It's been a little while since I've reviewed SEPIC, but I remember thinking at the time "Hey, THAT's the way to go!" -- it's not "the best of both worlds" (Step-up and step-down) but more like "a very acceptable combination of both worlds".
Step-down is fine, as long as your final (terminal) power-supply voltage is still higher than your V(out) needs to be -- otherwise, it just quits. You want to run the battery down to the last electron -- or you're wasting energy. If the step-down circuit quits before you've exhausted the battery, you're not getting the benefit of that electons. Also, the number of cells (if a multi-cell PS) enters into it... and may adversely effect the efficiency.
Fer example, the PAL... runs from a 9v battery. Burns a lot of electrons in heat, because of an in-efficient voltage regulator w/ who knows what drop-out voltage. Even if it were 20 mV drop-out, it's still inefficient, it just runs for a little longer. It will run from 9V (new battery) down to V(led)+V(drop-out). If it's a 1.8V drop-out, then you can run from 9V down to about 5.5V on the battery. Pretty low for a 9V. Now, figure the same thing w/ a stack of three 1.5V Alks... 4.5 V out, new. Get one of the really nice (but still in-efficient) low-drop-out voltage regulators - figure a 100 mV (.1V). The supply (batteries) will run from 4.5V down to 3.7V (for an LED flash... go figure). Well, you've gotten voltage (and current) regulation -- thus providing light and not torching that clear epoxy thingy, so you should be happy. 3.7V -- where the supply cct quits working, or, about 1.2V per Alk cell. Not efficient either in HOW the voltage is regulated, or in battery/ cell usage. Use 4 cells, and you get the per cell voltage down to .93V before the cctry quits working -- better, but still a LOT of juice left in those batts. You'd have to rig up a 5-6 cell stack to get the batteries to their exhaustion point (er, as considered by the battery companies) .6-.7V.
Maxim does have some step-up/step-down chips, but I don't recall if they do both at the same time, or if the chip can just be used in two diff't configurations.
I'd rather do the step-up, because I've seen ccts that will work down to .9 or .7 V input, so if you use two 1.5v cells, you really DO exhaust them... I'd be happier if there was a more efficient ONE cell solution, but the efficiency changes w/ load current (higher I(out) might mean much better efficiency) and source voltage (even if it works down to .7 V, it might run much more efficiently at 3V in than 1V).
But, that's part of what makes this so much fun to think about.
