I'm reminded of the old experiment:
- Hook up a battery to a voltage divider, which uses a CdS photocell in one leg.
- Watch voltage on scope.
- Hold photocell near scope trace. See deflection.
If you have the trace bright enough and fast enough (faster than the recombination rate of the CdS, usually < 1ms/div), the DC level gets pushed around, and under sensitive enough conditions (vert gain, supply voltage, resistor ratio, trace brightness, etc.), it can be deflected by several photocell widths. Go just a little too far, boop, it slips away.
If you do it on a slow sweep rate (~10ms/div?), you can see the dynamic response (and "drool" over time) as the beam sweeps towards and away from the photocell. Obviously, it won't "catch" if it's much more than a photocell diameter away from baseline.
What's interesting is it demonstrates positive feedback (catching then losing tracking corresponds to hysteresis), yet it's doing it with negative feedback (the signal remains stable while tracking).
Doing it with PV cells is interesting, because in principle, you can use a sufficiently small, extremely geared, motor to push around a fairly large panel for not much extra panel. Limited ultimately by required slew rate and positional accuracy, mechanical losses, and rotation of the Earth.
Tim