3S lithium cutoff circuit

[Bzzz]

Hi,

looking for advice on choosing a part and/or using clever circuitry.

I'm working on an ESP8266 LED thingy that is running on a 3S lithium battery pack, most likely one of the cheap car jump starter units. Uses the voltage from the batteries directly (most likely no protection given they peak well above 100 amps), charges using the integrated USB circuitry. 12.6V to 9ish volts is dope for the LED strings, everything else sits behind a 5V DC-DC (could be changed for a 3V3 one).

Now I'd like to accomplish two things
a) Get a rough state of charge reading via the A0 of the ESP
b) Integrate protection cutoff for when voltage drops below ~9V total / ~3V per cell

For the voltage reading I guess I would need a ~13:1 resistor divider in any case, as the ADC only offers 0...1V input range. So when using like 100k in total that would mean 126µA steady power consumption (worst case). I'd test higher resistor values and buffer the low segment before deciding for final values of course. I only need like 20% steps, so accuracy is really no requirement here.

As far as protection goes, there are very few units available that offer such a high input voltage. Most voltage detectors like the NCP300/301 series allow for 3V to 5V. My idea would be to cut power to the DC-DC via a MOSFET, which in turn would black out the ESP and related circuitry, meaning all MOSFETs driving the LEDs would also quickly die due to their gate discharge resistor. That way only the protection circuit would still sip countable chunks of electrons, but all major loads are disconnected. As I said, few components allow for a 9-10V detection threshold, such as the Ricoh R3119N090A-TR-FE or the Ablic S-1011E90-M6T1U4. Both are difficult to source for me. Others easily cost 5 times as much as a NCP300 and aren't much easier to buy either. As I only have plus and minus terminals from the battery available, I cannot use BMS chips that would monitor each individual lithium cell (without faking that with another resistor divider), so that's almost out of consideration. I also would like to have this protection on the ESP board, as this is not really a commercial thing yet, but I expect to build and maintain a few of these over the years. Not being locked in to buying a specific 12V power bank with integrated protection would be preferred, so that I can just hook up to any run of the mill 3S pack a decade down the line.

Do you think dividing the battery voltage down to let's say 4V and then use a more common voltage detector in that range would be a good idea? Those things don't consume much power and turning on a MOSFET (ideally only once) doesn't require much gate charge as well. Bonus question: Would integrating that into the ESP A0 voltage divider be a good idea, in order to minimize static power draw?

Thanks!

[Sceadwian]

Why can't you use a true BMS? It can't be that hard to tap the lines in between the cells. In order to implement proper protection you really need to as the overall voltage won't tell you much about the actual pack state when you really need it which is during cell imbalance.

[Bzzz]

Adding a true BMS chip would require me to open up the perfectly fine battery pack, find appropriate points to solder to (a unit in 5 years time might be different), and run additional wires to the ESP board. That'll nullify the benefits of buying the fully working 3S battery pack with all the bells, whistles and certifications in the first place. There's already a BMS in there which will watch over individual cell voltages during charging (not sure if all of them will use that to balance them). Tapping into these units would also require me to document what to do in there and how to replace a faulty one, which I'd very much like to avoid. Just plug in a new one and it's serviced.

[jbb]

If your pack has protection already and you're deploying a micro controller anyway, why not just do software under voltage cutoff?

You can design a circuit such that when the ESP declares empty battery, all the power cuts off and a physical 'on' button is required to turn it back on.

If you want a low power sleep mode, you could add a MOSFET switch to the voltage measurement divider and save some current.

For battery gauging, the basic voltage method isn't great.  You could put yourself in a much better position by including current measurements, which would allow you to compensate for pack resistance.  (Especially as you're drawing lots of current!)

[Bzzz]

Those jump starters have protection on their 5V output only. Which is not a surprise, as any milliohm in the way of several hundreds of amps when jump starting your car is undesirable. So again, those packs are just fine when charging over that USB circuitry, but the ESP board is hooked up at the "12V" terminals directly and therefore has no undervoltage protection. I could, in theory, power the ESP via the USB output, but I still would have to run the 12V connections over there AND most power banks cut off below a certain current drawn, which will kill this thing once it enters standby mode. Furthermore, different power banks have different thresholds, so figuring out a certain current to waste on one unit might not be sufficient for another.

Voltage measurements via voltage divider on A0 are perfectly fine. Any losses due to wiring resistance will only work to my advantage, as the battery voltage can never be lower than the measured voltage two connectors and one meter of wiring farther away unless the resistors of the divider have terrible and opposite temp coefficients. I'm drawing less than 1A absolutely worst case, 0.5A typical max (software limited), 0.3A typical. These packs offer something like 7.5Ah/30Wh which I have confirmed.

I'd like to keep the ESP out of anything power control related, as I not only fear crashing/hanging software as per Murphy, but there's also only one GPIO left. D0/GPIO16, to be precise, the worst useless piece of junk on the entire MCU, which not only isn't "GP", but also is able to mess up the boot process under certain circumstances.