I’ve seen a lot of Xee2’s xee2vids on Youtube, but I don’t remember seeing this one before. It’s dated Sept 2011, so it’s more than 2 years old. The Joule Thief flasher obtains its power from four red LEDs connected in parallel, which are used as solar cells and charge a 10 thousand uF capacitor. There is a ‘RK44’ diode between the LEDs and the capacitor, presumably to prevent current from discharging back through the LEDs AKA solar cells. But I don’t believe it’s necessary. The surface area of the LEDs, which I’d estimate to be a millimeter on the side, or one square millimeter, is one millionth of a solar cell that is one meter square. Consequently the leakage would be one millionth of that of a large solar panel, and any leakage should be unmeasurable. In any case, I’ve never known a LED to conduct measurable current in the reverse direction. I don’t know what a RK44 diode is, but by eliminating it we save a half volt or more of voltage drop.
But then I thought why put the four LEDs in parallel? Instead, put them in series and add the voltages up, to get four or more volts and that will allow us to change the circuit to a common two transistor flasher. Yeah, I know: I’m taking away all the fun with the Joule Thief. But it has about 50% losses in the coil and circuit, and eliminating the coil will save power. It may be possible to run two or maybe more 2 transistor flashers from the LEDs. But I digress, so back to the Joule Thief.
The 200 k resistor determines the flash rate along with the 3.3 uF capacitor. This also determines the current drawn by the capacitor, and the LEDs can only put out a very small curent. So if the flasher draws too much current and flashes intermittently, then it may be necessary to increase the 200k, and to get the flash back to its intended speed, reduce the 3.3 uF capacitor.
Note that the 3.3 uF capacitor has no polarity, there is no plus sign. This means that it should be non-polarized. But if a polarized electrolytic is used, it looks to me like the plus side would be at the bottom, which is connected to negative through the coil. However the capacitor may charge up in the opposite direction, so it would be wise to check the voltage across it to determine the correct polarity. Or else use a non-polarized cap as shown. If the resistor is high enough, the capacitor could be as low as 1 uF, which is available as a non-polarized type.
I have blogged my Supercharged Joule Thief Flasher, which presumably is more efficient than a conventional JT flasher. But to be honest, it is very difficult to measure the average battery current of any flasher, because the flasher pulls a pulse of high current from the battery, then waits a long time before pulling the next high current pulse. In order to average out the current, it requires a very large capacitor to filter out the current pulses, and get a stable average current. A digital ammeter will have the digits jumping around and it’s only a guesstimate of the actual average current. But flashers use only a very small current, so the battery lasts a long time. Most people would not be very concerned if the battery in a conventional JT flasher lasted only four months, and a supercharged JT flasher 7 or 8 months. Either one seems to be a long time.
I soldered up one of these on the bench this weekend. It works, but I have to get it right next to my 3 watt LED task light to charge up the capacitor. The main difference between this one and the one in Xee2’s schematic is that I didn’t use the RK44 diode. I also used the BC338 transistor, which is higher gain I believe than a MPSA06. Instead of 3.3 uF I used a 1 uF plastic capacitor. In order to keep the flash rate low, I had to increase the value of the resistor from 200k to 680k. As it is, the circuit works okay, but it draws enough current from the 10000 uF capacitor that when the LEDs are not charging it, the voltage drops over a few tens of seconds and the LED stops flashing. Discharging is to be expected, but the discharge speed seems higher than I would expect. One reason may be that the LEDs are letting current flow backwards, but I thought that this would be minimal.
Update Nov 18 – I got the stopwatch and made some measurements. The flash rate was about 2 per second. The inital voltage was 1.6V, and when I turned off the light, the 10000 uF capacitor took two minutes and fifty seconds to discharge to 1.4V. I think the cap had been sitting around for awhile and needed to have a charge or two put on it to ‘reform’ its oxide layer. Electrolytics may have a bit of leakage when they’re first charged up due to the oxide layer unforming during storage. But after awhile they reform and the leakage goes down to what is normal. All electrolytics have some slight leakage. The larger the capacitor, the greater the leakage, and 10 thousand microfarads is a very large capacitor, so leakage should be expected (I didn’t think about it earlier – my bad). So it now seems that the leakage backwards through the LED wasn’t a problem at all, it was the capacitor.
There are two things that I thought about after building this. First off, I don’t need solar cells to make my Xmas flashing house decorations run off free energy. They are now running from a single AA or AAA cell, which powers them for months, but the batteries eventually need replacing and they always leak and corrode all of the wiring that touches them. Replacing the battery with a capacitor and several LEDs is a great way to eliminate those problems. One thing has to be considered: the running time. I’m guessing that the 10000 uF capacitor, which is actually 0.01 F, is only going to run for 10 or 20 minutes before it’s discharged. So it’s going to take more than a dozen of them to get the flasher to run after dark until midnight or later. That’s too many to put onto a small flasher. I will have to obtain some 1 farad capacitors, and since they’re only 2.5V, the LEDs still will not overcharge them at 1.6V.