Gentlemen - I just joined the list and picked up on your discussion about measuring battery voltage, although I am a little late an probably out-of-context. But I wanted to relay an idea that paid off in a past application in case it helps: We made a little device with a PIC a few years ago that transmits an infra-red PPM message every few seconds and has a lithium battery. Life was usually about a year or more. We had no end of trouble trying to figure out when to report a low battery in the transmitted message. Voltage sensors would call a battery low when it actually had life left, or would not decide it was low until it was too late to report it. Then, I had a brainstorm! The PIC SRAM will retain data down to a volt or lower. Our IR PPM transmissions were momentary power hogs. The firmware would set a bit in the RAM prior to starting an IR transmission, and reset the bit when the transmission was done. If the CPU ever found itself waking up due to a watchdog reset or power-up, and found the bit to be set, then it knew that the IR transmitter current had sucked the battery down enough to crater the CPU. Then, we added a counter such that the device would send five messages in rapid succession once every 1000 messages (about once every few days), which would serve as a "battery test". If the five-message hit on the battery did not cause the CPU to go left-field, then we had enough battery to continue on. If the five-message blast killed us, then we still had enough energy to do single messages during which time we would report a low battery condition. In fact, we would slow down the rate of uplink in this mode just to stretch the battery life out a little longer. This method required no components and always got every possible amount of use out of the battery. If a CPU was, by variation of process or whatever, able to run at a lower voltage, then we would use more battery. No matter what, a low battery indication was a REAL indication of a battery near the end of its life. The Lithium batteries don't show voltage decreases until they are almost dead, so it very hard to check their state without a significant load being applied. We just used the load that was already there. I always thought that a system with a PIC could tie a resistor from a port pin to the VCC rail and then test it's own battery just by pulling the port pin low for a few seconds. If this is done very rarely, then it does not shorten the battery life enough to matter, but it provides a very accurate battery status check. The better the battery, the longer it can sustain the load before the CPU resets or craps-out. Chris Fox -- http://www.piclist.com hint: The PICList is archived three different ways. See http://www.piclist.com/#archives for details.