A fixed south-facing row tilted by beta, slope length c, stands h = c sin(beta) high and covers a horizontal depth d = c cos(beta). Seen along the rows, the sun has a profile angle theta (tan theta = tan(elevation) / cos(azimuth from south)). The top edge, at height H above the ground, throws a shadow H / tan(theta) towards the next row; if the row pitch p is smaller than d + h / tan(theta), the lower part of the next row is shaded.
declination (Spencer series), sin(alt) = sin(lat) sin(dec) + cos(lat) cos(dec) cos(H), H = 15 deg x (hour - 12)
profile angle: tan(theta) = up / south (components of the sun vector)
shaded slope length s = (d + h / tan(theta) - p) / (cos(beta) + sin(beta) / tan(theta)), fraction = s / c
no-shade pitch (21 Dec, 9:00 and 15:00): p = d + h / tan(theta_9h) ground cover ratio GCR = c / p
beam on the panel = DNI cos(incidence) (1 - shaded); diffuse x (1 + cos beta)/2; ground reflection 0.2 x (1 - cos beta)/2
levelised cost = (module cost per m2 + land rent x p / c) / (annual energy per m2 x efficiency)
- Shade is a winter-morning problem - the 21 December rule guarantees no shading at the worst time, but it wastes land; most of the year the sun is much higher, so a tighter pitch loses only a few percent of the annual energy.
- Energy per land area keeps rising as rows get closer (more panel per hectare outweighs the shading), so it has no interior optimum by itself; the pitch that minimises the levelised cost balances shading against the price of land, and moves wider as land gets cheaper.
- Shading is not linear in real modules - the "bypass strips" option rounds the shaded height up to whole substrings (three per module high), which usually costs more than the area fraction.
- Same model in the 2D twin - one shared physics module; the checks compare it with spherical trigonometry vs vector rotation, a 3D shadow projection, a brute-force ray test, step-size convergence and a fine scan.
- Honest simplifications - infinite rows (no end effects), clear sky with a simple air-mass beam model and a constant 10 % diffuse share, solar time without the equation of time, isotropic diffuse not reduced by neighbouring rows, no mutual reflections, no temperature, soiling or inverter-clipping losses, sun north of the east-west line ignored, northern hemisphere only. Cost figures are illustrative placeholders; real layout design uses hourly weather data and a yield tool.