Illuminating more than one very bright white LED requires a choice of configuration, either series or parallel. Of course, each configuration has design tradeoffs.
A parallel connection requires a lower voltage across each LED, yet ballast resistors or current sources are needed to accomplish matched light uniformity. Different levels of bias current, and thus light coming from each LED, create a disrupting light source. However, light matching with ballast resistors or current sources decreases battery life.
A series connection inherently has great current match, but it requires a higher voltage across the LED string. Common white LEDs are biased at 3.6 V at 20 mA (max) for useful illumination. The circuit in Figure 1 shows an inexpensive inductor boost circuit that regulates the brightness of a string of seven white LEDs.
The circuit is best described in two parts: one, the boost circuit function of Q1 and Q2, and two, the control circuit of Q3 and JFET1. Assume Q1 is off. With the battery voltage slightly above Q2's VVB, a positive Q2 base current \[iB = (battery voltage VBE)/RJET1\] would flow. Q2 turns on, which switches inductor L1 to ground.
Energy stored within L1's magnetic field builds as L1's current rises with a positive di/dt. As this current rises, it also flows through Q2's RSAT. (SD1 and the LED string are off.) Q2's collector voltage becomes sufficiently large to turn on Q1. Q1's base voltage is connected to Q2's collector by the feed-forward network of R1 and C1. R1 also serves as Q1's base current limit.
With Q1 turning on, the previous base drive to Q2 is now shunted to ground, and Q2 turns off. The switching off of Q2 enables L1's energy to be discharged into the LED string as the magnetic field collapses.
This flyback action of L1 forward-biases the LEDs at greater than 26 V, which emits photon illumination in the form of white light. The human eye integrates the LED's flashing frequency into a constant illumination. With L1 discharged, Q1 turns back off.
The self-oscillating action repeats under normal operation unless the battery voltage falls below Q2's VBE plus the IR drop of JFET1 (about 1 V). Then, Q2 no longer turns on. L1, Q2's RSAT, and the switching characteristics of Q1 and Q2 also contribute to the period and duty cycle of oscillation.
Figure 2 shows the flyback voltage at Q2's collector for battery voltages of 6 (color A curve) and 3 V (color B curve), using a Coilcraft DO1608-104 inductor and Red Line RL5-W10015 LEDs. This flyback voltage waveform is averaged by SD1 and C2 into a dc voltage of about 23 V.
Through R4, a small dc current (less than 20 A) biases Q3 as a VBE multiplier, which adjusts the channel resistance of JFET1 and in turn regulates the battery drain current for longer battery life. The voltage at the gate of JFET1 operates about 0.9 V above the battery-stack voltage. A p-JFET operates as a depletion-mode device. Its p-type channel conducts with zero VGS.
The source is connected to the battery terminal. Designers pull up on the gate (more positive than battery voltage) to turn off the channel. The more the gate is raised above the battery, the higher its channel resistance. This RJFET1 previously was discussed with the Q1 and Q2 oscillation.
So as the battery-stack voltage drops from 6 to 3 V, the boost frequency decreases (slight change in VGS of JFET1). With it, the LED brightness decreases slightly. Ideally, the control loop would keep the LED current constant. But human sensitivity to light is on a quasi-log scale. Therefore, the small linear decline in illumination isn't very noticeable until about a 2-V battery stack.
An alternate scheme would hold the battery power (voltage times current) constant. This would keep the LED brightness constant but shorten battery life due to the internal battery-resistance losses. Also, the circuit complexity would increase drastically. As a result, this simpler circuit's LED brightness varies a little over battery life and follows the battery current.
The LED string brightness can be slightly adjusted. Designers can adjust for manufacturing variations in the transistors and LEDs by making small changes in R2. Light output (lumens) then can be set to a fixed point from one circuit to the next.
At an end-of-life battery-stack voltage, a dim LED string could be shorted to connect just one LED, which would be very bright with as little as 1 V left on the battery stack. This one-LED connection would provide last-hope emergency light with the deadest of batteries.
Alkaline battery safety dictates the use of matched batteries. The safety problem arises when the weakest battery in the stack becomes completely discharged and the other batteries have enough energy to voltage-reverse the weakest, causing it to overheat and leak milky acid.
To accomplish matching, always change all four batteries with new batteries from the same package. Four AA alkaline batteries have a capacity rating of 4 1000 mAh, which would mean that the LEDs would be illuminated for about 61 hours. Test results of prototype units continuously illuminated for a little over two days (48 hours).