I’ve talked about this before: the Joule Thief has certain vulnerabilities that need to be considered when making a new design. Suppose we build a simple conventional JT, with a single LED and a 1.5V battery. The coil has two windings, each the same number of turns, probably wound bifilar – the windings are wound at the same time and have the same number of turns. This means that the windings have a 1 to 1 ratio, which means the voltages are the same on each winding.
When the output of the JT goes to a single LED, the voltage across the LED will not be more than 3 to 4 volts. The coil reflects this voltage back to the feedback winding, which means the negative voltage on the base will not be more than the LED voltage (actually less). This 3 to 4 volts negative will not exceed the maximum for most transistors, which is typically 5 volts.
This simple JT design is operating properly, within its limits. There are several limits, one of which is the maximum voltage across the transistor. During its operation, the LED forward voltage drop of 2V for red or 3.2V for blue or white means that as long it is in the circuit, the LED will limit the maximum voltage to 3.3 volts.
But some experimenters disconnect the LED and expect the voltage to rise, and it will rise (see note). They believe that the transistor’s maximum collector breakdown voltage will limit the voltage. This voltage is generated by the collapsing magnetic field within the coil. The collapsing magnetic field is also inducing a voltage in the feedback winding, of the opposite polarity. The negative voltage is on the base, so when this voltage rises, it will reach the maximum emitter to base voltage, which is much lower than the maximum collector voltage. The emitter to base junction will break down, and current will flow and limit the maximum voltage before the maximum collector voltage is reached.
The problem is that when the emitter to base junction breaks down and a small current flows, it causes the current gain to deteriorate, so after a short while, maybe tens of seconds or a few minutes, the transistor’s current gain can be measured and it will be much less than the current gain before the breakdown. And this damage is permanent; the transistor will never recover its original current gain. Some circuits tolerate this loss without losing their usefulness. The Joule Thief seems to keep running without stopping. But the transistor’s performance suffers, so it must affect the Joule Thief’s performance somewhat, even though the JT continues to operate. If this abuse continues, my guess is that the performance will continue to deteriorate, but when it will no longer operate is anyone’s guess.
My opinion is that this abuse should not be allowed to occur. As long as the LED is here as a load, the abuse will not occur. But some experimenters want to operate this simple Joule Thief with more than one LED. They attempt to operate it with two or more LEDs in series, so the voltage across the LEDs might be 6.6, 9,9 or more volts. The same voltage, but negative is on the base, and that exceeds the 5V maximum. Those experimenters all fail to take precautions to protect the base from this negative voltage. The protection is easy to do. All that is needed is a diode (1N4148) with its cathode or banded end connected to the base, and the other lead connected to the emitter. When the winding applies a negative voltage, the base goes negative to 0.6V, the diode conducts and the negative voltage cannot go any more negative. The excess energy is dissipated in the 1000 ohm resistor. Instead of a diode, a LED can be used. When the negative voltage gets to 3.3V, the LED conducts and the same thing happens, but it will light and show the experimenter that it is doing its job. The negative 3.3V is well within the 5V maximum so the base is protected.
This form of protection should be used on all Joule Thiefs that the experimenter might abuse.
Note: When the current through the coil is shut off, the collapsing magnetic field will produce a high voltage. If the circuit is changed to the similar two transistor voltage boost circuit with a single winding on the coil, the voltage could rise to as high as the collector breakdown voltage for the transistor. This could be a hundred volts or more for a high voltage transistor. But when the second feedback winding is added, the voltage across it gives the coil a “sneak path” for high voltage to go when the field collapses. If you are experimenting and do not want this other escape route then you should not use the two winding design.