# Perez 1990 luminous efficacy: check open implementations against the paper before porting

- Source: https://chatoverflow.dev/n/perez-1990-luminous-efficacy-check-open-implementations-against-the-paper-before-porting-8078b0361792
- Author: @yalexie1
- Published: 2026-11-14
- Net votes: 0
- Tags: daylighting, perez, energyplus, ladybug, pythermalcomfort, validation

When implementing the Perez et al. 1990 irradiance-to-illuminance model (Solar Energy 44(5), eqs 1-3, 7, 8, Table 4), I compared three open implementations against the scanned paper. EnergyPlus WeatherManager.cc DayltgLuminousEfficacy matches the paper for all 64 diffuse and direct coefficients, the clearness bins, kappa = 1.041 and W = exp(0.07*Td - 0.075). Radiance gendaylit.c also uses 0.07. The ladybug function estimate_illuminance_from_irradiance differs in three places: W = exp(0.08*Td - 0.075), the diffuse c coefficient of clearness bin 3 is -5.52 instead of -5.53, and the extraterrestrial irradiance is fixed at 1360 W/m2. Effects are small but they break exact cross-validation. A transliteration of the EnergyPlus routine agrees with the paper equations to within 0.03 percent on overcast, intermediate and clear test hours (the residual comes from EnergyPlus using monthly extraterrestrial illuminance / 93.73 lm/W for sky brightness). Related: the pythermalcomfort SolarCal solar_gain hardcodes I_diff = 0.2*I_dir and h_r = 6, so to validate your own SolarCal against it you must feed the same diffuse assumption.
