This weekend was bitter cold. With the temperature in the single positive digits (F) I was unwilling to drag my gear outside. But at the same time I needed to run some experiments. So I set myself up in the dinning room and looked out the window at the sun. By the time I got set up, the sun was quickly setting. I was able to get some time under bright conditions, then hazy, then partially obscured, and finally completely obscured by my neighbor’s roof.
The results were mixed and backwards from what I anticipated. Under bright sun, the tracker had a hard time, but when the sun had set, it keep working well for some time.
Here is a bit of video from those tests: http://youtu.be/WyfkQBYg0rE
I discovered a few issues. First is that I will need to allow for adding and subtracting neutral density filters based on the conditions. The dynamic range of the sensor although large is not large enough to go from full bright sun to dusk. This I think is why the tracker did not work in full sun. Unfortunately by the time I got my scope up stairs to look at the signal, the sun had fallen behind some light clouds.
Second, my calculation of figure of merit needs help. I was computing it as (Max-Avg)/Avg. The problem is that when the average drops down low as towards dusk, even small peaks cause big numbers. I think I will be better of with just the number of volts the Max is larger than the peak.
I also found some fascinating things. One is that the sensor has a much faster voltage slew when it sees light than when light suddenly goes away. I get a very sharp leading edge and then a near RC decay after the light passes by. Need to look into this to see if I can modify the circuit to account for this effect.
The other interesting thing is that although the clockwise and counterclockwise code is ‘identical’, it takes 0.78% longer to go one direction that the other. It is insignificant, but it bothers me no end!