My application features a Li-Ion battery (NCR18650B) which I'm charging with the standard CC/CV method. Various constraints force me to use a generic buck converter (with current limiter) for the charging, instead of a more specialized IC. I have temperature monitoring and MCU control over the charging process.

I'm wondering what's the recommended approach to calculate the feedback resistors that set the buck converter's output voltage. This voltage should canonicaly be 4.20V ±0.05V, but I'm using 1% resistors, and another 1% variance comes from the feedback reference; in total, about 3% of uncertainity. For example, I can select resistors that give me 4.20V typically, but span a range 4.08..4.32 with all tolerances considered ("the simplistic approach"). I can instead design conservatively and select a divider that sets the voltage to 4.10V typically, with [3.98..4.22] variance, which is well within spec ("the conservative approach").

In the current PCB revision I've designed with the conservative approach and in the lab my devices consistently charge to about 4.05V OCV (I believe because the Vref is actually not uniformly distributed in the specified range, but skewed towards lower values). The bigger problem is in the actual usage scenarios, where the energy source is not a reliable wall-wart, but an intermittent source whose availability pattern is unpredictable. As it is "free to use", I want to optimize time-to-charge. Charging with voltage (typically) limited to 4.05 means the charging gets into the CV mode too soon, and the battery could rarely ever exceed 70% SoC.


If the cells feature protection circuitry, which is the recommended, industry-standard way to select the CV charging voltage? The simplistic one (target 4.20 volts, don't care about tolerances) or the conservative (consider worst-possible tolerances, so your charging voltage could never exceed 4.20V)?

  • 3
    \$\begingroup\$ There's a charge cycles/capacity tradeoff in the 3.92 to 4.20+ voltage range. Do you want capacity or life? How much does capacity cost? 0.1% resistors cost about 50p each, and would eliminate 2% of your variation. Or you could adjust. Or you could select. Just some questions so you can tighten up your spec of what you want, and think how much you're prepared to pay for it. \$\endgroup\$ – Neil_UK Mar 18 '18 at 17:42
  • 2
    \$\begingroup\$ Off the shelf charger IC's are trimmed during manufacture to make sure the CV voltage is within spec. If you continue down this road, you will need to add a trimming step also to maintain the same accuracy. You can do this by adding a trim pot to the voltage divider. The goal is to have enough adjustability in the potentiometer to overcome the up-to 3% error. \$\endgroup\$ – mkeith Mar 18 '18 at 18:09
  • 2
    \$\begingroup\$ I don't think it is acceptable to rely on the protection circuit. I think the maximum CV for the charger should be less than or equal to the minimum OVP threshold for the protection circuit. \$\endgroup\$ – mkeith Mar 18 '18 at 18:11
  • \$\begingroup\$ I found one datasheet which listed a maximum CV setting of 4.23V, not 4.25. batteryonestop.com/baotongusa/products/datasheets/li-ion/… \$\endgroup\$ – mkeith Mar 18 '18 at 18:17
  • 1
    \$\begingroup\$ It will probably be OK. You would get a physically small trim pot. You would still use the feeback divider, but use the trim pot to fine tune only. For example, if the main divider is 10k and 2k, then the trim pot could maybe be 220 Ohms or something. Another option is to trim using optional resistors. You place two extra resistors in the divider network. Build with 0 Ohms installed. If the CV set point is out of range, replace the 0 Ohm trim resistors with small value resistors as needed to bring it back in spec. Obviously not good for high volume low-margin products. But OK for low volume. \$\endgroup\$ – mkeith Mar 18 '18 at 21:33

There are an infinite number of profile paths to get to a given SoC.

At a low enough current rate, holding Absorb / CV long enough can overcharge even at a much lower voltage.

Overcharge meaning "losing significant life cycles if done consistently".

The good news is, very little usable SoC Ah capacity is lost by incrementally dropping Voltage, or stopping charge at a higher / earlier endAmps setpoint.

So, yes, if your regulation is sloppy, be very conservative. You could even at low C rates, "just stop" at a voltage setpoint, no CV at all, and sacrifice under 3% capacity.

At high C rates, can be less conservative, any resting voltage with the batt isolated over 3.6-7 is "close enough" to full if you can't measure actual SoC through controlled load testing.

| improve this answer | |

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.