SDCM lighting specifications are usually written in one line: 3-step, single batch, spares from the same code. What the number counts is how far each fixture’s actual colour point sits from its nominal colour temperature, measured in MacAdam ellipses. A 3-step bin is roughly 0.004 in CIE 1976 u′v′; a 5-step bin is roughly 0.007. Both sound small. On a 12 m ceiling with 40 downlights in a row, the difference is a clean line of light versus two panels that clearly read greener than their neighbours.
1. What an SDCM step actually means
The unit comes from David MacAdam’s 1942 work on colour discrimination. He mapped, for a given base colour, the region of chromaticity that an observer cannot distinguish from it. That region is an ellipse, not a circle, and its size depends on where you are in colour space. One step, one SDCM, is one such ellipse.
The practical definition used in lighting is looser than the science. A 3-step bin is a region three ellipses wide around the nominal point, and a 5-step bin five ellipses wide. Suppliers draw those regions as quadrangles in CIE 1960 uv space, which is why an ANSI bin is a polygon rather than a true ellipse. The number on the datasheet is an approximation of a tolerance, and it is worth remembering that when two suppliers both claim 3-step.
Converted to the modern CIE 1976 u′v′ space that colour work now uses, the numbers are these:
| 色度ビン | Approximate Δu′v′ from nominal | Visible as a difference? |
|---|---|---|
| 1-step | ≈ 0.0015 | Not to the eye, even side by side |
| 2-step | ≈ 0.003 | Only in a direct A/B comparison |
| 3-step | ≈ 0.004 | Borderline in a continuous run, acceptable in a room |
| 5-step | ≈ 0.007 | Visible on a long ceiling, obvious on a wall wash |
| 7-step | ≈ 0.010 | Clearly different colours |
One caution on the numbers: they describe distance from the nominal point, not from each other. Two fixtures at opposite edges of a 5-step bin can be 0.014 apart, which is a visible mismatch even though every unit is “within tolerance”. That is the core problem with binning as a quality claim, and the reason batch control matters more than the bin label.
2. 3-step vs 5-step in a real room
In a single fixture on a single wall, nobody can tell a 3-step product from a 5-step product. In an array, everyone can. The eye is very good at detecting a step change between adjacent surfaces, and very poor at judging absolute colour. That asymmetry is what makes consistency a separate specification from accuracy.

| 用途 | Maximum bin | Why |
|---|---|---|
| Feature wall, art, retail display | 2-step | Adjacent fixtures at close range; any step reads as a defect |
| Guestroom, restaurant ceiling | 3-step | Fixtures spaced 1.2–2 m; steps visible only if extreme |
| Corridor in a continuous run | 3-step, single batch | Long sightlines expose edge-of-bin variation |
| Wall wash and cove | 3-step | Grazing light amplifies small differences |
| Back of house, plant, car park | 5-step | Nobody compares surfaces at close range |
The corridor case deserves its own paragraph. A 30 m corridor with downlights every 1.5 m gives the eye a repeating reference and a long viewing distance. Under those conditions a 5-step bin produces a visible beat: two or three units in the run read slightly green, the rest read neutral, and the pattern looks like dirt on the ceiling. Cleaning does not help. Re-lamping does.
Wall washing is the other amplifier. When light grazes a plaster surface at a shallow angle, small chromaticity differences show up as vertical bands. A 3-step bin is usually fine; a 5-step bin in the same detail will show banding that no amount of aiming fixes.
3. Batch codes and binning
An LED bin code is not one number. A typical code carries the colour bin, the flux bin and the forward voltage bin, often with a date or batch reference appended. Two fixtures can share a colour bin and differ in flux by 10%, which shows up as a brightness step in a run even though the colour is consistent.

| Element | What it controls | Failure if unmatched |
|---|---|---|
| Colour bin (CCT + SDCM) | Chromaticity | Green or pink cast on individual units |
| Flux bin | Luminous output | Visible brightness steps in a run |
| Voltage bin | Forward voltage | Driver loading, efficiency, occasional drop-out |
| Batch / date code | Production run | Drift between phases of a refurbishment |
Batch matters because phosphor deposition changes slightly between production runs. A supplier can hold a 3-step colour bin across a year and still deliver two batches that differ by 2–3 steps at the edges. That is within the bin and still visible on site, which is why the PO clause in section 6 asks for a single batch rather than just a bin.
