Roast My Circuit:  LiPo Battery Protection for Raspberry Pi

Relays are horribly power inefficient - the draw of the coil could easily add several tens to hundreds of mA to your power consumption. Perhaps consider buying a "latching" relay if you can - those only need a "pulse" of power on a coil to change the state, thus not taking any power while otherwise running. Electromechanical relays may be a bit vulnerable to vibration and bumps ...

Not sure the 1000uF would keep the circuit latched that long ... relays are pretty power hungry.

To be honest, you're probably better off with a proper 3s battery pack protector circuit that will disconnect the battery regardless - no intervention from a Pi or anything, as that could crash or hang. Such a circuit would unconditionally protect the battery while only using very small amounts of current (tens of uA). Usually they also incorporate cell balancing as well, but need connections to the intermediate tap points. They often also provide short circuit and over-voltage protection too.

Then you're basically reduced to having the RPi detect the voltage of the battery and shut down below a given voltage. Something like a cheap uC with an ADC input could provide that data to the Pi, or if you're daring, an ADC hooked into an I2C/SPI port in much the same way.

In essence, that would achieve a similar result, but minus the mechanical relay and power consumption associated with it, and the risk of software crashes causing the protection to fail.

- Gough

I like the sound of the latching relay.  My breadboarded circuit drew about 43mA, which is acceptable for the application, but 0mA - now that's a bargain.

The capacitor in my circuit is actually on the base of the transistor that is completing grounding the relay.  So, it is very slow to drain.  The breadboard proved it to be about 20 seconds to drain through the base - plenty of time for the Pi to shutdown cleanly after the pin voltage is dropped.

To address the software failure concern, I'll add a second voltage supervisor.  One supervisor will use a voltage divider as a LOW alarm that triggers a user warning and a shutdown timer for a clean shutdown by the PI.  The other will be monitoring for the LOW-LOW interlock which simply punches the latching relay regardless of the PI state.  Ideally, that one would never be reached.  This will be the best of both worlds and will prevent SD card corruption from hard power downs.

I'll update the schematic once I find one of those latching relays.

Thanks for the assessment.

Sean

This latching relay looks like a winner - plenty of amps, lapping, and works good with my battery:

     G5RLK1EDC12 - Omron - General Purpose Relay, G5RL-K Series, Power (newark.com)

Indeed, for applications where the relay is staying "on" for long periods, latching relays are the norm.

Sorry, I didn't notice the capacitor was on the transistor - you might want to make sure your Pi turns "off" the supply by going high impedance rather than low on the pin that is connected to the transistor base/RC combination, otherwise the Pi could try to discharge the cap via the GPIO which would probably burn it out or make it very unhappy (depending on where your resistors are placed).

- Gough

Concept art for the new circuit.

I'm using a dual N&P channel MOSFET surface mount chip along with two voltage supervisors.

The clean software shutdown is triggered by a voltage supervisor that is sensing off a voltage divider.  I call it the Low Alarm point.

To be safe from glitches, a transistor only branch in the circuit handles the Low Low condition and does a hard shutdown.  Hopefully, it will never be needed.

See ya',

Sean

Interesting sketch ... but perhaps there are still improvements afoot, at least from my point of view. The voltage supervisor function isn't all that clear (in terms of how it is implemented) - most latching relays only expect to have a short "pulse" input, so if it's supposed to be turned off, then perhaps having it hooked to the battery long term will cause current drain from the battery through the voltage supervisor if its output just "remains high" (the opposite of what you want).

Instead, I suggest you might have the voltage supervisors powered from the output of the latching relay itself, so if there is no "initial" pulse to start the circuit, everything is essentially open circuit. Likewise, same with the voltage divider for sense - the way you have it is it will always drain the battery over time (and not an insignificant drain at that). That also takes care of not having current drain through the voltage supervisor and its associated transistor when powering off - because as soon as enough current passes to latch the relay off, it will cut power to the supervisor circuit itself.

Another thing that you're probably missing is the "flyback" diodes for the relay coils - when you de-energize coils, the inductive kick you get results in a high voltage spike that will eventually destroy most semiconductors. A diode placed in reverse across the coil is customary to "snub" out this spike.

- Gough

Good point on the voltage supervisor blocks.  They would at least have some micro amps.  I'll move it's inputs downstream of the relay as you said. 

Another good point on the flyback diode.  I remember doing this with my motor circuits.  I'll add it in.

This whole circuit block is being added to a bigger circuit that also contains a speaker amplifier, display screen port, and some auxiliary headers for I2C devices.  Looks like it will all fit well in my camera project reducing the cobbled together protoboards I have now to 1/3rd the footprint. 

The board is basically a Pi Hat that will make my future battery powered/portable projects much more end user friendly.  I'll share it on OSHPARK.

Thanks for the roasting!

Sean

Still a work in progress, but its coming together...