Desk
Desk lighting: the spec that matters isn't brightness
Lumens tell you almost nothing about a desk lamp. Lux at the surface, CRI and R9, colour temperature, flicker and glare tell you the rest.
Desk lamps are sold on one number, and it’s the wrong one. A listing leads with 1,200 lumens and says nothing about how far above the desk that light sits, what colours it renders, or whether it pulses when dimmed. Those are the things you notice after an hour of reading.
Lumens describe the lamp; lux describes your desk
A lumen is total light output in every direction. Lux is what lands on a surface: one lumen spread over one square metre. Only the second one is about you.
The gap between them is geometry. Illuminance from a small source falls off with the square of distance, so doubling the height of a lamp head above the desk cuts light on the page to roughly a quarter. A 1,200-lumen head on a tall fixed post can deliver less usable light than a 500-lumen head sitting close and aimed properly.
Office standards put reading and typing work near 500 lux at the desk, higher for fine detail. If a lamp publishes a lux figure it should come with a distance, usually 30 or 40 cm directly under the head. A lux number without a stated distance is not a specification. The conclusion is unglamorous: an arm that puts light where you want it is worth more than raw output.
CRI is the average, R9 is the one they leave out
Colour Rendering Index compares a source against a reference across a set of test colours. The headline Ra figure averages the first eight, all of them muted pastels. Saturated colours are not in that average.
R9 is deep red, and it sits outside Ra. Red is where cheap white LEDs are weakest: a blue-pump LED with a thin red phosphor can score a respectable Ra in the low 80s while rendering skin, wood, food and warm ink as flat and slightly grey.
So the question is not “is CRI above 90” but “is R9 published, and is it positive”. Makers with a good R9 tend to print it, and silence is informative. Newer datasheets sometimes use TM-30 instead, reporting Rf for fidelity and Rg for saturation, which signals a manufacturer who measured rather than estimated.
Kelvin, and why the same lamp is wrong at 9pm
Correlated colour temperature describes the tint of white light: around 2700K is warm, 4000K neutral, 5000K and above daylight-like. There is no single correct value, because the requirement changes across the day. Blue-rich light is more alerting, since the retinal receptors that drive circadian timing are most sensitive to blue wavelengths. A bright 6500K lamp late at night is a poor idea for anyone who then wants to sleep.
The practical answer is tunable white, or two settings at minimum, plus a dimmer with real range at the bottom. A lamp that only reaches 40 percent of full output is not usable in a dark room.
Flicker, and the camera in your pocket
Most LED lamps dim by pulse-width modulation: the LED switches fully on and off very fast, and the duty ratio sets apparent brightness. At high frequency with shallow modulation depth it is invisible. Done cheaply, at low frequency and full depth, it produces flicker that many people register as eye strain, headache, or a stroboscopic smearing on moving hands. Cheap lamps flicker because low-frequency PWM needs almost no components, while steady drivers need better ones.
You can test for this without instruments. Point a phone camera at the lamp, dim it low, and look for rolling bands across the preview; slow-motion video makes it clearer. Waving a pencil quickly under the light is the other check: a smooth blur means steady light, a row of separate pencil images means it is switching. Both catch bad flicker rather than certifying its absence.
Glare is about area, not output
Perceived glare tracks luminance, which is intensity divided by the area you see it over. The same output through a bigger, better-diffused surface is far more comfortable than a small bright chip, which is why a wide diffused bar beats a compact head of the same wattage. If you can see the emitter from your seated position, the design has failed regardless of its numbers.
For screen work the argument goes further. A monitor is already a light source, so the goal is asymmetric: light the desk, keyboard and paper, and keep light off the screen face. Light striking a display washes out contrast and, on a glossy panel, reflects the lamp back at you. Bar lamps that clip to the top of a monitor exist for that reason, throwing a controlled forward-and-down beam with nothing spilling onto the glass.
What to look for
- A published lux figure with the measurement distance stated, not a bare lumen count
- R9 as a separate number, or a TM-30 Rf/Rg pair, rather than Ra alone
- Tunable colour temperature, or at least a warm option near 2700K for evening use
- A dimmer that reaches genuinely low output, and a stated PWM or flicker frequency
- A large diffusing surface, with shade geometry that hides the emitter from seated eye height
- An arm that sets head height and angle, since that controls illuminance more than output does
A lamp listing R9, driver frequency and lux at a stated distance was engineered rather than assembled. Where a datasheet gives only lumens and a colour temperature, assume the rest was never measured.