Can I put Li-ion batteries in series and discharge them at different voltages?

I need to run a pump (drawing 95mA at peak flowrate) at +14V and the rest of my circuit runs on +3.3V.

To save on battery costs, I was thinking of purchasing four 18650s at 3.6V 3500mAh per cell and stacking them in series to produce ~14.4V needed for the pump (it can run between 10-15V and we are going to use PWM to control the voltage i.e. the flowrate).

Could I also run two terminals from one cell (at 3.6V) and regulate that to 3.3V for my lower voltage loads? I understand that this will cause one of the four cells to discharge faster than the other three, but I plan on using a BQ29209 balancing chip to maximize the capacity I can get out of them.

Would it be wiser to just use two buck converters from 14.4V down to 12V and 3.3V respectively?

Ideally, if I had more money, I would put 4x 3.6V in series = 14.4V @ 3500mAh, and then another 2x 3.6V in parallel = 3.6V @ 7000mAh to get more capacity.

• There is a major flaw with your plan nobody else has mentioned. The maximum voltage of a Lithium Ion battery is 4.2V when fully charged. 4x4.2 = 16.8V. So the maximum voltage when fully charged is higher than the 15V max you listed for your pump. I think 3 pieces of battery might work OK, though. And I agree with other answers also, but no need to repeat what they said. Dec 29, 2018 at 4:02

Would it be wiser to just use two buck converters from 14.4V down to 12V and 3.3V respectively?

Yes, that would be a MUCH better idea. Tapping off 3.6 V for making 3.3 V is a BAD idea as you would be using that single cell differently (loading it more) compared to the other cells. Then this cell's charge state will at some point be different from the other cells.

Also if you charge the cells the other cells could be fully charged while that single cell isn't fully charged yet. You would need separate charging circuits to handle this properly.

Also that single cell will wear out sooner than the other cells.

Using a buck converter to make 3.3 V from the 14.4 V is easy and much better design wise as all cells will be treated equally.

As @Bimpelrekkie already mentioned, the buck converter is easy and relatively cheap (I assume you do not need any high currents there) to implement and much better idea from almost any point of view.

Your proposal has at least two significant problems:

1. Single Li-ion cell is not good source for circuit which needs 3.3 V. Typical end-of-discharge voltage for LiIon cell is less than 3.0 V, but you would need to stop discharging somewhere slight over 3.3 V (depends on LDO dropout voltage), so lot of cell capacity will remain unusable. (If your MCU circuit can work with 3.0 V or even less, than this is not so much issue.)

2. Balancing with BQ29209 means, that you are simply putting a resistor in series with higher-voltage cell. It means that you other cells are discharged with the same current as the first one, but energy is lost in joule heat. It is in no way "... to maximize the capacity I can get out of them" Besides, I am not sure you can easily adapt BQ29209 to balance 4 cells in series.

• Thank you @Martin. What I meant with my capacity optimization comment is that, by putting the cells in series, I'm getting a higher voltage but still the same capacity. I'm wondering if it would be better to put the four cells in parallel to quadruple my capacity, and then step up the 3.6V to 10 or 12V from there for my pump. Or perhaps two pairs of series-connected cells in parallel, doubling both the voltage and capacity? Dec 29, 2018 at 0:42

Your proposed design is flawed, consider either step up or down. I give the considerations for each.

Pros

• 3.3V Buck converter for 1mA to 0.1A is much simpler than a Boost regulator in component count.
• 14.4V Boost converter is better regulated than a 4S = 12 to 16V pack and delivers constant energy when needed and may be disabled when not for high efficiency.

Cons

• a 14.4 Boost converter is more complicated ( but easy to follow TI webBench designs )

Other

• Both Buck and Boost in this range will be about 85% efficient but the Boost can be disabled when not needed.
• a 3v~4V 4P pack with a buck-boost converter for 3.3V can extend the capacity of the pack unless you use an 3.0V ultra low dropout LDO .

• whichever pack config voltage consumes the most energy ( watt-h ) is the preferred voltage. 4P= 3 to 4V or 4S=12~16V

• only you can calculate this and use in your design spec for efficiency, energy storage, discharge time, complexity and cost.