Lumens, Candela, and Lux: How to Read DXZ Lighting Reports

Most spec sheets throw lumens, candela and lux at you in the same paragraph, as if they were three ways of saying “bright.”

They are not. They answer three different questions. And confusing them is how buyers end up with a lamp that looks impressive on paper and disappointing on the road.

Lumens describe total luminous flux. Candela describes luminous intensity in a direction. Lux describes illuminance on a surface. We use selected DXZ reports to show why these three measurements answer different questions. The paired records share model identifiers — but they were not necessarily produced from the same physical sample, or on the same date.

Dataset

58 selected model-level matches

Equipment

Sphere + GO-HD

Measurement

H = 1.000 m · 13.2 V

Source

DXZ in-house archive

Lumens vs Candela vs Lux: The Short Answer

If you remember one thing: lumen ratings are a property of the fixture; lux readings are a property of the situation. The fixture doesn’t change with position. The illuminance reading does — it depends on distance, direction and the orientation of the receiving surface.

Lumens (lm): How Much Light Leaves the Lamp

A lumen is the total luminous flux a source emits, summed over every direction. This is the number on the box, the one quoted in every product listing — and the one most buyers compare.

It is measured in an integrating sphere: a large hollow sphere with a highly reflective white interior. The lamp is mounted inside, the sphere mixes the emitted light into a uniform field, and a spectrometer reads the total flux.

Every direction contributes equally, so the sphere is the right instrument for a “total output” number. It is the wrong instrument for anything to do with beam shape.

Across our white-light production — 165 white-light sphere reports, each carrying a recorded flux value — total output runs from 240.7 lm to 9,051 lm, with a median of 1,659 lm. That span covers everything from compact marker units to full-width light bars.

It is a useful number. On its own, though, it tells you almost nothing about what the lamp will light up. Here is why.

Consider two lamps from the same 4C product family:

The second lamp produces about 35 % more total light — yet its measured peak illuminance is only about one-eighth as high. Both numbers are correct. Both come from our own reports.

Beam distribution is one possible explanation, but these two pairs alone don’t isolate the cause. Both records were produced near 13.2 V, with the lamp on the same mounting geometry. But the sphere run and the GO-HD run are separated in date. The two GO-HD scans also used slightly different survey areas (200 × 40 mm and 200 × 38 mm). So the pair is not a controlled experiment. What it does show: the difference isn’t total output — it is where the output goes.

Candela (cd): How Concentrated the Beam Is

Candela is luminous intensity — the light emitted in one particular direction, expressed as lumens per steradian (lm/sr). If lumens are the total water in the tank, candela is the pressure at the nozzle.

This is where beam optics live. Push the same LED output through a narrow reflector and the solid angle shrinks, so candela rises. Push it through a wide flood optic and the light is spread across a bigger cone, so candela falls.

Total lumens can stay identical while candela changes by an order of magnitude.

This is why candela vs lumens is the comparison that actually matters for driving lights, spot lights and long-range auxiliaries. A beam that has to reach down the road is a beam-concentration problem, not a total-output problem.

It is also why the peaks you see in our own data do not line up with the lumen column.

Lux (lx): How Much Light Reaches the Surface

Lux is illuminance — one lumen per square meter. It quantifies the light falling on a target surface. It is widely used when specifying how much light reaches a work area or a road surface.

It is the only one of the three units that changes with position. Take a point-source approximation, with the receiving surface perpendicular to the incident direction. Now the inverse-square law applies: double the distance, and illuminance drops to one quarter. Move the measurement point off-axis, or tilt the receiving surface, and the intensity in that direction sets the reading — not the distance alone.

Across our 100 unique GO-HD illuminance reports, measured on the road plane at a fixed mount height of H = 1.000 m:

• Peak illuminance ranges from 18.18 lx to 652.0 lx, with a median of 108.75 lx.
• Every report is a road-facing beam measurement — a partial measurement window (vertical −10°…+5°, horizontal −30°…+30°), not a full-sphere distribution. That is the right method for a forward-facing lamp, and we don’t present it as anything else.

One honest note about this figure: we deliberately plot illustrative fixed-intensity sources rather than our own lamps. Our GO-HD reports measure illuminance on a road plane. Turning that into a candela curve would require assumptions the measurement itself does not justify. So we keep the physics clean and the product data in Figure 3, where it belongs.

