Both fixtures accept the same instruction, a number between 0 and 255, and both get dimmer as the number falls. Everything else about the two processes is different, and the differences show up in exactly the places a designer cares about: the bottom of a slow fade, the colour at half, and what a camera makes of it.
A filament is a thermal device pretending to be an electrical one
A conventional dimmer does not reduce voltage smoothly. It chops the mains waveform, conducting for part of each half cycle and blocking the rest, so what arrives at the lamp is a series of fragments. The lamp does not care, because a tungsten filament is a lump of hot metal with thermal inertia. It cannot cool appreciably between fragments, so it integrates them into steady light.
That inertia produces the behaviour everyone who has worked with tungsten knows. Response is not instant: there is a perceptible lag at the top of a snap and a lingering glow at the bottom of a cut. A fade to zero does not end at zero, it ends in a decaying ember, and designers have used that ember as an effect for a century.
Dimming a filament is also changing its colour
Incandescent light is thermal radiation, so its spectrum is determined by temperature. Reduce the power and the filament runs cooler, and cooler means redder. A lamp at full might sit around 3200 K; taken down to a low level it drifts several hundred kelvin warmer in appearance, visibly amber.
This is why tungsten dimming reads as emotional rather than merely quantitative. A dim tungsten stage is not just a darker stage, it is a warmer one, and the association with candlelight and dusk is doing work that a designer did not have to ask for. Fixed colour LED fixtures do not do this at all unless they have been built to imitate it, which is why so many manufacturers now offer an emulation mode.
A diode has no memory of the last instant
LEDs respond to current essentially instantaneously and store no heat that turns back into light. Two dimming methods dominate. Pulse width modulation switches the diode fully on and fully off very rapidly and varies the proportion of on time, which holds the colour constant because the diode is always run at the same current. Constant current reduction lowers the drive current instead, which avoids the pulsing but allows a small colour shift as the operating point moves.
Because nothing averages the pulses, the modulation is present in the light itself. The eye usually does not see it; a camera shutter frequently does, and that interaction has its own article in this section.
Why slow LED fades step and slow tungsten fades do not
A single byte of control divides the range into 256 values. On a tungsten fixture the filament's thermal lag smears the boundary between one value and the next, and a ten second fade looks continuous. On an LED there is nothing to smear it: each value change is a discrete jump in output, and near the bottom of the curve, where each step is a large proportion of the remaining light, those jumps become visible as stepping.
Two things address it. Sixteen bit dimming allocates a second control slot to fine resolution, which is the same trade described in the article on DMX universes: smoothness costs slots. And the dimming curve itself matters more than it used to, because perception of brightness is closer to logarithmic than linear, so an evenly spaced set of numbers does not produce an evenly perceived fade.
Fade time is not the same as fade shape
A console asks for a duration, and the duration is the least interesting part of the instruction. What the audience reads is the shape: whether the light leaves evenly, holds and then drops, or eases out at the end. On tungsten a substantial part of that shape was supplied for free by the filament, which is why the same numeric fade time transplanted to an LED rig so often feels abrupt in the last second.
Rebuilding the shape deliberately, with a curve chosen at the fixture or a profile applied at the console, is ordinary work rather than an advanced technique, and it is usually the difference between a rig that feels harsh and one that does not.
The bottom of the curve, and the failure nobody expects
Many LED drivers have a minimum stable output below which they cannot regulate, so the last part of a fade can drop out abruptly instead of arriving at black. The cure is a fixture whose driver was designed for theatrical fades rather than for architectural use, and this is one of the few specifications worth interrogating before purchase.
A related trap belongs to retrofits. Putting an LED lamp into a circuit fed by a phase control dimmer asks a switched mode power supply to work from a chopped waveform it was not designed for. Flicker, buzz and refusal to strike are the usual results, and the fault is the combination rather than either component.
What we cannot verify
Modulation frequencies, minimum dimming levels and curve behaviour are published by fixture manufacturers, usually without stating the measurement method, the drive level or the operating temperature, and LED behaviour changes with all three. We reproduce none of those figures here. The relationships described are physical; how a particular fixture implements them is a question for a measured comparison in the room, with the camera and the console that will be used.
The short version
- A phase dimmer chops the waveform; the filament's heat averages it back.
- Thermal lag gives tungsten its slow top, its ember, and its forgiving fades.
- Dimming a filament lowers its colour temperature, so dimmer also means warmer.
- PWM holds LED colour constant; current reduction avoids pulsing but shifts colour.
- Without thermal smoothing, eight bit fades step visibly near the bottom.
- Driver minimum output and phase dimmer retrofits cause most dropout complaints.