Stray Light Analysis and Control

This map uses Eric C. Fest, Stray Light Analysis and Control (SPIE Press, 2013). It is the best of the five references for interpreting KrakenOS’ non-sequential branches, ghosts, CAD occlusion, diffuse scattering, and path history. It also defines the absolute radiometric quantities that KrakenOS does not yet calculate as a complete calibrated system.

Coverage by chapter

Fest-to-KrakenOS coverage
Book section	Match	KrakenOS implementation
Ch. 1, terminology	Related	Surface roles, interaction types, branch paths, media transitions, and detector hits provide concrete counterparts to path terminology.
Ch. 2, radiometry	Partial/not modelled	Relative branch power and detector sums exist; general radiance, throughput, PST, VGI, NEI, and thermal detector response do not.
Ch. 3, basic ray tracing	Direct/partial	Non-sequential geometry, branching, ray depth, detector acceptance, ancestry-like paths, and scatter targeting are implemented. Statistical convergence analysis and a general backward tracer are absent.
Ch. 4, optical roughness scatter	Partial	Optional pySCATMECH BRDF models include microroughness/PSD physics and weight deterministic child directions. KrakenOS does not fit Fest’s models from metrology data itself.
Chs. 5–6, contamination and black-surface scatter	Partial/related	Imported BRDF parameters and Lambertian, cosine-lobe, Oren–Nayar, or pySCATMECH surfaces can approximate them; contamination budgets and measured black-coating databases are absent.
Ch. 7, ghosts and diffraction	Direct/partial	Fresnel/coating split branches, TIR, and path/power records model ghosts. Aperture diffraction is available in separate field/PSF tools, not as a BDDF child in the non-sequential tracer.
Ch. 8, stray-light-aware optical design	Related	Out-of-field launches, detector maps, branch filtering, and layout edits support the process; KrakenOS does not optimize directly for PST.
Ch. 9, baffles and cold shields	Partial/related	CAD or analytic absorbing/scattering geometry can test baffles. There is no automatic baffle designer or thermal cold-shield emission model.
Ch. 10, BSDF/TIS and system measurements	Not modelled/related	External measured BRDF values may be supplied, but KrakenOS is not a goniometer or stray-light measurement reduction package.
Ch. 11, engineering process	Related	Saved layouts, reports, validators, and reproducible traces help with verification, but do not implement the organizational process.

Radiometric boundary: Chapter 2

Section 2.1.6, Eq. (2.17), printed p. 22, defines throughput (étendue) between source and collector areas:

\[G=\iint \frac{dA_s\cos\theta_s\,dA_c\cos\theta_c}{d^2} =\int dA_s\cos\theta_s\,d\Omega_c.\]

Equation (2.20) gives the index-weighted invariant \(n_1^2G_1=n_2^2G_2\). KrakenOS traces the geometry needed to estimate area and solid-angle acceptance, but it does not expose a general étendue integrator or enforce radiance conservation as an absolute unit contract.

Section 2.1.10, Eq. (2.27), printed p. 25, defines

\[\operatorname{BSDF}(\theta_i,\phi_i,\theta_s,\phi_s) =\frac{dL_s}{dE_i},\]

and Eq. (2.32), printed p. 28, integrates it over projected solid angle:

\[\operatorname{TIS}=\int_{2\pi} \operatorname{BSDF}\cos\theta_s\,d\Omega_s.\]

scatter_backend.pyscatmech_scalar_brdf evaluates a selected scalar BRDF at an incident/outgoing direction. KrakenSys.__LambertianScatterSamples multiplies sampled BRDF values by \(\cos\theta_s\), normalizes the sampled weights, and distributes the user-specified total reflectance among child rays. This preserves the configured child-power budget, but it is not a quadrature evaluation of Eq. (2.32), nor does it turn the result into absolute radiance. The built-in Lambertian and lobe models are teaching/engineering approximations with the same limitation.

