Mike Davis has some land in Arizona that is off the grid. He has generally used a small homemade solar array to keep a small battery bank charged, but he had to move the panels manually every few hours for optimal exposure to the sun. He decided it was time to automate this process, and he did so with an antique television antenna rotator that he got for $15 at a garage sale.
The first thing he had to do was build the frame. Building a wooden frame is pretty easy, but he had to design this one so that it would hold the array at the optimal angle for final deployment on his property (at 34.6 degrees North latitude). The frame itself isn't anything special. It is composed mainly of 2x4 lumber and can be broken down for easy transportation. This was necessary because he was building it in his workshop in Florida and would drive it to his land in Arizona. Davis tells us on his website that he will be using it only during a specific time of the year, so there was no need to make the vertical angle adjustable.
After the physical construction was done, it was time to move on to the electronics. His initial thoughts were that he could salvage the original motor controller and just add an external system for tracking. However, once he dug in, he found that the system probably couldn't handle the abuse he intended to throw at it. The system had been designed for rare rotations for correction, but he planned on rotating for all the daylight hours every day.
He then set out to design a new control system. Having worked with MBED at work, he felt comfortable using that as the brain. He says on his site that the MBED platform is extreme overkill for this, but his familiarity allowed him to build the system quicker. Having that extreme overkill also means that he has plenty of room to expand the system or add features in the future, such as statistics logging.
The MBED reads voltage from two small photovoltaic cells placed at right angles to each other. It rotates until the voltages on the two panels match. However, Davis found that this did not work in the Arizona sun as well as it did on his Florida workbench. The sun was too bright, and the difference between the panels was too minimal at any angle. He remedied this by adding a small piece of metal to shade the center of the two sensors. This occulting bar would cast a shadow that would make the difference in output voltage more extreme based on the angle.
Davis has many more pictures of the project, a full schematic, and the code for the MBED on his website. You can see a video of the final result below.
It is easy to calculate the position of the sun very accurately based on location and time. I do not understand why an optimization algorithm is used when you could just calculate the position. The MBED has an on board real time clock and the location could be input each time it was moved so a dedicated GOS unit on the tracker would not be required.
An on GPS could easily be used if the designer thought the drift on the RTC was too great. The location of the system could be read in from the GPS when started at a new location. Measurements of the location could be read in for a pre-determined period and the location would then just be the average of those results, this would provide a better location. Given the distances involved any location the GPS gave you would suffice anytime the GPS had a lock.
this is exactly what one company did recently. Instead of having rotating solar cells, they were cylindrical with a reflector below them that was also cylindrical. When done in a larger quantity this is much more efficient than rotating a flat panel (they claim).
Here they are if you're curious. http://www.solyndra.com/technology-products/
Good thoughts, Luis, and it made me think of another approach. Why not put a tube slightly greater than the diameter of the solar cell that would channel the maximum amount of sunlight when the tube is pointed directly at the sun? Then the rig could be calibrated to maintain the maximum voltage from the cell, which would keep it aimed at the sun.
The way the correction has been applied to rotate the panels to get maximum sunlight thruout the day, a similar correction can be done to track the sun's north south movement during the year. This is especially required for the regions which are away from the equatorial regions.
We at school tried to do this almost same thing. But We were using a stepper motor and it was more of a simulation as we didn't have real solar panels.
But, reading about this makes me question... how about using the same current that the solar panel yields as a measure or as the level to judge if the panel is receiving the light at the expected levels. Mmmm just by thinking on it I guess it would be more difficult as some variations are added to the system like the fact that the sunlight varies in the course of the day and also varies with the seasons.
Thus, perhaps this is something to consider for using all that horse power in the MBED brain ... to make it able to get the weather info and from that define a range of sunlight intensity. Then wage that with some tables which model the expected intensity during the day and season and vuola! now you can spare some hardware and save some bucks! If it's mass produced then you'd be saving a lot of dollars man!
I get it. So the problem wasn't that the sun was brighter, but was more evenly distributed -- the distance from one side of the board wasn't any different from the other side. That makes sense. I'm still not sure how he keeps the increasing angle of the sun against the shade from increasingly throwing off the calibration, but I'll just accept the notion that he found a solution. Thanks!