I found this single cell white LED torch (flashlight) from EDN. The author claims it is efficient.
I like the Q5 active rectifier concept. It seems that designers are doing this more often, especially with high power designs. But they use MOSFETs, which do not draw gate current like a BJT draws base current. Since this is a low power design, every milliwatt not used is a milliwatt saved. So the Q5, Q4 and 3.3k resistor must waste some power, but is it less than a Schottky diode alone? I can’t guesstimate on this; I would have to measure both circuits.
That having been said, my theory has always been that adding a rectifier and filter capacitor is not necessary and is another point where power is lost. A single LED without any other components is the least lossy, in my opinion. However others have claimed that adding the rectifier and capacitor makes the led brighter, and I have confirmed that it’s a valid claim. I don’t know how this happens, though. To me it contradicts common sense.
In a conventional Joule Thief, the oscillator uses feedback from itself, so the frequency or pulse rate is determined by the oscillator itself. It runs at whatever frequency is optimum for the coil, mainly. But in this torch design, there is no feedback from the coil. The frequency (and waveform) is determined by the astable multivibrator Q1 and Q2. Therefore there is no way for the circuit to determine if the frequency is optimum for the coil. This may be okay in a product where the coil and frequency are optimized for the circuit. But for the experimenter, who may not have the exact parts, this could mean that the circuit may never get optimized. The 100k variable resistor may never adjust it to its brightest and highest efficiency.
I think the only way to find out is to temporarily disconnect one of the 1 nF capacitors and connect a pulse generator to the base. The range of frequencies can then be determined for the particular coil used. Or else connect the pulse generator to the left side of the 3.3k resistor to drive the high current transistor directly. The frequency and width of the pulse can then be controlled to get the best results for the coil, and then the parts values for the multivibrator can then be changed to duplicate the pulse generator.
Back to the circuit. The standard JT or the two transistor voltage boost circuit are simpler, and do an adequate job. I’ll have to ask the author or wait for someone to tell me why the design needed this astable multivibrator. The coil is not tuned, so given a coil of a certain inductance, the designer could have optimized the circuit for that coil and then eliminated the variable resistor.
More to follow, if I can get time from my other projects.
