My project involves a lot of Wixels powered by single LiPO cells, and I need to reliably cut power as completely as possible if the source voltage drops to around 3V. Otherwise I risk damage to the cells. Originally I was just going into Sleep mode, but the best i can do with an un-modified Wixel is still over 80 uA, which will still eat away at the cell if the circuit is left unattended for weeks.
Now I know snipping the regulator will allow for a much lower current, but this is a difficult and error prone task, and then I would need an external regulator (a completely charged LiPO may be 4.2V, higher than the CPU would tollerate). So long story short, I'm trying to develop an external method of shutting down the wixel, and this is my idea. If anyone can see any problems, or better yet see a way to do this with even less parts, please comment!
The idea is this.
1) On power up, the P Channel MOSFET should turn on, because the capacitor will take time to charge, thus holding the gate voltage low for a second or two. The MOSFET I'm thinking of using is a low voltage and low "on resistance" power MOSFET by International Rectifier: IRLML6401 (info here: http://www.irf.com/part/_/A~IRLML6401 )
2) Before the voltage on C1 can rise appreciably, the wixel (MCU) will be running, and immediately output a voltage on an I/O point, connected to the NPN transistor. This should keep the MOSFET's gate voltage low and keep the power to the wixel on. Note, that transistor "should" also prevent leakage through the wixel's I/O outputs when the device is un-powered. (I guess I could have used a FET for that one too)
3) The wixel will continually monitor its regulated 3V3 voltage using standard ADC calls. When the LiPO cell really starts getting low, this will drop below the 3.3V point. I'll first respond by alerting the user of eminant shutdown, and soon will set all my I/O I/O to "inputs" and drop into sleep mode. This should remove the bias on the NPN transistor, allowing C1 to charge. Very soon here won't be enough gate voltage to the MOSFET, and it will turn off.
4) I would think at that point, I should be down to nano amps, the combined leakage of the two transistors in the circuit. Certainly way better than the 80uA I was dealing with before.
5) To reset (hopefully after a battery charge), the user would just have to open the switch long enough for the charge on C1 to naturally dissipate. I suspect several seconds should do it, but a high value bleed resistor might be needed
I confess haven't tested any of this yet, and am hoping to get some "sanity check" ideas before I start gathering parts. In particular, if this can be done with less parts, I'd like to know. I'm already employing a MAX1555 IC, along with a couple more resistors and caps to handle safe charging for the LiPO cell, from an external 5VDC source, and so the extra circuitry needed to use a LiPO cell has already become more elaborate than I'd originally hoped. And if its a dumb idea, feel free to shoot!