Radiometric vs photometric light specification: comparing intensity

Two lights sit on your bench: a white one whose datasheet quotes lux and a red one that quotes radiant watts. The paper numbers make the white light look overwhelming. The difference is the units, not the lights. Photometric units — lumen, lux, candela — weight optical power by the human eye's response curve. Radiometric units — watt, W/m², W/sr — count optical power wavelength-blind, the way a physics detector does. A silicon camera is much closer to that detector than to your eye. What follows covers what each family measures, why photometric numbers underrate the red, blue, IR and UV LEDs that cameras see well, and how to convert between lux and W/m² before you shortlist anything.

What is the difference between radiometric and photometric light specs?

Radiometric units count optical power itself — radiant flux in W, irradiance in W/m². Photometric units weight that power by the eye's daytime response, giving lumens, lux and candela. The weighting peaks at 683 lm/W at 555 nm and falls to zero at the edges of the visible band.

Both systems describe the same photons through a different detector. Radiometry answers how much optical power is present; photometry answers how bright it looks to a person. Commercial lighting is sold to people, so its datasheets are photometric, and vision lighting inherited both habits. Every photometric quantity has a radiometric twin:

The bridge between the columns is the photopic response V(λ). It is tabulated by the CIE and normalized to 1.000 at 555 nm. Photometric value = 683 × V(λ) × radiometric value, applied wavelength by wavelength. That single weighting causes every confusion between the two columns.

Why do red, blue, IR and UV lights read low or zero in lux?

Because the eye-weighting function V(λ) falls away from 555 nm. Three LEDs each radiating a nominal 1 W convert to about 683 lm at 555 nm yellow-green, 181 lm at 630 nm red, 62 lm at 470 nm blue and 0 lm in IR or UV — same radiant power, wildly different photometric readings.

The arithmetic, per radiant watt:

Same optical power, an eleven-to-one spread from green to blue, and hard zeros outside the visible band. Not because the emitters are weak — because the unit is defined by an eye that cannot see them.

This creates a classic procurement trap. A light-starved application needs short exposures to freeze motion, someone is sent to buy the brightest wavelength, and a photometric comparison table hands them yellow-green every time. A silicon sensor's quantum efficiency at 630 nm, 660 nm or 850 nm bears no relation to V(λ); many monochrome cameras are at their most sensitive exactly where the eye has given up. For wavelength choice driven by part contrast rather than raw output, start from which LED colour gives the best contrast in machine vision and light colour and monochrome camera contrast: bright vs dark.

Datasheets for infrared and UV illuminators therefore quote radiometric figures — radiant flux or irradiance — because photometric numbers would read zero. The backlights category includes infrared models such as the LS-BKL-0416, an 850 nm backlight with a 136.4 x 136.4 mm lit area, built for constant-on use with no driver needed. In photometric units any 850 nm source is zero by definition; to a silicon camera it is simply a working light.

Which unit should you use to compare machine vision lights?

Use irradiance in W/m² at a stated working distance for front lights — or at the emitting surface for backlights — whenever the comparison includes a monochromatic, IR or UV light. Lux is fair only when every candidate is broadband white.

The judge is the camera, and the camera's signal is irradiance folded with the sensor's quantum efficiency, integrated over the spectrum and multiplied by exposure time. Neither unit family reproduces that integral exactly. But W/m² is wavelength-blind and honest about delivered power, while lux actively penalises the wavelengths where monochrome sensors often peak. So:

• All-white shortlist: lux comparisons are fair, because the eye weighting penalises every candidate equally.
• Any monochromatic, IR or UV light in the running: compare in W/m², converting any lux figures first.
• Final ranking: measured gray value on the real part at fixed exposure and gain — the ROI reference method for measuring light intensity in machine vision is the quick way.

Compare like with like in the bar lights category: the LS-BAR-1168 is a 423 x 40 mm red diffused bar, and the LS-BAR-1276 is a 372 x 28 mm white diffused bar. Two bars of similar size and radiant output can sit far apart on paper once one is quoted photometrically and the other radiometrically, because the red spectrum lands where V(λ) is small.

How do you convert lux to W/m²?

Divide lux by 683 × V(λ) for a single wavelength. So 1,500 lx from a 630 nm light is 1,500 ÷ 181 ≈ 8.3 W/m². For white light no single factor exists, because the whole spectrum is weighted.

For broadband sources you need the spectrum. Published solar conversions put 1,000 W/m² of daylight at about 120,000 lx, roughly 120 lx per W/m². A typical white LED spectrum converts nearer 300 lx per W/m², because its power sits where V(λ) is large.

Worked example: is the red bar really thirteen times dimmer?

Two example datasheet lines, both at a 200 mm working distance:

• Light A, white bar light: 20,000 lx at 200 mm
• Light B, red 630 nm bar light: 1,500 lx at 200 mm

Photometric ratio: 20,000 ÷ 1,500 ≈ 13.3, so the white light looks thirteen times brighter. Convert both to irradiance:

• Light B: 683 × V(630 nm) = 683 × 0.265 ≈ 181 lm per radiant watt, so 1,500 lx ÷ 181 lm/W ≈ 8.3 W/m².
• Light A: a typical white LED spectrum converts at roughly 300 lm per radiant watt, so 20,000 lx ÷ 300 lm/W ≈ 67 W/m².

