For various reasons I'm happy to describe in the comments section, I would like to charge large (>30S @ 1-30Ahr) packs directly from USB. Speed is not critical and multi-day (or week) charging is not an issue. However, what is an issue is size, safety, and complexity, which are hard to come by at these cell counts (unless you have an extensive engineering department to test, test, test).

I reason that since a USB supply is 5V (@ 500mA-5A, depending on PD or not), it's conceptually easy to inject power into a 4.2V cell. If we could split the pack up into individual cells we could connect each cell to the 5V supply and let it soak up some charge. No need to develop any kind of switching mode power supply or balancer.

However, in a real-world case, these cells are inseparable and so the charger has to deal with the fact that there might be 100V or greater across the pack. Furthermore, the devil is in the details, since pragmatically it would be very dangerous (i.e. prone to user failure) to completely charge one cell before moving on to another.

I feel like the below approach should work for an arbitrary number of cells. The idea is that a microcontroller switches on both ground and positive terminals for a single cell, while switching off all other cells. A zener diode protects against the microcontroller running amok and trying to overcharge the cells.

enter image description here

  • The system switches at a high rate ("high" as determined by the LC filter)
  • An LC smooths the current so that the battery isn't seeing a sharp PWM signal
  • A zener diode provides passive protection against the microcontroller running amok and trying to overcharge the cells
  • Requires a USB power supply with good current regulation (to keep itself from blowing up)
  • Not drawn are all the things required to make this work in a real-world setting, e.g. appropriate FETs vs transistors, the sensing circuit to measure input current, voltage, temperature; PWM control of the input voltage; current limiting through the zener diode, etc...

Seems simple, does it work as easily as this? Are there some obvious pitfalls in this design?

My working assumption is that this doesn't exist in the market because most people don't want to charge a >1kWhr, 100V battery off of USB. If I'm wrong, please use the comment section to disabuse me of this notion.

  • \$\begingroup\$ I haven't been brave enough to explore charging lithium batteries without a dedicated controller, but one thing I notice is missing from your implementation is current limiting. Slapping 5V straight on a 4.1V zener and battery is gonna throw as much current as possible from the USB port through those - how much through each would be dependent on the diode resistance and battery internal resistance. \$\endgroup\$
    – Orotavia
    Apr 29 '20 at 1:33
  • \$\begingroup\$ You're not wrong, all that is handled by the microcontroller. Drawing out the entire schematic would distract from the crucial question here, which is whether the proposal works as intended. The zener's role is to protect passively in case the uC doesn't do its job anymore. \$\endgroup\$ Apr 29 '20 at 4:48
  • \$\begingroup\$ Just a few comments. 1) limit the base current of the upper BJT as most microcontrollers cannot sink a lot of current. 2) Shouldn't the LC be flipped in order to form a LC filter? 3) Is the battery flipped? 4) What is limiting the current of zener? \$\endgroup\$
    – vtolentino
    Apr 29 '20 at 19:20
  • \$\begingroup\$ Agreed. All those are intentionally left out. There is a lot of upstream glue which is required to go from a theoretical system to a real one. The transistors will likely be replaced by FETs, etc... But does it isolate the cells the way I think it will? (BTW, the inductor is bracketed by two charge banks, to the left by a cap and to the right by the battery.) \$\endgroup\$ Apr 30 '20 at 0:11

The problem is that transistors are not exactly switches, and they especially do not provide isolation between the sides. If you assume the top cell is "on" and its lower terminal is at ground, the lower terminal of the bottom cell is at -5V and the transistor is reverse biased. Their voltage tolerance for this situation is usually quite bad, and it will also be difficult to turn it on. Have you tried building this in Falstad and simulating it? What happens?

(The whole thing seems considerably more complicated with higher part count than just building a boost converter and balancer?)

  • \$\begingroup\$ I'm not sure it's much more complicated than a balancer, since a balancer has to have at least half the FETs shown. This system also has fewer ways to go wrong, because whereas the failure state for bad programming of the balancer's microcontroller is overcharged or undercharged cells, this depends on passives and other transistor/FET properties. I feel that's easier to idiot-proof. \$\endgroup\$ Apr 30 '20 at 16:17
  • \$\begingroup\$ That's a good comment about the reverse biasing. The scale of this gets worse with each cell. By the time I have 30S, the lowest FET/transistor is reverse biased to ~120V. If the FET/transistor cannot handle it, could I not resolve this problem by using properly rated diodes? I have not simulated it, I'd hoped to avoid a simulator until I had some reasonable confidence the idea could exist. \$\endgroup\$ Apr 30 '20 at 16:37
  • \$\begingroup\$ This also obviously depends on the microcontroller working! I'm sure turning on the wrong transistors would blow up something. It shouldn't take long to set up your schematic in Falstad. \$\endgroup\$
    – pjc50
    Apr 30 '20 at 18:56

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.