A luminaire schedule says "dim to 1%". On the mock-up the slider runs to the bottom, the light is still clearly on, and the last quarter of the travel changes nothing visible. Nothing is broken. The 1% was a statement about current inside one component, and what reaches the room is the product of two separate limits plus a perceptual curve that compresses both.
A bedside reading wall lamp with the globe diffuser and reading head on separate circuits.
One name, two electrical schemes
0-10V began as an analog interface for fluorescent control gear, defined in IEC 60929 Annex E, where the band ran from 1 V at minimum to 10 V at full output and anything below the bottom of the band meant off. The 0-10 V lighting control entry still describes that original behaviour best. LED drivers then adopted the connector and the voltage band but kept their own internal translation, so two things that look identical on a drawing are not interchangeable in the field.
The first distinction is who owns the voltage source.
| Current sinking | Current sourcing | |
|---|---|---|
| 10 V source | the driver | the controller |
| Controller behaves as | a variable resistor to common | an output stage |
| Broken control wire | controller draws nothing, output returns to full | depends on the controller design |
| Passive potentiometer | works | does not work |
| Usual home | architectural and hospitality | theatrical and legacy racks |
In a sinking system the driver puts out a reference through a series resistance and the controller shunts it down, so the control voltage is a divider rather than a command:
Vc = Vs * Rc / (Rc + Rs) That single line explains most commissioning surprises. A controller cannot hold a lower Vc than its own hardware floor allows, and no amount of driver resolution recovers the range underneath it. It is also why the fail-safe direction differs: a cut control wire in a sinking system releases the shunt and returns the fixture to full output, which is safe to see and awkward to explain to a guest.
The floor is a product, not a number
Two floors exist, and the delivered one is the higher of them: the driver's minimum regulated output, and the output the driver produces at the lowest voltage the controller can hold. Neither document states the product, because each of them is describing only its own half.
| Documented item | Usually written as | What it really bounds |
|---|---|---|
| Minimum dim level | percent of rated current | how far the driver will regulate |
| Cutoff threshold | a voltage, sometimes a curve | where output stops rather than dims |
| Controller minimum | a footnote on a datasheet | the bottom of the usable band |
| Dimming curve | a graph, rarely a table | how voltage becomes current |
| Sink current per channel | milliamps | how many drivers one channel can carry |
A driver quoted at 1% minimum is quoting a current ratio against its own rated output. If its curve is linear in current and the controller stops at 0.8 V, the room receives whatever that pair produces, which is often nearer 5% to 10% of flux. Neither datasheet is wrong. They are answers to different questions, and the schedule usually asks a third one, which is what the room will look like at eleven at night.
Why 10% flux does not look like 10%
Light output is roughly linear in drive current, but the eye is not linear in light. Perceived lightness follows a cube-root relationship, so a small change in flux near the bottom of the range is far more visible than the same change at the top.
def lightness(flux_rel): """CIE L* from relative luminance, Y/Yn in 0..1.""" if flux_rel <= 0.008856: return 903.3 * flux_rel return 116 * flux_rel ** (1 / 3) - 16 for f in (1.00, 0.50, 0.10, 0.01): print(f"{f * 100:5.1f}% flux -> L* {lightness(f):5.1f}") Run it and the scale falls away fast: full output reads 100, half reads about 76, 10% reads about 38, and 1% reads about 9. Two consequences follow for anyone specifying a bottom end.
The first is that a 1% floor is not a gentle glow, it is close to dark, and a room asked to use it as a night level will end up with a different fixture or a complaint. The second is that a linear-in-current curve makes the bottom of the travel coarse. One percentage point of output is worth under half a point of L* at the top of the range and close to two points near 10%, which is why the last part of a slider feels steppy while the first part feels inert.
What to ask for instead
Ask for the curve, not the floor. A driver that maps voltage to output on a square-law or logarithmic curve spends its travel where the eye can resolve it, and a 5% floor on that curve reads far lower than 5% on a linear one.
- Request the dimming curve as a table of voltage against measured output, not a marketing percentage.
- Confirm the scheme before wiring: a sinking controller on a sourcing driver will look plausible at full output and fail at the bottom.
- Budget sink current per channel, then re-check it after a fixture count changes.
- Establish that off is off: a limited dimming range needs a switched live or a relay, and a driver left idling is not off.
- Keep the control pair together in the record, because a compatible-looking replacement driver is the usual cause of a second site visit.
Coupling matters as much as curve. The interface carries around a milliamp, so voltage drop along a long run stays small, but the pair behaves like an antenna next to mains cable, and a controller that synthesises its output with a fast switching stage can leave a ripple that some drivers average badly. Screen the run, keep it out of the same conduit as the feed, and test the assembly rather than the parts.
What this means in practice
- Convert every dimming claim into flux at the room before accepting it, and state the curve it assumes.
- Treat the controller and the driver as one component; specify them together or not at all.
- Measure the bottom of the range on the mock-up, at night, at the level the room will actually use.
- Record the paired part numbers with the curve, so the next replacement is a decision rather than an experiment.
If the low end is where a project keeps failing, the sequence in our notes on low-end dimming compatibility tests is a better starting point than another driver swap.
Sonny Yu — Zerouno Lighting. We make alabaster, brass and wood fixtures for guest rooms and refurbishment projects.

