Angle of Incidence on a Tilted Surface

At solar noon, for a collector facing due south in the northern hemisphere, the angle between the sun's rays and the panel's normal is $\theta = |\phi - \delta - \beta|$ — latitude, less declination, less tilt. Set $\beta = \phi - \delta$ and the angle is zero: the rays arrive square. This is the noon special case of the general five-term incidence equation in Duffie and Beckman, which adds hour-angle and surface-azimuth terms for every other moment of the day.

Because beam irradiance on the surface falls as $\cos\theta$, the cosine's flatness near zero is a piece of good news worth internalising. Being 10° off costs 1.5%; being 20° off costs 6%; even 30° off costs only 13%. That forgiveness is why fixed arrays work at all, why a roof's existing pitch is usually good enough, and why chasing the last few degrees of tilt rarely pays for the mounting hardware it takes.

The familiar rule "tilt equals latitude" falls straight out of the equation: it is exactly right at the equinoxes, when $\delta = 0$. Follow the optimum through a year, though, and it swings 47° — from latitude minus 23.45° in June to latitude plus 23.45° in December. Two refinements follow. For maximum annual energy the best fixed tilt is typically a few degrees below latitude, because summer days are longer and clearer and contribute more to the total. For a winter-critical off-grid system it is steeper than latitude, often latitude plus 15°, which also sheds snow — and snow on the glass costs far more than any tilt error.

Three caveats keep this honest. It is a noon relation for an equator-facing surface; a panel pointed east or west of south needs the full incidence equation. It describes only the beam component, while diffuse sky radiation arrives from all directions and actually favours a flatter tilt, which is why the true optimum in a cloudy climate sits below what this equation alone suggests. And seasonal adjustment on a ground mount is worth roughly 4–6% of annual energy for two visits a year with a wrench — real, but far less than most people assume before they measure it.