The pacemaker is the Forger–Jewett–Kronauer (1999) van der Pol-type limit-cycle oscillator driven through Kronauer's light-transduction "Process L", with the light-transduction constants of Forger, Jewett & Kronauer (1999, J Biol Rhythms 14:532) and the oscillator constants (μ, q, k, τx) that St Hilaire et al. (2007, J Theor Biol 247:583) also use:
dx/dt = (π/12) [ xc + μ(x/3 + 4x³/3 − 256x^7/105) + B ]
dxc/dt = (π/12) [ q xc B − x((24/(0.99729 τx))² + k B) ]
B = Gα(1 − n)(1 − 0.4x)(1 − 0.4xc)
α = α0 (I/I0)^p (I in lux, per minute)
dn/dt = 60 [ α(1 − n) − β n ]
μ=0.13 q=1/3 k=0.55 G=33.75 α0=0.05/min I0=9500 lux p=0.5 β=0.0075/min
x is the main oscillator; its minimum is used here as the marker of the core-body-temperature minimum (an offset of up to about an hour between the two is a modelling convention, so the absolute clock hour shown is approximate). n is the fraction of photoreceptor capacity already used, so light saturates: the same lux has less effect the longer it has been on. Time runs in hours and is integrated with fourth-order Runge–Kutta.
What you can test: the free-running period in darkness equals τx; when the light schedule (the zeitgeber) is close enough to τx the clock locks to exactly 24 h and the phase of the x-minimum relative to waking is set by the light level; outside the entrainment range (try the "Long clock" scenario) the phase keeps drifting. The phase-response curve is measured by simulating one pulse at each of 24 circadian phases from the dark limit cycle and comparing the later x-minima with an undisturbed run: light before the minimum delays the clock, light after it advances it.
Simplifications: a single population-average oscillator of an average adult; no age, chronotype, sleep-homeostasis or non-photic (food, exercise) inputs; light is a plain lux number with no spectrum (real melanopsin response peaks near 480 nm blue light); the outdoor sky strip is only decoration. Cross-checked against the St Hilaire et al. (2007) CellML model: μ=0.13, q=1/3, k=0.55, I0=9500 lux and τx=24.2 h agree; that paper’s own light stage is a refit (G=37, α0=0.1, p=0.6, β=0.007, with an extra I/(I+100) factor), whereas this lab uses the original Forger et al. (1999) values G=33.75, α0=0.05, p=0.5, β=0.0075 — these appear in published reimplementations but were not re-read from the 1999 paper itself, and the value of μ differs between published variants (0.13 here). Individual people differ. Not medical advice.