(SOLVED) Charging 18650's with a CC/CV buck converter

[Thermoelectric]

Hey all,

For a small event I'm involved with, we use packs of 18650's to run lighting overnight and have to recharge them each day. Last year we used a bunch (say 20) of 3A XL4005 (or XL4015) based CC/CV converters, one per pack, and I ran into an odd issue. Setting the supplies for 4.2V CV and say 2A CC, they'd charge in constant current to about 4V then start tapering off the charge current before reaching the 4.2V setpoint. As this slowed charging substantially, I tweaked the voltage up on the supplies a little. However, as you'd expect, this resulted in some packs getting overcharged and destroyed.

I assumed this was just specific to that particular type of CC/CV buck converter, however got a few different styles and I've noticed the exact same characteristics. I've filmed a quick overview of this - https://youtu.be/WnNlf9rCMps

Any ideas on why the buck converters behave in this way? I assumed they should run in CC mode right up until they hit the CV setpoint, however it appears not to be the case. Alternatively, does anyone have any suggestions on something that may work better? Current packs are 1S14P so needs to be low voltage, high current and preferrably cheap. I've got a LiPo charger but even that won't pump out much power for 1S packs.

Cheers.

[Siwastaja]

Looking at XL4005 datasheet, it has no CC mode, it's just a standard voltage regulator. The current limit is high, non-adjustable, and to protect the MOSFET of the regulator IC. So whatever module you are using implements the adjustable current limit somehow - we don't know how. That current limit circuit clearly doesn't work properly.

This being said, the current starting tapering down already at 4.0V isn't necessarily a bad thing, this increases the cell lifetime especially if the initial charging rate is quite high. For actual numbers, I would suggest tapering a 1C charging current down to 0.5C between 4.0 and 4.2V. Of course, if it tapers too low too early, the increased charging time becomes an inconvinience.

BTW, be very careful charging cells with some random power supply modules. Try to verify there is no failure mode of Vin shorting to the Vout; add a classical fuse rated just above the nominal charging current which blows if this happens. Note, some power supplies can blow up if there is a voltage connected to the output while the input is not connected, or if the input is shorted. Connect a voltage to the output and see if it powers the input back; if this is the case, you might want to add a series diode on the input side.

[Thermoelectric]

For that particular board I assume the CC mode was set by an external means - see the board on eBay (that one was actually a 4015, but I've got another 4005 board here with me, same setup) - https://www.ebay.com.au/itm/5A-DC-to-DC-CC-CV-Lithium-Battery-Step-down-Charging-Board-Led-Power-Convert-JCA/193263079849

The packs being 1S14P allow for the higher charging currents up to 4.2V as each cell would only see 1/14 of the current, but I totally get where you're coming from.

I think I just answered my own question though, the taper I'm seeing is due to the boards only being 2 wire, not 4 wire measurement, just checked with the current setup and I'm seeing that discrepancy in the voltage drop of the wires going to the battery. Can you think of any ways to compensate for this short of trying to break out the sense leads on these boards?

[tunk]

Onstate LED Lighting has a few youtube vidoes about these,
and in some of them I think he suggests some modifications.
A web search will show some reverse engineered schematics
if you want to look into it yourself. Also note that these does not
stop charging when reaching 10% of the starting current.

[Siwastaja]

I think I just answered my own question though, the taper I'm seeing is due to the boards only being 2 wire, not 4 wire measurement, just checked with the current setup and I'm seeing that discrepancy in the voltage drop of the wires going to the battery. Can you think of any ways to compensate for this short of trying to break out the sense leads on these boards?

This is the likely cause, yes.

The obvious solution: reduce the resistance in wiring. If you see the changeover to CV at 4.00V instead of 4.20V, this means you are dropping 0.20V in the cables, which is approx. 6% power loss, a significant part of the total system losses (assuming the cells are 99% efficient and the DC/DC module maybe 85%).

Use thicker, and/or shorter wires to the pack, and if it isn't enough, consider replacing the crappy quality screw connector with something better (like directly soldering the wires to the module) to reduce the contact resistance.

So move the very low voltage charger DC/DC modules as close to the pack as possible; the higher voltage, lower current input side tolerates thinner wire.

You can't completely fix the issue with this, but I doubt you need to. Once you get the cable loss to a level which makes sense from the efficiency improvement viewpoint, your charging time problem has almost vanished as well.

[Thermoelectric]

You know what, I might just ditch those packs entirely and make some 12V packs instead, should solve more than a few issues we were having. Thanks for the insight all, much appreciated!

[Siwastaja]

1s packs have their advantages; you don't need to think about balancing or cell-level monitoring, HV and LV limits are absolutely trivial to implement. The cost is the high current, low-voltage wiring, though. What's consuming the power? You say lighting, but what kind? You just run LEDs with series resistors, or maybe you have a proper constant-current LED driver?

