This is a genuinely 2D scattering problem, not a squashed 3D one. The 3D version of this simulator sums flat-panel physical-optics returns (σ ∝ A²/λ², a cosⁿθ specular falloff). Here the airframe is a flat plan-view outline and every edge is treated as a 2D diffracting strip — the classical physical-optics result for a finite strip illuminated broadside, which has a completely different functional form:
σ₂D(θ) = (k·L² / π) · cos²θ · sinc²(k·L·sinθ)
k = 2π/λ, θ = angle between edge normal and radar direction
sinc(x) = sin(x)/x (sinc(0) = 1)
σ₂D total ≈ Σ σ₂D,edge (incoherent sum over sunlit edges, θ < 90°)
Two things fall straight out of that formula and are worth watching for: σ₂D scales as 1/λ here (not 1/λ² like the 3D panel model — a 2D "echo width" has units of length, not area), and the sinc² term produces real diffraction sidelobes flanking the main specular glint, instead of the 3D model's smooth cosⁿθ roll-off.
- Faceted stealth — every edge sits at its own oblique angle, mirrored left-right but never repeated, so no two edges present the same normal to a sweeping radar; large glints get pushed to a couple of predictable, narrow slivers instead of a broad reflective arc.
- Conventional (boxy) — the fuselage sides run parallel to the body axis and the tail is a flat cap perpendicular to it, so as azimuth sweeps through 90°/270° (broadside) and 180° (tail-on) the radar's line of sight lines up square with a long flat edge — a genuine "corner reflector" spike, visibly higher than anything the faceted outline produces at any azimuth.
- Band selector — σ₂D ∝ 1/λ, so X-band (3 cm) returns are roughly 8× stronger than L-band (23 cm) from the same geometry; the sinc² lobe is also narrower at X-band (shorter wavelength diffracts less), so glints are brighter but more sharply localized in azimuth.
Simplification note: like the 3D version, this sums edge returns incoherently (power, not phase) and ignores edge-to-edge diffraction coupling and traveling waves — a teaching model of the shaping principle, not a certified RCS prediction code. Numerically verified: the model reduces exactly to σ₂D(0)=2L²/λ at normal incidence, is symmetric in θ, has its first diffraction null at sinθ=λ/(2L), and — over a full 360° sweep of both outlines — the boxy shape's peak echo width and its fraction of azimuth spent above a strong-return threshold both come out higher than the faceted shape's, matching the intended stealth-shaping result.