I spent a few minutes this morning twiddling with the circuit design, with an eye towards 6V operation.

First off, I changed D1 to a 3.6V Zener. I happened to use a 1W diode; as it turns out, the 1W parts take substantially more current to approach their Zener voltage than the 1/2W part I used in my 12V circuit (1/2W 5.1V NTE5010A's spec iZ = 20mA, 1W 5.1V NTE135A iZ = 49mA, 1W 3.6V NTE134A iZ = 69mA). Even with the 1/2W part, most of the overall circuit current draw is used for the Zener diode, so boosting it to accommodate a 1W part is even more wasteful. In order to accommodate the 1W diode, I dropped R1's value from 10k to 1k.

Here's a summary of the other changes: 1) Ditch D2; 2) R6 up from 10k to 22k; 3) R4 and R5 to 1Mohm (or lower, for more hysteresis); 4) R8+R9 dropped drastically. In my test circuit I eliminated R8 and R9 entirely, and the LED current draw was only a couple of mA... but that leaves the opamp outputs exposed and vulnerable to short circuits.

As far as the opamp U1 goes, I swapped out the NTE889M and tried the National TL082 (Rat Shack part 276-1715) to see how it worked. The TL082 pulled more current (about 4.5mA more in the 12V configuration) and didn't put out quite as much current for the LEDs.

When it came to setting the voltage thresholds, I just used 5.5V and 5.0V for testing. For a 6V SLA system, I'd probably just use 5.25V for "critical" and then set "low" based on my run-time; for a NiMH system, I guess I'd choose the "critical" voltage based on the number of cells, at 1.0 or 1.1V per cell.

Best of luck,

JAB