Non-sequential paths: Chapter 3

Section 3.1 requires both optical and mechanical geometry plus surface optical properties. system.NsTrace intersects analytic surfaces, optical solids, and promoted CAD faces; per-face roles can select optical, mirror, absorbing, diffuse-scatter, detector, or mechanical behaviour. Each hit records surface, face, point, direction, normal, media transition, interaction type/model, branch path, power, and OPL.

Section 3.2.7, printed pp. 54–55, describes ancestry as the number of ray splits and uses it to classify first- and higher-order stray paths. KrakenOS’ closest direct representation is a hierarchical branch path such as root/R/T/scatter01 plus branch depth and parent/child records. It is more descriptive than one ancestry integer because it retains interaction kind. Power/depth thresholds bound the tree.

Section 3.2.8 discusses Monte Carlo ray splitting. KrakenOS normally creates a deterministic set of hemispherical or targeted scatter directions and normalizes their powers. It can therefore compare repeatable paths, but its default scatter operation is not Fest’s stochastic ray-splitting estimator and does not report estimator variance or convergence versus ray count.

Section 3.2.3’s backward trace is also a real gap. Launching a ray in the opposite direction or reversing a sequential prescription can answer some reciprocal geometry questions, but KrakenOS has no general adjoint radiometric tracer that starts at every detector pixel and preserves the book’s source and throughput bookkeeping automatically.

Roughness and BRDF models: Chapters 4–6

The optional pySCATMECH backend connects most closely to Chapter 4. A surface may select Microroughness_BRDF_Model and pass its wavelength-dependent parameters; the backend converts KrakenOS vectors to incidence, scatter, and azimuth angles and requests a scalar BRDF. Other pySCATMECH models can be named through the same interface.

docs/source/manual/diffuse_scattering.rst documents the complete setup and fallback rules. If pySCATMECH is unavailable or returns no usable weights, KrakenOS falls back to a built-in model. A successful trace therefore needs a backend-status check before it is cited as a microroughness prediction.

KrakenOS does not derive a PSD from a surface map, infer contamination level, fit an Harvey–Shack/ABg model, validate reciprocity, or integrate TIS from a measured data set. Those preprocessing and verification steps must be done outside the tracer.

Ghost reflections: Chapter 7

For an uncoated dielectric boundary, Sec. 7.1.1 uses the Fresnel amplitude coefficients. KrakenOS evaluates their S/P vector equivalent and converts to power. At normal incidence this reduces to

\[R=\left|\frac{n_1-n_2}{n_1+n_2}\right|^2, \qquad T=1-R\]

for lossless real media. Coating tables can replace the interface split, and bulk absorption reduces power along a segment. NsTrace follows reflected and transmitted descendants, so a two-reflection lens ghost appears as an explicit path rather than a paraxial estimate.

Path existence alone is not a quantitative ghost result. Coating wavelength and angle, complex phase, bulk loss, aperture clipping, focus, source extent, detector sampling, scatter, and branch cutoffs all affect the final artifact.

Practical KrakenOS workflow

Model every optically visible mechanical surface, not only the prescription.
Assign measured or defensible Fresnel, coating, absorption, and scatter data.
Launch both in-field and plausible out-of-field sources over angle and wavelength.
Inspect detector-reaching branch paths before interpreting a heat map.
Increase ray/scatter sampling and branch limits until the quantity of interest is stable.
Normalize to a stated incident power, then perform the missing absolute radiometric conversion externally when watts, radiance, irradiance, PST, or detector signal are required.

Important gaps

No complete absolute-radiance engine, blackbody/internal thermal emission, PST/PSNIT/VGI/NEI calculator, statistically controlled Monte Carlo estimator, backward/adjoint tracer, contamination budget, diffraction-as-BDDF branch, automatic baffle designer, BSDF/TIS measurement reducer, or uncertainty propagator was found in the reviewed implementation.