I don't know if I'm being silly, and should just use a series resistor to power my Luxeon LW3C LED from a 6 V power pack. It just seems a bit over the top to try to increase the efficiency of the circuit from 3.7/6.0*100 % = 62 % to maybe 90 %. Yet it seems one of those challenges that can't remain unanswered.

The schematic shows both solutions, the "high tech" current switcher (buck converter) and the "low tech" resistor. Hm. OK, this is how it works. The LMV7239 comparator is connected as a Schmitt-trigger circuit with a little hysteresis. The delay of the chip itself (45 ns propagation delay) causes the resulting switching frequency to be a bit lower. Now, when the power is applied current starts to flow (the capacitor in the FET's gate is pulled to ground so it is switched on) the voltage at the comparator's + input also rises. At 90 mV the circuit toggles, the output goes high and the FET is switched off. When the voltage across the current sense resistor (0.15 Ohm) has dropped below the threshhold voltage the circuit toggles again and it all starts anew. The reference voltage is not stabilized so the current drops as the battery gets depleted. The diode in series with the supply voltage is a protective measure. As the comparator's maximum rating for the supply voltage is only 5.5 volts and my power pack cranks out 6.6 V when fully charged it has to be reduced. With its 0.6 V drop the diode reduces the supply voltage to about 6 volts. Still too much but my test sample stayed happy with it. This also explains the ridiculous arrangement for driving the FET. With the lower VDD it won't switch off properly. *Sigh* Just can't seem to keep it simple!

The resistor is easy: apply voltage and current flows. Resistor gets hot.

So back to the switcher. All my favorite electronicks are SMDs nowadays. Some people are afraid of these supertiny parts, but with magnifying lamps and eye magnifiers this is not an issue. Maybe shaky hands is... In any case, it does allow for a really small circuit. After some deliberation I decided on a piece of epoxy perfboard measuring 13 by 13 mm. While not ideal in the sense of robustness it is really handy to whack it together without too much effort. So I did. With my trusty Weller (with the tiniest tip I had!) I soldered everything on the little board. The fat coil came last, on the other side. (Rats! No pic of that!) Incidentally, this coil came from the power supply of a discarded satellite digibox. The schottky diode was liberated from the same laptop backlite converter used in my reading light project. This converter also had an SMD coil on it but it was too large in diameter. Note the way the ground pin of the comparator is wired. What looks like a fat blob of solder actually is a tiny ball! It will probably disappear when I add a decoupling capacitor across the chip.

Final results: I'm a bit disappointed with the amount of light generated. It is a wider spot, but the intensity is about the same as with my old 2.4 W halogen beam. The bracket also gets pretty hot. And I haven't found a way to mount it in a headlight fixture, yet. There's also a parasitic effect: apparently due to insufficient decoupling of the fast comparator's supply voltage it is sometimes prone to erratic oscillation. Funny enough this has no effect on the light whatsoever. The circuit does get hot, while in normal operation it gets only slightly warm to the touch. A propos warm: the efficiency of the switcher is a phenomenal 95 %! The switching frequency is around 90 kHz, quite a bit lower than in the simulation. But that is probably due to the different comparator used in the simulation.

Latest update: It turned out that the capacitive coupling to the FET's gate was not such a bad idea after all. It nicely provides undervoltage lockout when the battery voltage drops below four volts. If this happens, the comparator wants to switch on the FET fully as the pulses disappear (100 % duty cycle). The capacitor prevents this. In case of semiconductor latchup inside the comparator this will prevent damage to the LED as well. So far I haven't observed this (I don't like to increase the supply voltage of the comparator too much over its absolute maximum rating of 5.5 V) but if it happens this fault condition won't cause a loss of an expensive Luxeon. Now I can skip trying an LMV397MF, which is a comparator with open collector (drain) output but 30 V max supply voltage. Open collectors are BAD as they will slow down the FET switching off which leads to extra losses. This will wreak havoc on my efficiency! I also added that dreaded decoupling capacitor and this solved all stability problems. There's only a little glitch left at the transitions. Must be a layout issue...














© Zappy TV 2005