Ask for the batch code on the fixture label, not only on the carton. Cartons get separated from fixtures during installation, and a maintenance team replacing a failed unit three years later needs the code on the hardware to order a match. This is the same argument as keeping matched spares, covered in spare parts for hotel lighting.
4. Mixing suppliers and batches
Mixing suppliers in one continuous ceiling is the most reliable way to produce a visible mismatch, even when both suppliers meet the same 3-step clause. The reason is the shape of the bin. Two 3-step polygons drawn around slightly different nominal points can overlap by less than half their area, so a fixture at the edge of one is four steps from a fixture at the edge of the other.
The rules that avoid it are short:
- One supplier per continuous ceiling plane, without exception.
- One batch per plane; if the quantity spans batches, split at a physical break such as a column line or a door.
- Order the full quantity in one release, plus 5% spares from the same batch.
- For phased handovers, order the later phase against the original batch code, not against the current catalogue.
Refurbishments are where this becomes expensive. A 2019 batch and a 2024 batch of the same SKU can sit two steps apart, which is invisible in a showroom and obvious where the two meet at the end of a corridor. Where the original batch is unavailable, the honest options are to re-lamp the whole plane or to place the join behind a physical break. Neither is cheap, and both are avoidable with a spares order at first delivery.
There is a second-order effect worth knowing. Ageing shifts chromaticity, and different batches age at slightly different rates. Fixtures that match at handover can drift apart over 10,000 hours if the thermal design differs between the two production runs. On a long-life project, matching the batch at installation is the only control you have.
5. Testing consistency on site
Consistency is measurable in half an hour with a handheld spectroradiometer, and it is worth doing before snagging rather than after. The method is straightforward.
Measure at least one fixture in twenty, at 1 m from the emitting surface, at full output, after 30 minutes of stabilisation. Record CCT and the u′v′ coordinates rather than just the CCT. Compute the Δu′v′ between adjacent fixtures in a run — the number that matters is not the spread against nominal but the step between neighbours.
| Check | Pass criterion | Action if it fails |
|---|---|---|
| Adjacent fixtures in a run | Δu′v′ ≤ 0.003 | Re-lamp the odd units from the same batch |
| Run against nominal CCT | Within 150 K of specified | Check against submitted bin code; reject if outside |
| Fixture against its own report | Δu′v′ ≤ 0.002 | Batch does not match the submitted sample |
| After 1 hour at full output | No measurable shift | Thermal design issue; check heatsink and void temperature |
The one-hour check catches a specific failure that the datasheet never mentions: colour shift as the fixture heats up. A unit that starts neutral and drifts green after 40 minutes will pass a cold measurement and fail in service. Measure cold, then measure again when the fixture has been running for an hour.
Visual checks still matter, and they are free. Stand at one end of the corridor at night with only that circuit on, and look along the run. Any unit that reads differently from its neighbours will show up in that view even if it passes a number. Photograph it with a fixed white balance, because the camera exaggerates what the eye tolerates.
6. Writing SDCM into the PO
The specification sets the requirement; the purchase order makes it enforceable. Five clauses cover the failure modes described above.
| Clause | Wording to use |
|---|---|
| Colour tolerance | All luminaires to be within 3-step MacAdam (Δu′v′ ≤ 0.004) of the specified CCT. |
| Single batch | Each continuous ceiling plane to be supplied from one production batch; batch code to be stated on the delivery note and on the fixture label. |
| Flux matching | Flux bin tolerance within 5% across each order line. |
| Spares | 5% of quantity supplied as spares from the same batch, labelled with the batch code. |
| Remedy | Fixtures failing the on-site check to be replaced from a matching batch at the supplier’s cost. |
Add the evidence requirement that makes the clauses testable: an LM-79 report per SKU, less than 24 months old, measured on the complete luminaire, plus the ANSI bin code on the submittal. A supplier who can produce both is a supplier who tracks production; one who cannot is guessing.
Finally, write the acceptance test into the programme. Consistency checks belong before practical completion, when replacement stock is still available and access equipment is on site, not three months later when the scaffolding has gone. The full set of metrics behind these numbers is in the light quality guide.
次のステップ
Batch consistency is a production question, not a catalogue question, so the useful starting point is the quantity per ceiling plane and the handover programme. Send the fixture schedule with quantities and phases, and the team will confirm which items can be supplied from a single batch, what the lead time looks like, and how the spares should be split across phases. Start at the capabilities page.