Converting Between Them (and When the Conversion Breaks)

Two relationships link the three units. One is trustworthy under stated conditions. The other needs a caveat you shouldn’t skip.

Candela ↔ Lux: exact in principle, approximate in practice

E (lx) = I (cd, in the direction of the measurement) × cos θ / d²    — when the surface is perpendicular, cos θ = 1: E = I / d²

Say a lux meter reads 50 lx at 5 m from a directional source. That source’s intensity is 50 × 5² = 1,250 cd in that direction, with the receiving surface perpendicular to the incident ray and d the actual light-path distance. The simplified form E = I / d² holds only at that orientation. Where the surface is tilted, the cosine term of the general form applies instead.

Caveat, stated plainly: the inverse-square law assumes a point source and a receiving surface perpendicular to the incident direction. A real lamp has a physical aperture, and a vehicle lamp is an extended source. So near-field geometry and scene reflections can break the simple form at short range. Treat this as an engineering relationship for well-separated source and receiver — not a certified calibration. And remember that d is the light-path distance, which is not the same as the mounting height H at every point on a road plane.

Lumens ↔ Candela: there is no exact conversion

I_avg (cd) = Φ (lm) ÷ [2π × (1 − cos(θ/2))]

That formula holds for an idealized uniform circular cone. It estimates the average luminous intensity inside that cone — not the lamp’s true peak candela.

Real automotive and work-light beams are not uniform — brightness varies across the beam. So there is no exact conversion from total lumens to peak candela for a real lamp. The uniform-cone shortcut is sound for estimating an average intensity inside the cone. It should not be applied unconditionally to the peak intensity of a real lamp. For a real lamp, measure peak candela from the angular intensity distribution.

If you searched for “how many lumens in a candela” hoping for a single number, here is the honest answer. The factor doesn’t exist without the beam geometry — and it doesn’t exist without an assumption about how light is distributed inside the beam. For anything contractual, measure the distribution.

What the 58 Model-Level Matches Show

Theory is cheap. Here is the data.

We matched our archives and found 58 white-light model identifiers that carry both an integrating-sphere report (total luminous flux) and a GO-HD illuminance report (peak illuminance).

These are model-level comparisons across archived tests — not measurements of the same physical specimen taken simultaneously. The two reports for a given model were produced on different dates, and we treat the match as same model identifier, not same unit.

What the chart shows:

• Across the displayed range of these 58 model-level matches, a luminous-flux value alone does not uniquely determine the GO-HD peak illuminance recorded here — the plotted results vary. A scatterplot that is not a straight line does not by itself prove that no predictive relationship exists. No straight-line fit or correlation is claimed here.
• Recorded peak illuminance per 1,000 lm varies by more than 10× across these matches — from about 18 to 220 lx per 1,000 lm. Two models with a similar lumen figure can sit at opposite ends of that scale.
• The single most instructive pair in this archive remains 4C-BW-37Y vs 4C-F64F: about 35 % more total lumens, yet a recorded peak illuminance only about one-eighth as high. Both values are the recorded figures; the pair does not by itself establish which optical difference produced them.

A note on rigour. We matched on a model identifier present in both archives, white-light models only. Every exclusion is recorded with its reason, rather than averaged away:

• five models whose two reports disagree on measured value — 4C-BW-F10, 4C-BW-37F, 4C-BW-F-Y44, 4C-BW-Y12-X, 4C-Y32F. The re-scans used different survey-area settings, so the two readings are not directly comparable — but neither original value is wrong;
• one 24 V model, which is not comparable with our 13.2 V series;
• one color-changing model, for which white-light photometric metrics are not a meaningful comparison basis.

The plotted set is the 58 model-level matches that remained after those exclusions. The exclusion list above is the selection rule in full. Every excluded record keeps its original measured value. And where a report pair can’t be retrieved and re-checked, read the plotted relationship as a description of this particular collection only. Not as a statement about LED lamps in general.

So Which Number Should You Specify?

A lumen figure with no beam data doesn’t tell you how far the lamp reaches. For a long-range application, ask for the directional intensity, the beam pattern, and the task illuminance at the distance you actually use. For a flood or area application, ask about coverage and uniformity across the target. Which number is decisive depends on the installation, not on the unit alone.

FAQ