My apologies to any of the 1400 or so people who've read this Instructable in the day since it was published; I described an idea I was sure would work, and I was so excited to get it out there where people could try it, that I didn't wait until I could get the parts I needed to fully test it myself, and I posted based on a partial test and an assumption about the performance of small motors that turned out not to be accurate.
Now normally I would take a faulty Instructable down (and run away and hide), but in this one instance I'm going to leave this online after editing it to correct my mistakes, because I thought that there was still a good idea here that could be developed even if it had to be by someone with more electronics experience than me, and I wanted to put this germ of an idea out there so that maybe it could inspire someone to come up with a solution that works. I offered the 1Yr pro membership upgrade I received when this went front-page as a prize, and as you can see from the comments a lively discussion ensued. (I gave the prize to jtlowe for his suggestion of using a clockwork rotating platform, but although there's no more prizes to give, I would still welcome any more suggestions you all can offer).
The problem is sun tracking: pointing a solar panel directly at the sun so that it can harvest significantly more light - and the difficult aspect of the problem is that the cost of adding a sun tracker to a solar panel in order to gain X% extra output has to be less than X% extra cost, otherwise it's more cost effective to simply add more solar panels.
The solution that I had was to take two small PV cells from a couple of solar garden lights, and connect them not in series or even in parallel, but head to head, connecting the ground lead from one to the ground lead from the other, and determining which of the two panels was receiving more light by looking at which one was able to generate more voltage than the other. For instance if one generated 2V and the other generated 3V, then the voltage between the two positive outputs would be 1 volt, and that volt would be used to drive a DC motor to turn the platform in the direction of the cell that was reading the stronger light signal. (Although in practice the voltage was actually less because driving current through a PV cell against its natural direction (since it acts a little like a diode) causes a voltage drop in excess of just the voltage that would be cancelled out, though that's not relevant to the problem)
Although that is indeed what happens which you can confirm by looking at the voltage on a voltmeter, what I didn't realize was that these panels don't generate enough amperage to drive even the smallest motor, as far as I can tell. I'm talking about 3V motors that need so little power that they'll spin from a single half-dead AA battery!
So what I'm going to show you here is half a solution, and I'm asking the smart readers of Instructables to help come up with the other half.
Step 1: What you'll needGrab a couple of those cheap garden path solar lights, the kind they sell at the Dollar Store for... well, a dollar. (If you miss the occasional Dollar Store deal on those, damn, you'll probably be out of pocket by two dollars from a more expensive store! :-) )
That's it - to test the circuit you won't need any more components. Just your trusty volt meter, and a soldering iron... (I finally treated myself to a programmable temperature controlled iron from Radio Shack and I have to say I'm really enjoying using it compared to what I had before. The volt meter is that cheap one that's on sale for a couple of dollars every weekend at Harbor Freight)
When I was trying to drive the motor and realised I didn't have enough power I added a second pair of PV cells in parallel, so some of the images below have 4 cells and some have 2. As long as we're just using these as light sensors, 2 will do fine - I didn't feel it was necessary to reshoot the photos...
Step 2: Here's the clever part...
Unscrew the top of the light and pull the battery terminals from the plastic case. With the cheapest lights, you need to pull the lugs on the battery terminals back a little before you can pull the AAA battery out; more expensive lights have a more accessible AA battery holder but it may not be as easy to pull the battery terminals out of the holder.
One way of building this circuit is to leave the circuit board in place and use the wires connected to the battery pads - in this version, it's easiest just to pull both negative battery pads from each light and solder them together. On my lights, these wires are black (ground) and yellow (positive).
(The circuit board is a "Joule Thief" circuit. It scavenges otherwise unusable low voltages and then sends a burst of higher voltage to charge the battery. So the output from that circuit is either zero or 1.5V or higher.)
However it's probably better that you clip off the red and black wires coming directly from the PV cells, maybe a centimeter short of the circuit board (in case you want a little wire left to make subsequent experiments with that board easier?) and use the output of the PV cells directly, which depending on the strength of the incident light will range from 0 to about 3V,
In either case, connect the black wires from both cells. (Step 3: Test it!
Connect your voltmeter to the other two wires - the red ones from the PV cells or the yellow ones from the Joule Thief circuit boards,
Now either test this in the sun or use a flashlight . Compare the voltage when the light shines evenly on both photo-cells versus when it shines on each of them one at a time.
In my first test I got 1.5V when illuminating the left-hand cell, and MINUS 1.2V when illuminating the right. The voltage went close to zero when they were equally illuminated. The discrepancy turned out to be from one poorly performing PV cell so I swapped it out with another one I swiped from my garden and then they both output almost identical voltages for the same illumination.
(The raw output directly from the good PV cells was about 3.4V and from the bad one, 2.5V.)
Step 4: Outdoor test
Here I've connected the circuit to a voltmeter (as I mentioned earlier, I doubled up the cells to get more power, but it didn't really help) and took 5 readings, from well left of the sun, through half as far left, then straight on, then the same angles to the right of the sun, You can see from the readings on the voltmeter that this is a good indication of which way you need to turn and when you're pointing in the right direction.
