RustyBolt.Info/wordpress

Background:
I should explain what I’ve been considering. My door has southern exposure so I would like to put a solar cell on it along with a long-running joule thief. The LED will be pointed at the keyhole in the lock. During the day the solar cell will charge a supercapacitor and run the Joule Thief after dark. When it’s dark, the keyhole will be easily seen and opened. I have another option and that is to use a battery. This could be a AA cell or AAA cell. I have been experimenting with low power Joule Thiefs and their battery life can be a month or more. If I use a rechargeable cell, the solar cell could recharge it during the daytime.

I found a Joule Thief on Youtube from xee2vids that he calls Self Running.  This JT has two things that are different than the conventional JT: the cathode of the LED is connected to the positive, not the negative.  Also the resistor is much higher; he uses 51k.  Also a small value capacitor is across the resistor.   In the version I built I used 150 k resistor and a 470 Picofarad capacitor.

The reason for connecting the LED cathode to the positive is so that if the voltage on the input rises above the LED’s forward voltage, nothing happens.  If it was connected to the negative then the higher supply voltage would light the LED without the Joule Thief.  In this case Xee2vids and I are using 6 volts for the supply voltage and the LED must be connected to the positive to prevent it from lighting.

The reason for using such a high supply voltage is because xee2vids’ circuit uses a 10000 microfarad capacitor for the supply.  In order to make the Joule Thief run longer he charges it up to 6 volts.  In my case, I used a bank of 1 Farad super capacitors charged up to 6 volts.  I connected nine 1 farad supercapacitors in series parallel to give 1 farad at 7.5 volts. The JT has a Fair-Rite 2673002402 coil core with turns of wire for each winding. The transistor is a PN2222A.

I charged up my JT to 6 volts and timed the lit LED. At the start it was moderately bright. After 30 minutes the voltage had dropped to 2.4 volts and the LED was still moderately bright, just a little bit dimmer. After 45 minutes the voltage had dropped to 1.66 volts and the LED had dimmed noticeably and the light was not good enough for illumination.

My conclusion is that this circuit can run for about a half an hour with a one farad capacitor initially charged to 6 volts. If I increased the capacitance to 10 Farads, the led should stay lit for about 5 hours. A capacitance of 15 Farads should give about 7.5 hours, which is enough to keep the circuit lit from sunset until midnight on a winter day.

A typical Solar garden light runs about this long using cheap nickel-cadmium rechargeable AAA cell. It’s a lot cheaper then 15 farad capacitor but the capacitor doesn’t lose its capacity after a year or two. Obviously this price difference is the reason why solar garden lights don’t use capacitors, they use cheap rechargeable batteries.

To get 15 farads at 7 volts, I would need to put 3 capacitors in series, each capacitor would be 45 farads. The closest value would be 50 farads. This capacitor would be about the size of a single AA cell, and three of them of course would be much larger.

I have to design a small enclosure that can be fastened to the door and hold the LED properly pointed in the right direction. Most of the room inside the enclosure will be taken up by the super capacitors.

Quantsuff reminded me of his Hak Light in an email.  I haven’t visited his web page in awhile so my memory needed refreshing.  I checked it out, and I thought about another use for his circuit.  At the dollar store, they sell (when they’re not out of stock) the CR2032 or CR2025 coin cells on a card for a dollar.  I forget how many there are, maybe 5.  But they’re dirt cheap compared to Radio Shack where you pay $5 apiece.  I know that I could run a single LED with one coin cell.  But with his circuit I could put three of the coin cells in series and get a much longer battery life.  And it would act as a touch switch too.  The coin cells are good for about 5 milliamps. Three times 3 volts times 5 mA is 45 MW, and at 95% efficiency that gives 40 MW, which is plenty for a single white LED.

Also, three coin cells are much smaller and lighter than a 9 volt battery.  They’re cheaper too, but the dollar store sells two 9 volt zinc carbon heavy duty batteries for about a dollar or a dollar and a half.  If I wanted to stay at the same size as a 9 volt battery, I would use 2 AAA cells.

Back to QS’s circuit.  This is a small, cheap and simple circuit that is much more efficient than a single resistor.  He shows two circuits and the one I’m talking about is the one that uses a choke, not a coil.    It makes a lot of sense to use the circuit because it saves about one quarter of the battery power that would otherwise  be wasted with the resistor.  I  think an adequate choice for a transistor for Q2 would be the SS8050.  They’re only about five dollars for a bag of 100, and they can handle 1 amp or more, which should be enough.

