can be verified on delivery, and nothing here obliges the supplier to ship the same thing twice.
Samsung LM301B · bin J5 · 3-step MacAdam · LM-80 report attached
Every clause is checkable against a document, and every lot has to match the first one.
A 10-slide explainer, ready to post straight to LinkedIn as an image carousel. Same content, faster reading.
10 slides · LinkedIn-ready. Download all 10 as high-res PNGs and post them straight to LinkedIn as a carousel. PDF version still available.
Practical lighting know-how from NLC Academy. No spam, unsubscribe anytime.
By subscribing you agree to receive NLC emails.