Assuming your load is constant then, you just need a fixed duty cycle. Either a controller IC from Linear, Texas et al. or a 555 tuned to the right duty ratio and a good frequency should do the job nicely. The actual boost/buck circuit is pretty simple, but make sure to choose good parts; FETs that switch quickly, quick Schottkey diodes etc.
I know it's cheating, and no longer a tracking solution, but I was looking at another cheapo ebay item to perhaps improve my setup, namely
this 3$ DC-DC board. All I would have to do to boost effeciency, is tweak for a nominal range, 18V drain and 12V source. If $3 gets me 15-20% more out of my small 2x10W panels, I'm all for it. What is unclear to me, is whether the LM2596 will boost current as it drops the voltage?!
Just to add; one thing you will find with PV cells is that the internal resistance of the cells varies dramatically over temperature and with irradiance, this means that the maximum power point does tend to move fairly dramatically hour by hour.
Fair point. My panels are amorph ones though, so temperature and brightness reaction tends to be much more nominal, but not as optimal, as mono/poly panels.
We found we got good results by adding a light sensor input to our control algorithm, maybe for your setup this isn't as much of an issue though
Interesting! However, surely, regardless of what the light sensor may suggest for your heuristic based control algorithm, nothing beats a sweep across the 12V - 22V range, multiplying current and voltage output? Are you using the change in brightness to trigger this sweep?