All case studies
Illumination and stray light

Irradiance of a Lambertian source and ghost reflections of a glass plate

The etendue non-sequential ray tracer compared with exact radiometric view factors and with the Fresnel reflectance at normal incidence

Irradiance of a Lambertian source and ghost reflections of a glass plate

Irradiance of a Lambertian source and ghost reflections of

a glass plate

The etendue non-sequential ray tracer compared with exact radiometric view factors and with the

Fresnel reflectance at normal incidence

The questionThe reference
Does etendue put the right amount of light in the rightBoth questions have exact answers. The irradiance of a
place? An extended source must be sampled so that itsuniform Lambertian disc on a parallel plane follows from
irradiance is right at every distance. A ray that meets athe view factors tabulated by Howell (cases B-12, B-14
glass face must split into reflected and transmitted partsand C-41). A plate of index n reflects R = ((n − 1)/(n + 1))²
that carry the right power, through any number ofat each face (Hecht, § 4.6). Summing the reflections in
reflections.power gives a transmission 2n/(n² + 1) and ghost-to-

image ratios R² and R⁴.

1 The two systems

The disc emits 1 W with a cosine law. Its irradiance is read on a plane PART VALUE at one to ten source radii, where the inverse-square law is wrong by Source disc, 5 mm radius, Lambertian, 1 W, up to a factor of two. The plate is lit at normal incidence, so 555 nm polarisation plays no part. Its index goes up to 4 so that the higher Rays 2 × 2 million per detector distance ghosts are strong enough to read. Receiver 400 × 400 mm plane at h = 5, 10, 20, 50 mm

Plate 5 mm thick, n = 1.5, 2, 3, 4, no absorption

Beam on the plate collimated, 2 mm radius, normal incidence

2 How far etendue is from the exact values

How close is the irradiance? Within a third of a percent everywhere, and within the noise of the random sampling. The ghost ratios of the plate agree to fourteen digits, because a reflection that is not too faint is followed deterministically.

Figure 1. a. Relative difference between etendue and the exact value for the on-axis irradiance E₀ at four distances, the irradiance E(s) off axis at 20 mm, and the power Φ(ρ) within a radius ρ. b. Transmission and ghost-to-image ratios of the plate against its index. Image power for n = 1.5: 0.921600000000 of the source, exact 0.9216.

0.32% 2.6σ 0.001600 29 of 29

Largest difference from the Largest difference in units of Two-reflection ghost of an n = Comparisons within the limits

exact irradiance or encircledthe Monte Carlo noise. All 131.5 plate relative to its image.written before the
power, over 13 comparisonslie within 3σ.Exact: 0.0016.measurement.

on a Lambertian disc.

etendue.io References: Howell, Catalog of Configuration Factors; Hecht, Optics, § 4.6 · etendue, November 2026

The numbers, and what they do not say

Every comparison, the noise it was judged against, and what was not met.

Table 2. Lambertian disc, 5 mm radius, 1 W

Quantity Exact etendue Diff. (%) σ (%) Diff. / σ Verdict

On-axis irradiance, h = 5 mm 0.0062337 0.0062377 +0.06 0.33 +0.2 met

On-axis irradiance, h = 10 mm 0.0024619 0.002461 −0.04 0.28 −0.1 met

On-axis irradiance, h = 20 mm7.2958×10⁻⁴7.2976×10⁻⁴+0.030.30+0.1 met
On-axis irradiance, h = 50 mm1.2247×10⁻⁴1.2248×10⁻⁴+0.010.83+0.0 met
Irradiance 10 mm off axis, h = 20 mm4.8812×10⁻⁴4.8718×10⁻⁴−0.190.25−0.8 met
Irradiance 20 mm off axis, h = 20 mm1.9784×10⁻⁴1.9800×10⁻⁴+0.080.23+0.3 met
Irradiance 30 mm off axis, h = 20 mm7.7040×10⁻⁵7.7080×10⁻⁵+0.050.31+0.2 met
Irradiance 40 mm off axis, h = 20 mm3.2407×10⁻⁵3.2303×10⁻⁵−0.320.41−0.8 met
Irradiance 60 mm off axis, h = 20 mm8.0233×10⁻⁶8.0447×10⁻⁶+0.270.62+0.4 met