(You'll note that the absolute values of the signals are pretty low - even when one sensor was blacked out completely and the other was in full South Texas sunlight, the best differential I've seen was 1.65V.)
Step 5: Now I need your help to control a motor with this circuit!
Now, I had hoped that if you feed this DC signal into a motor, it would spin one way or the other or not at all depending on whether the incoming voltage was positive or negative. (DC gearmotors are reversible by reversing the input polarity). This motor would rotate the platform holding the PV cell sensors as well as a larger solar panel such as a USB phone charger, and when it rotated enough to face the sun, the outputs from the two PV cells would balance and the motor would halt due to receiving 0V.
But unfortunately that plan didn't work because the PV cells didn't generate enough current to drive the motor.
I have some ideas about how we can salvage this part of the design which I explain in the next section, but this failure does make the system more complicated and far less cost effective. (Unless one of our readers can come up with a solution that will save the day?)
The best way to use the two PV cells as a direction finder is to exaggerate the amount of light falling on one versus the other when the cells are side by side but not pointing directly at the sun. The standard way to do this is to put some sort of divider between them like a piece of card, pointing forwards, which casts a shadow when the light source is to one side but which has no relevant shadow when the light is directly in front of the equipment.
A subtle improvement over this is to use a double-sided mirror. This has the advantage that more light is caught by the cell that is facing the sun, which may be a factor in the early morning when the sun is just coming up, especially if we can also find some way for these cells to provide enough juice to drive the motor that rotates the platform as well.
When this apparatus points in the direction of the sun, the light falling on each cell is equally strong and the voltages generated by each cell cancel out, so you'll see 0V on your meter. Turn the meter left and right and you'll see the voltage go up or go below zero as in the photos attached.
By the way, I can't take all the credit for this idea - I based it on the underlying principle behind a circuit that a friend designed back in the 70's: he used a couple of infra-red sensitive photo-transistors tied emitter-to-emitter to detect a person's body heat, in order to turn a model skull (at least I hope it was a model...) so that it would follow people as they walked through his room! Basically a Halloween Haunt trick way ahead of its time. Where-ever you are now, Steve McGloin, that was a brilliant idea you had! This hack uses the same principle, of pitting two current sources head to head and letting the stronger signal win.
Step 6: Hack a TrekBot toy for motor parts?
(The picture for this section is just a mock up of a USB charger panel on a platform with the two PV sensors. I considered making a rotating platform from a CD stack case; I also looked at using a Lazy Susan cake stand. But until we work out a way to drive the analog DC motor cheaply, there's no point in wasting time working on the platform.)
I mentioned that initially the two cells I used were mismatched. One generated about 2.5V max voltage rather than the other's 3.4V, so the circuit was far from 0 when the illumination was balanced. ideally you would pick two cells that matched better, but... it doesn't actually matter that much, because (assuming you have removed the Joule Thief circuit that's trying to regulate the charging voltage) your motor and platform will still turn until the two outputs are balanced. Although your cells may not be pointing directly at the light source, you can skew the platform to a slight angle relative to the sensor cells so that what is on it points directly at the light once the platform has stopped rotating - this is the beauty of negative feedback analog control systems!
What I've described here would just turn your solar panel platform east or west in the direction of the sun - but not up or down as the sun arcs across the sky during the day. However you do not need a second motor to follow the changing elevation - if you build what's called a 'polar mount', you can use a passive physical mechanism to change the angle of tilt rather than any sort of powered motor. (Attaching the sensors to the mount along with your primary panel, so that they are also tilted to the optimal azimuth wouldn't hurt either, but isn't necessary - they're quite forgiving)
Finally - I do have an idea how to resurrect the idea of driving a motor from the PV cells (rather than stealing power from the larger 'payload' panel that the platform would be supporting): I have a $10 toy that I bought from Woot for reuse as parts - a "TrekBot" which contains a couple of motors and some good gearing. It charges up from a handheld controller that contains three button cell batteries - I suspect it has an internal super-capacitor rather than a battery. The wheels have enough torque that I was able to spin a Lazy Susan just by friction-driving the wheel against the rim of the platform. If we could make the PV cells charge up something like this, then switch to voltage-comparator mode once it was charged up, it could adjust the platform's orientation in a more bursty manner than the original continuously-updating design. This level of analog design is probably beyond me, and I really don't want to use a PIC in this device, even if it is possible to buy the bare chips for a couple of dollars. It just strikes me as inelegant :-) But I may end up going that route if it's the only way to do this on cheap. At the moment, the rotating platform looks to be the most expensive component.
Actually that's not quite true. The USB charger I have - made by Soldius - appears to cost around $100 nowadays. I swear I got it for a fraction of that some years ago, though I forget now how much exactly, Maybe $20. But if this solar tracker ends up costing $20, it'll be a failure. We need to come in around $5 to be worth the effort.
That's as much as I have to contribute. I hope you can use this as part of a new tracker design. Good luck!