A long time ago, a pair of white LED + red laser flashlights powered by an AA cell diverged: one flashlight worked fine, the other always had a dead battery. The latter ended up on my “one of these days” pile, from which it recently emerged and accompanied me to a Squidwrench Tuesday session:
The black wire trailing from the innards goes to the battery negative terminal, with the aluminum body providing the positive terminal connection to the wavy-washer spring contact visible atop the rear PCB inside the front shell.
The switch connects each red wire to the battery negative terminal, so there’s a color code issue in full effect. The two red wires burrow through holes in the rear PCB (shown above) and connect to the negative terminal of the laser module (the brass cylinder near the top) and the negative terminal ring on the front PCB holding the seven white LEDs:
Continuing the color code issue, the black wire from the laser is its positive terminal. The out-of-focus wire (an LED pin) sticking up near the top of the picture carries the positive connection to the LED ring. The red wires from the switch are the negative connections for the LEDs and laser.
Voltages applied to the LED ring and the currents flowing therein:
Seven LEDs at 20 mA each = 140 mA, so the voltage booster must crank out slightly more than 3.2 V. They’re not the brightest white LEDs I’ve ever seen, but suffice for a small flashlight.
A crude sketch of the PCB layout, with a completely incorrect schematic based on the mistaken assumption the SOT23-3 package was an NPN transistor:
Obviously, that’s just not ever going to oscillate, even if the 2603 topmark meant a 2SC2603 transistor, which it doesn’t.
A bit more searching suggests it’s a stripped-down Semtech SC2603A boost converter, normally presented in a SOT23-6 package. If you order a few million of ’em, you can strip off three unused pins, do some internal rebonding, and (presumably) come out with an SOT23-3:
That topology makes more sense!
Before going further, I had to rationalize the colors:
Soldering longer leads to the PCB allows current & voltage measurements:
With the LEDs and laser disconnected, the converter seems to be struggling to keep the capacitor charged:
Those purple spikes come from the current probe at 200 mA/div: maybe half an amp in 5 μs pulses at 6 kHz works out to a 15 mA average current, which is pretty close to the 11 mA I measured; it’s not obvious the Siglent SDM3045 meter was intended to handle such a tiny duty cycle.
Obviously, the output capacitor is junk and, after removing it, the AADE L/C meter says NOT A CAPACITOR. Perhaps it never was one?
Measuring the cap in the good (well, the other flashlight) suggests something around 100 nF, so I installed a random 110 nF cap from the stash. The current peaks are about the same size:
The cap voltage (not shown) is now nearly constant and the 50 Hz PWM rate reduces the average battery current to 100-ish μA:
Not great, but tolerable; a 1000 mA·h battery will go flat in a few months.
The LED current runs a bit hotter than I expected:
The bottom is about 200 mA and the average might be 350 to 400 mA.
Compared with the other flashlight:
So the cap is maybe a bit too small, but it likely doesn’t matter.
Done!
The Smell of Molten Projects in the Morning 