The MCU also has BOD (brownout detection) enabled, but that doesn’t activate until 1.8V. So, if it gets that low, it’ll shut the MCU off even more than it normally can. Which probably won’t have a significant effect on the parasitic drain.
But LVP (which shuts off at 2.8V) only works when the light is actually on. And it sounds like whatever is draining happens when the light is off. So get out a DMM and start measuring things. It’s probably a hardware issue.
It would make the Q8 suitable for the 3V Luxeon MZ leds (on the 4040 Q8 board): great CRI plus great power with an easy mod, as long as the reflectors are heavy OP to prevent the donut hole!
Not just MZ, but there would be less beam color separation with most other LEDs too - OP reflectors always give smoother, more uniform beam which is important for high CRI builds.
I am very interested in a 3x18650 light and mentioned it in an interest thread a long time ago. The smaller diameter and lighter weight will make it easier and more likely to be picked up and put in my pocket. The 3 cells will allow for very good runtimes and being larger than a single cell light, it should handle heat a bit better. People with smaller hands might prefer it as well.
There have been a few 3x18650 lights released and the Firefly seems to be a success, but it costs quite a bit more than the Q8 does. I don’t care about usb charging at all and would prefer to not have it because of waterproofness if the port is covered by a small rubber flap or plug. If its hidden by the threads, I might like it more. It would have to be a solid design and one that would not add too much weight or bulk or cost for it to interest me.
It’s possible to get a dpst swtich, so you have one pole that “bootstraps” the MCU by turning on a transistor, and then you have the other that is used purely for click detection. It may even be possible to do this on a spst switch with some electronics trickery. When the MCU detects i’s below X volts, it sets the pin holding the transistor to “off” and the whole thing goes totally dark. No power until the button is clicked again.
Hardware wise, it would be best to use a mosfet to minimize quiescent current, and make sure to use very high value resistors for the “on” signal to the mosfet. Given the relatively high capacitance of the mosfet gate, you might even be able to get away with putting a capacitor on the line to hold it high, and just give it a periodic pulse to maintain voltage when in sleep, if you desired to keep the MCU running.
Have you measured the power used by the Q8 driver during standby, or after LVP activates?
In stock form, with the button LED off, standby drain is about 0.02 or 0.03 mA. Or after reflashing it with different fuse values, it’s even lower. If I recall correctly, it was something like 0.00002 mA. And with four cells, divide all these values by four. This makes the standby power virtually indistinguishable from zero, because it’s lower than the cell’s self-discharge rate. With the right fuse values and with the button LED off, I’ve found that I can actually take the battery out for a few minutes without making the MCU shut down. And it can still wake up on button press.
Atmel calls this “picopower”. It makes standby mode use such a small amount of power that it’s not necessary to design in clever power-disconnect circuits.