Radiometric ratio: 67 ÷ 8.3 ≈ 8. The white light genuinely delivers more power — eight times, not thirteen. And none of that says anything about contrast: on anything that reflects 630 nm strongly, light B can give the higher-contrast image at a fraction of the power.

Cross-check with geometry. If light B's datasheet also lists radiant intensity, the second formula predicts the same answer. At 0.33 W/sr and d = 0.2 m: E = 0.33 ÷ 0.2² = 0.33 ÷ 0.04 = 8.25 ≈ 8.3 W/m². Back through the eye weighting, 8.3 W/m² × 181 lm/W ≈ 1,500 lx. The loop closes on both units.

Exposure makes the decision practical. Camera signal scales with irradiance × exposure time, so cutting exposure from 500 µs to 250 µs to freeze faster motion doubles the irradiance you need: light B must reach about 16.6 W/m² at the part, from overdrive, a second bar, or opening the lens one stop — f/4 to f/2.8 gathers about twice the light, since collection scales as 1/N².

For single emitters specified in millicandela, the same logic runs in smaller units: the conversion from LED luminous intensity in mcd to optical power in mW shows the arithmetic step by step.

What does a source watt or lumen rating actually tell you?

Only what leaves the emitter. A source watt or lumen says nothing about what lands on your part, because it ignores distance and beam geometry — two lights with identical source watts can differ several-fold in irradiance at your working distance.

A source watt or lumen is a flux total, not a density. What the part receives depends on how that flux is spread: into a hemisphere by a bare emitter, into the narrow beam of a bar light, or into a thin line by a collimated source. Radiant intensity plus the inverse-square law takes you from W/sr to W/m² at distance d. Real near-field numbers drift from the ideal as the part gets close. For backlights the part sits almost on the glass, so the figure that matters is radiance or luminance at the emitting surface.

The number to demand from a vendor is therefore intensity at your working distance, at a stated drive, with a uniformity figure over your field of view — and your working distance is not free: it falls out of lens focal length and field of view.

A structured way to force these numbers out before purchase is machine vision lens and lighting selection: a test plan.

Which drive conditions must match for a fair comparison?

The same drive mode and current. An overdriven strobe rating and a continuous rating are different worlds, and a light quoted at full drive cannot be compared with one quoted at a conservative test current.

LEDs are current devices — radiant output scales with forward current — so a datasheet intensity figure means nothing until you know the current it was taken at. Constant-current drive also holds output steady as the LEDs warm, which is what makes figures repeatable between bench and line. The light controllers category carries constant-current units for exactly this: the LS-CTL-0180 is a constant-current controller with 4 channels, 120 W, for 24 V lights.

If you plan to strobe, compare strobe ratings at your pulse width and duty cycle, not continuous ratings — LED strobe duty cycle: matching lights and controllers covers the limits. Where output must hold over a shift, closed-loop lighting control for repeatable vision images adds the feedback. A PLC can dim a channel from an analog output using PID light intensity control with analog output scaling.

When is a lux meter still the right instrument?

Use lux when every candidate is white or broadband — the eye weighting then penalises them all equally — and for anything specified for people, such as lighting around the cell. The moment a monochromatic, IR or UV light joins the list, switch to radiometric measurement.

A lux meter is a silicon detector behind a photopic filter: it applies V(λ) in hardware, defined only across roughly 380–780 nm. It ranks two white lights fairly and calls your infrared light dead. White illuminators add one more photometric wrinkle — their spectra shift with correlated colour temperature, which white LED colour temperature specification for repeatable inspection untangles.

What should you check before ordering?

Demand four things for every candidate: intensity at a stated working distance, the drive current and mode behind that figure, beam geometry or radiant intensity, and uniformity across your field of view — then confirm the winner with gray values on the real part.

• Unit family — W/m², or lux only inside an all-white shortlist; never source watts or lumens alone.
• Wavelength — note V(λ) at that wavelength before trusting any lux figure: 1.000 at 555 nm, about 0.265 at 630 nm, zero in IR and UV.
• Distance — your working distance, not the vendor's reference distance; scale point-like sources by 1/d².
• Drive — same current, same mode, continuous or strobe, for every candidate.
• Geometry — beam angle or radiant intensity, so you can compute irradiance at your distance yourself.
• Uniformity — a hot centre with dark corners wrecks gauging; ask for the uniformity figure over your imaged area.
• Camera — quantum efficiency at your wavelength, and the passband of any lens filter in front of it.
• Gray-value check — a fixed-ROI measurement on the real part at final exposure, before the order is signed.

Shortlist parts in the categories above, pick the drive electronics in the light controller finder, or run the whole station through the configurator.