[Thermoelectric]

As bad as it sounds, we were running LED strip from old TV's from little CV boost converters. These strips didn't have any current limiting resistors so I'm surprised it worked at all.

Thinking we'll just scrap the current setup and go for 12V/3S packs with a BMS, and use it to drive 12V LED strip instead. Makes charging easier, gives a much more controllable lighting level, and gives high voltage and low voltage protection to the pack. If we're gonna do it, might as well do it right.

[Siwastaja]

Thinking we'll just scrap the current setup and go for 12V/3S packs with a BMS, and use it to drive 12V LED strip instead. Makes charging easier, gives a much more controllable lighting level, and gives high voltage and low voltage protection to the pack. If we're gonna do it, might as well do it right.

That's not necessarily the "right" way, however. 3s pack ranges from about 9.5V to 12.6V, you can't drive a 12V LED strip with that (unless you are OK with huge brightness variations and basically dismissing the last 10-20% of the battery capacity), it needs regulated voltage, so you either need a SEPIC type converter, or go to at least 4s pack and a buck.

Less complex and more efficient way is to use a proper constant current driver for the LEDs, which is either buck or boost type. You get rid off the series resistor loss in the LED strips.

Unless your LED power requirements are so large that a 3.6V wiring losses become a problem, the boost type has many advantages, like the simplification of the BMS.

Many are available, this is a very typical problem, typical use case is powering backlight LEDs from a single li-ion cell:
https://www.digikey.com/products/en/integrated-circuits-ics/pmic-led-drivers/745?k=&pkeyword=&sv=0&pv1098=408383&sf=0&FV=-8%7C745%2C183%7C337512&quantity=&ColumnSort=0&page=1&stock=1&pageSize=25

In practice, this would be 3-4, maybe 5 leds in series per each string (too much voltage boost would drop the efficiency), with a boost-type constant-current LED driver IC driving each string. Such converters typically are approx. 80% efficient, so definitely better than the combination of first regulating the pack down to 12V, then dropping some more in the series resistors to "regulate" LED current.

I recently picked SC4541 because of the integration - the solution area isn't that much more than a simple series resistor! It's a bit more expensive, though, but increased efficiency means savings in battery cell cost.

Pay attention to the LED efficiency, there is huge variation. If you pick a product delivering say 130 lm/W instead of 65 lm/W, you can halve the power and double the runtime. Some random Ebay LED strip may be just 20-30 lm/W.

If you have LEDs everywhere, large total power, and a lot of wiring, then a 4s to 6s pack with local buck-type regulators is more efficient. It should still optimally be a constant-current LED driver so you can skip the series resistors. (If you want to do it "right", that is!)

[Thermoelectric]

True that, we've used a similar setup in for other lighting and the packs usually cut out at maybe 10.8V, but you're right, we'd be subject to brightness variations. However for what they're doing (art piece ambient lighting and "street" lights in the middle of a paddock), I don't think it's a massive issue.

If we went and used a constant current supply for the LED's, we'd need some arrangement of LED's that don't already have a series string resistor. Can you think of anything off the top of your head where these would be premade? Short of TV backlights (which might run at 25-30V), I can't think of anything that would actually fit this purpose.

I've previously looked into using constant current boost converters for said TV backlights but they're nowhere near as cheap or readily available as CC buck converters, so I'd still be looking at a 2S/3S/4S etc pack and a buck CC depending on the particular lights available (don't particularly feel like spinning my own).

[Siwastaja]

No, I don't think there is a easily available pre-made solution, at least not at a significant price.

I like to design everything myself, hence the suggestions.

For something ready-made, the 12V LED strips are likely easiest; just pick a proper supplier who sells products with datasheets you can trust, pick those with good efficiency (well beyond 100 lm/W).

10.8V cutoff for 3s pack, you can't be serious? That's approximately 50% SoC under no load; under load, you are using even less than a half of the pack capacity.

[Thermoelectric]

10.8V cutoff for 3s pack, you can't be serious? That's approximately 50% SoC under no load; under load, you are using even less than a half of the pack capacity.

You make a very good point. For some reason was thinking that 4.2-3.7V was the typical range for 18650's. Gonna have to look into those packs, maybe there's something whacky going on with their BMS's. Sounds like a SEPIC converter will likely be the best solution then if we go with 12V LED strip. Cheers!

[Siwastaja]

If you are really forced to use a 3s pack, then sepic it is.

Buck from a 4s-6s pack would be simpler and offer more savings in the wiring.

[Thermoelectric]

Not forced, however looking to use secondhand cells so I'd just have to be a bit more cautious with arranging the packs by cell capacity, 4s probably wouldn't be that much harder than 3s though. Definitely some food for thought!