This circuit has a weak point, it has no short circuit protection.  Over the last few years I have collected a few cell phone chargers that plug into the car’s cigarette lighter outlet.  These have died, and at least one of them burned up inside so badly that the values of the resistors could not be read.  The cigarette lighter plug has a fuse inside, but the circuit sacrificed itself to protect the fuse!

Another circuit that I built A few years ago was the one from Roman Black’s website.  It’s a 2 transistor circuit similar to this one.  I found out while experimenting that it, too did not like a short circuit on the output.  The short circuit stops the circuit from oscillating and the full input voltage may be applied to the output which can cause excessive voltage and/or current, and can burn up the circuit and whatever is connected to the output, too.  This shouldn’t be a problem as long as just LEDs are being used.  But the circuit really needs overcurrent protection to prevent it from destroying itself.

A .TXT file I had from 2009, probably from my late great Watsonseblog.  If it had a pic attached, I’d like to add it, but I’m unable to determine what picture.

Joule Thief using Air Core Inductor – Feb 13, 2009
I saw Quantsuff’s Valentine’s Day Flashing Heart on Instructables and it inspired me to experiment with some air core inductors.  The Joule Thief doesn’t have to have an inductor using a toroid, but the toroid helps a lot, as I’ll explain below. The coil has to have inductance so it can store a charge in the magnetic field and then dump it into the LED when the transistor turns off. You can use a coil of wire with an air core.

With a toroid core, the wire is much shorter and the resistance lower, so it will have less losses in the wire. And the coil will be much smaller, small enough to fit into the base of a flashlight bulb. You can’t do that with a large air core coil.  But if you have space for an air core coil, you can make a Joule Thief work with one.

Let’s say you use a toroid, and wind it with a foot of 24 gauge wire.  It might have an inductance of 100 to 300 microhenrys, which is great for a JT.   That wire will have a DC resistance of 25.7 milliohms or .0257 ohm.  That’s so low it will have very little loss.

In the instructable, QS said to use 6 feet of wire on a spray can lid.  I wound six feet of 24 gauge telephone wire on a spray can lid 2-1/4 inches in diameter. I then slid the coil off and taped up the windings. The winding had about 8.5 microhenrys inductance, which is quite low for a Joule Thief (typically 100 microhenrys or so), and much lower than what QS said it should be (22 microhenrys). The telephone wire has two wires twisted into a pair, one is white with blue stripes, the other is blue with white stripes.  You can use finer wire for the feedback winding, but the primary winding should be heavy wire to minimize losses.

The 6 feet of 24 gauge wire had a resistance of .15 ohm.  That doesn’t seem like much, but when you are working with only 1 volt and several hundred milliamps, that resistance can cause losses.  I calculated that it could drop about 1/20 volt, and have more than a milliamp of loss.  That’s more than 1 percent of the total LED power.

The coil drew about a hundred milliamps and ran at over 300kHz, which is too high for this circuit.  I didn’t think that a coil of 8.5 microhenrys was enough inductance. It should be 22 microhenrys, as QS stated.  So I wound another coil with ten feet of the same wire.  This one came out much better, it measured 25 uH.  When I taped it up, it was a bit higher, closer to 30 uH.   I connected this to the Joule Thief circuit and it drew 100 milliamps from the 1.5V supply and ran at 139 kHz.  This was much closer to what a Joule Thief should be.

But the DC resistance of ten feet of 24 gauge wire was over a quarter ohm, 0.257 ohm to be more specific. The loss was more than twice that of the first coil I made.  This is not a good thing.  What I really should do is use heavier wire, or put another winding in parallel with the primary winding.

Update Sep 13, 2009

I have since wound several more air core coils, some with 24 gauge insulated telephone wire, some with enameled wire, and some with both.  I have found that because of the plastic insulation on the telephone wire, the turns of the winding can’t get close together and this reduces the inductance and makes the coil larger, allowing fewer turns in the same space.  This is not a big drawback if the coil is wound on a core of, say, ferrite, since the permeability of the core increases the inductance and less wire is needed.  But when the coil is wound on an air core, the permeability is low, actually unity, and the wire has to be longer to get the same inductance.  Thus for best air core performance, it is best to use enameled wire with thin insulation.

I pursued another way of winding a coil.  I used 6 feet of four pair telephone wire, which when each pair was connected to the next, was equivalent to a single pair of 24 feet wound into a coil.  The wire is cheap and easy to find.  But connecting the pair to one another and taping them up takes time and is tedious.  This can be automated somewhat with telephone type blocks.  Another disadvantage is the insulated jacket puts more space between the wires, lowering the inductance somewhat.  And as in the above case, this increases the resistance per winding.  But I did manage to get a Joule thief working with this coil.  I used my Supercharged JT circuit.