Kirk left me a comment asking how much inductance is too much. I’ll try to go into detail as to how the inductance and other factors affect a Joule Thief. First a couple things should be defined. This discussion is confined to a conventional JT, with a single transistor and a coil with two windings: a primary winding and a feedback winding. Also, the battery voltage is limited to a maximum of a single cell, which is about 1.55 to 1.6 volts when the cell is fresh, but averages about 1.5V during most of the cell’s life, then tapers off as the cell becomes depleted. We will assume that the JT is designed to be run at 1.5V. Also the LED is a typical white or blue with a Vf of 3.3V.
The job the JT has to do is transfer a ‘bucket’ of electromagnetic energy from one voltage, 1.5V, to a higher voltage, typically 3.3V for a white LED. The typical 5mm white LED works optimally at 20 milliamps, which, multiplied by 3.3V, equals 66 milliwattts. So we want the JT to give the LED a ‘bucket’ of energy equal to 66 milliwatts, and in doing this, the JT itself will use somewhere around that much, as losses in the transistor, resistor, coil windings and core. So, for example, the total power we pull out of the cell averages 120 milliwatts divided by 1.5V, a current of about 80 milliamps. But the transistor is turned on only about 30 to 40 percent of the time, so the transistor actually takes a gulp of current of 200 to 250 milliamps.
Ok, we go to the Wikipedia entry for inductor and read about it. It says that Estored is equal to one half times the inductance times the current squared. If you wind an inductor, the more turns it has, the more inductance it has and the more energy it can store. But as you wind more turns on it, the resistance of the wire increases and the higher resistance reduces the current, since we are working with a fixed 1.5V. If the wire resistance is doubled, the current is cut in half, but going by the above formula, the energy stored is 1/4 (current squared). In order to increase the number of turns but keep the resistance the same, we have to use thicker, heavier wire. As we add turns, the inductor grows from the size of a pea to the size of a grape, or even bigger, the size of a golf ball. The inductor becomes so big that it takes up more space than the rest of the circuit. This can be a problem if we have limited space. Also the heavier copper wire costs more.
But instead, we decide to go the other way. If we reduce the number of turns to half, the resistance is reduced to half, and the current can double. But since the energy stored is equal to the current squared, the double current gives us four times as much energy stored.
So if you want your inductor to fit on the corner of your PC board and not take up the whole board, then it’s better to have fewer turns, lower inductance and lower resistance so that the 200 or more milliamps can flow and give us the stored energy equal to 120 milliwatts. As we see above, there is nothing wrong with increasing the number of turns, as long as the wire is heavier so that the current does not decrease. This means that the inductor will get larger, but if you have no problem with a large inductor, then the inductance can be as much as you want. I’ve seen chokes that have more than 1 henry that would work, but the resistance of the winding would be high enough to limit the current to a few tens of milliamps or less. The JT will draw lower current and the LED will not be full brightness.
So we have a choice of an inductor with a lot of turns of heavy wire, with a weight of several pounds or kilograms and is larger than your fist, but it works well in a JT circuit. The frequency may be low, but your eyes can’t tell that the LED is not on all the time. Or we can choose a much smaller inductor that fits on the PC board and has an inductance of a fraction of a millihenry, but the resistance is low, the current is high, the LED is brightly lit, and the frequency is several tens of kilohertz, but again the eyes can’t tell that the LED is not on all the time. The choice is yours. If you have no choice and have to use a very large inductor, then you’re limited to what you have. But given the choice, most people would choose small, so it will fit on the PC board.
Type of Inductor
We can use a toroid core, because it can be small, have a relatively high inductance and a low resistance, and only a dozen or so turns, so it is easy to wind. The inductor doesn’t have to be a toroid core, but they’re easy to find in old, not working PC power supplies or a CFL light. I would just order some cores, such as the Fair-Rite 2673002402 for 11 or 12 dollars U.S for a bag of 100 from Mouser.
A few days ago I blogged some cores I scrounged from CFLs. They are small, and seem to have a good amount of inductance with some small wire, which is short so it will have low resistance. For the neophyte, the CFL seems to be a cheap source of cores in small quantities. The dollar store sells new CFL lights for a dollar, so you can get a toroid core and several other parts from each light, making it a reasonable cost, readily available source for JT cores. I used a small blade screwdriver to fit in the groove that is where the two halves of the case snap together. A couple twists and it will start to come apart. I snip the wires and voila! I have the guts laying in my hand. That sounds a bit unappealing, doesn’t it?
Conclusion
So the answer to Kirk’s question is really about how much compromise you have to make when it comes to cost, size, energy storage, resistance and construction of the inductor, such as toroid, bobbin, air core, etc. Most will make a decision towards the smaller size, so they will have to compromise in the amount of inductance. I’ve found that 100 microhenrys is a reasonable amount for a conventional Joule Thief, but this can vary widely. The inductance can go below 100 uH, but you have to remember that as the inductance decreases, less energy is stored. Also, the energy ‘buckets’ are smaller, so in order to keep the LED the same brightness, the frequency has to go up to transfer more of the smaller ‘buckets’. The frequency can go up, but if it gets above a few hundred kilohertz, there is an increase in the chance that the frequency or its harmonics will cause interference with the AM radio band.
Another factor is the size of the wire. I often use regular 24 AWG solid insulated telephone wire which can be removed from old telephone cable or even from cat5 cable (it’s awfully kinky, though). It’s cheap (free!) and easy to find. The insulation makes it thicker so fewer turns can be put on to a toroid core. If enameled wire is used, the insulation is very thin and maximizes the number of turns that will fit on to the toroid.
Thus we have multiple factors that have to be considered when making these decisions. Inductors allow the experimenter to customize their component by winding their own. With the information I’ve given, the experimenter can make some informed decisions on how to best wind their own.
Back to experimenting…
I’ve successfully used 150uH common mode chokes scavenged from power gear in e-waste bins at work. No hand winding necessary, and they drive multiple 20mA ultrabright white LEDs with no problem. I fabbed a series of circa 1cm square PCBs to assemble JT circuits on with these – coil, resistor, transistor, LED (or sometimes a socket to connect an LED or another board), and 2 pin header for power.
More recently, I’ve taken to winding 0.1mm enameled wire around the body of a 100uH axial inductor. These are dirt cheap, readily available, easy to wind (anchor and spin – loads easier than a toroid) and results in a compact coil.
There are several ways to assemble these – obviously direct to a PCB is the easiest way, but for breadboarding, it’s easy to solder one lead of the winding to the axial lead on the opposite end, making that the supply side of the coil, and the other end is soldered to a 1Kohm resistor, which becomes the “leg” for that to push into a breadboard (and subsequently runs to the base of the transistor). It’s nominally more power efficient to have the supply side of the winding run to the resistor (so the flyback action in the coil snaps the base fully off quicker), and solder the base side of the winding to a clipped component lead, but this leaves you with two winding contacts to break if you mishandle it instead of one.
The slightest daub of clear acrylic let to dry will hold the windings.
If the chokes are very inexpensive, then consider using two of them, like I did in this blog.
By hand-winding the second winding onto the inductor, you should recover more of the energy from the coil collapse into the second winding, which should cut the base off quicker, which means it toggles between full open and close quicker, spending less time in an intermediate state, where the transistor is dissipating energy across the Collector-Emitter junction.
The assembly process is quite straightforward – certainly not as trivial as just soldering two inductors down to your PCB, but still less than a looping windings on a toroid, and it results in a very compact coil.