Power within 5 mm radius, h = 20 mm 0.055728 0.05568 −0.09 0.63 −0.1 met

Power within 10 mm radius, h = 20 mm0.192240.19218−0.030.13−0.2 met
Power within 20 mm radius, h = 20 mm0.492190.49199−0.040.05−0.8 met
Power within 40 mm radius, h = 20 mm0.797980.79746−0.070.03−2.6 met

Irradiance in W/mm², averaged over the same detector cell as the exact value; power as a fraction of the emitted power. h: distance from the source to the plane. σ: Monte Carlo noise, from two independent draws of two million rays and from the binomial law. A value is met when it lies within 3σ of the exact one.

Table 3. Glass plate, 400 rays: relative difference from theTable 4. Transmission bias, 40
exact valueindependent draws

n Transmission Image 2-refl. ghost 4-refl. ghosts n Mean diff. Std. error

1.5 −2.2×10⁻⁷ −8.1×10⁻¹⁵ 9.8×10⁻¹⁵ −8.4×10⁻² 1.5 1.3×10⁻⁷ 1.3×10⁻⁷

2 5.9×10⁻⁶ −6.7×10⁻¹⁵ 4.1×10⁻¹⁵ 4.1×10⁻² 2 −8.7×10⁻⁸ 6.8×10⁻⁷

3 8.0×10⁻⁶ 3.4×10⁻¹⁵ 0 0 3 8.1×10⁻⁷ 9.4×10⁻⁷

4 3.5×10⁻⁶ 8.4×10⁻¹⁵ −1.0×10⁻¹⁴ −5.8×10⁻¹⁵ 4 5.5×10⁻⁷ 1.2×10⁻⁶

Ghosts are counted relative to the image. In red: not met. A disc source of the same radius, 0.01° divergent, 500 rays per draw.

Where the noise is large fringes of a plate. A ray tracer sums powers, as does the

reference. The four-reflection ghosts of the n = 1.5 and 2 plates fall

^(b)^(e)^(l)^(o)^(w)^(1)^(0)^(⁻)^(⁴)^(o)^(f)^(t)^(h)^(e)^(s)^(o)^(u)^(r)^(c)^(e)^(,)^(w)^(h)^(e)^(r)^(e)^(f)^(a)^(i)^(n)^(t)^(r)^(e)^(f)^(l)^(e)^(c)^(t)^(i)^(o)^(n)^(s)^(a)^(r)^(e)^(e)^(n)^(d)^(e)^(d) How it was computed at random. They survive in 6 of 400 and 282 of 400 rays, so Each scene is built and traced as the non-sequential editor their value carries a large but known noise, and their does it, and read by the irradiance map, encircled energy, difference (−8%, +4%) lies within it. flux budget and ghost analyses of etendue. No value was

What this does not cover tuned. The 100-ray and 400-ray traces of the plate use

different random seeds, so that their difference measures Oblique incidence, where s and p polarisation reflect the noise of the random ending. Table 4 checks the differently, and coherent effects such as the Fabry-Perot transmission for a bias over 40 further independent draws.

References

[1] J. R. Howell, A Catalog of Radiation Heat Transfer Configuration Factors, 3rd ed., online, cases B-12, B-14, C-41; after H. C. Hottel, Trans. ASME 53 (1931) 265-273, and D. C. Hamilton, W. R. Morgan, NACA TN 2836 (1952). [2] E. Hecht, Optics, 5th ed., Pearson (2017), § 4.6, reflectance at normal incidence.

etendue.io Limits set before the measurement: 3σ of the Monte Carlo noise (disc); 3σ of the 100 to 400 ray difference, plus 10⁻⁹ (plate).