LiFePO4 balancing.

[paulca]

It's been a while since I looked into lithium cell balancing.  I already see a few "new" techs thrown at it.  Like "active flying capacitor balancers", which looks like a set of capacitors, a PWM circuit and some mosfets which connect the capacitors to each cell in turn.  Such that higher cells charge the caps and lower cells discharge them.  Efficiency seems to be the issue, certainly when testing extremes which they probably would never be used over.

It made me start thinking of active balancing again as I had similar or better ideas for top balancing.  Passive resistors are just pants.

I figured I could do better than a PWM circuit.

An other example I seen "out there" which is more my style was using a completely separate MCU (STM32G030) per cell, running from the cell and connecting back to the mother MCU via opti-couplers, probably just UART.  I like it, but it does seem a bit overkill.  Also it was designed to clip into one of those 18650 grid arrays, not LiFe bulk cells.  And... it doesn't actually do anything special in terms of balancing... just a passive power resistor to cap the top cells.  Making it a bit snake oil for Lithium shedphiles.

So.  My thoughts are to take the two ideas and merge them.

A single MCU connected to (initially) cell 1.  It sees 2.5-4.2V.  Via a 50Ohm resistor it charges a supercap.  The MCU spin monitors that cap and waits for it to stabilise.  Notes the voltage.  Disconnects entirely from cell 1, pauses a few miliseconds and reconnects to cell 2.  Again via the 50Ohm resistor the supercap equalises to that cell voltage and it notes the voltage and moves on cell, 3, 4 etc.

After a pass of all cells it can now make intelligent decisions about which cells need the most attention and focus there.  It can also determine when the cells are "close enough" and put itself into a shutdown state, disconnected from all cells.  Waking up periodically to scan the cells, balance if absolutely necessary and return to sleep.  Data can be reported back to base via Wifi from an ESP-AT module.

I figure as long as the capacitance and surges are managed during disconnects/reconnects and the device is always 100% floating, even if powered from a wall wort it should be doable.

[paulca]

One challenge which might need addressing separately are the "fail safes".  Low/High voltage cut outs for protection. Overcurrent protection (asides a basic fuse!).  Soft fusing, cell fusing etc.  I doubt a "floating" device can help much here except to monitor and inform any "intelligent" control system something needs done.  For me, that is not enough, so a completely separate set of interlocks in the form of cell or pack cut outs ...  I can probably source those pre-made, it's just hard to find ones that DONT also try and balance or charge monitor these days.

[tszaboo]

LiFePO4 has very stable voltage across it SOC, so small measurement errors (and balancing then) leads to imbalance in the pack. So they are usually balanced when the SOC is low, maybe below 10% or when it is almost fully charged. If you do it when it's charged, you can get away with cheaper parts, with some wasted energy, that's why it's usually done that way.
Otherwise yeah, why not put a motor on each cell, and a gearbox selects which cell to charge and discharge. I'm sure we can come up with many bad ways of balancing cells.

[paulca]

Well.  For now I am top balancing them manually to 3.65V with a PSU and 5A.  They were about 100mV across the pack out of balance at 14.40V pack charge, which is a bit too much.  I'll go round them tonight bring each to 3.65V individually until they take less than an amp, then repeat.  Then I'll leave them with only the solar charge controller load on them till morning and remember to check them before the sun comes out.

I have ordered a simple little passive balancer/monitor.  The plan being, if I balance them first, then it shouldn't have much to do.  It only has like 100R resistors or something, so shouldn't waste too much until I decide if I need a better solution and if then what that is.

I believe I am correct they "can" technically go to 4.1/4.2V, but there is little point past about 3.7V.

[lastguy]

Where you get the power source for each cell? it might be easier for 12-16V battery array, how about high volt?
if you use isolated power source for each cell, system would have high cost I think?

[Red Squirrel]

I have not gotten to a point where I play with these, considering the cells are very hard to even get here but my philosophy would be that each cell bank (multiple cells in parallel) would essentially have its own isolated charger circuit and be a single module.  Then you have multiple of these modules in series to get the total voltage you need.  They would have a separate port for the charger which would be powered by a common bus which would be connected to the charger input for the entire device. Ideally you'd want it to have a rather wide input range so you can charge from various sources such as mains or the output of a solar charge controller.    This would also be rather modular as you can just keep adding more of these modules in series to get a higher output voltage.    Would that idea actually work?   The main issue I see is in setups like solar where the batteries are being discharged at same time as being charged. Charging lithium batteries requires to monitor how much current is going in so if there's also current going out, it makes it a bit tricky to do right.   I guess in an application where there is also a load, you can have the charging circuitry actually power the load directly, then also charge the batteries out of circuit.  Ex: with relays and a huge capacitor bank so there is zero switch over time.

[paulca]

That works.  I am upgrading from ~12V pack to a ~24V 8 series pack.

Current only flows in net.  So if your charge controller is producing 5A into the battery and the load is using 2A then only 3A flows down the conductors to the battery.  When the sun goes down and the panel is producing nothing, 2A is flow out of the battery.

LiFePO4 100Ah cells are "happy" up to 0.5C charge rate, although 0.2C is meant to be gentler.  That's still 20 amps - 50 amps!

[Siwastaja]

Forget all redistributive (some call it "active") balancers, it's academic masturbation and those few who have tried to make it in production have failed by making complex, expensive and unreliable implementations.

Especially if a redistributive balance suffers from poor efficiency (any that only can push charge to neighbor cell, so has to go through multiple conversion steps to reach the destination; or any that uses capacitive charge pumps), then it's a total oxymoron. Because the only reason to use redistribution is to save energy.

And that gets us to the point. It is analogous to designing a switch-mode converter (instead of resistor) to drive a 1.6V 5mA indicator LED from 3.3V "to save energy". How much energy can you save and is it ever worth the increased cost and complexity?

Turns out, the only source of imbalance in li-ion cells is the difference between self-discharge currents. What a dissipative balancer does, it effectively turns every cell into the pack to have the same self-discharge current than the one with highest self-discharge. A redistributive balancer is able to return some % of that energy to other cells.

It is good to understand the existence and principles of redistributive balancing, because a future battery technology could emerge, one which has appealing energy density, price per kWh, or some other parameter, but which suffers from high/unstable self-discharge or poor, unstable coulombic efficiency, necessitating large amount of balancing "all the time". But with currently available li-ion, this is not a concern.

Another case for redistributive balancing is, if cells cannot be matched in capacity, so you transfer charge during runtime to dynamically run all cells from 100% to 0%, to avoid having excess charge left in larger cells at end-of-discharge. But this requires large redistribution power capability, so is really expensive.

Example case: we built a 20kWh demonstration system with second-life LFP in an EV conversion, some of the cells had potential damage due to overdischarge (as destroyed by a fancy, expensive, active balancing BMS). Using dissipative balancing with mere 40mA balancing current never caused issues in that system, despite some 240Ah or so cell capacity.

But that one did use a clever algorithm allowing the balancer to stay on for longer than just minutes during the CV phase. Competitors had balancing currents in range of 200 - 500 mA, which already creates a possible thermal problem in case something goes wrong with the control.

[Siwastaja]

LiFePO4 has very stable voltage across it SOC, so small measurement errors (and balancing then) leads to imbalance in the pack.

Balancing at top maximizes available energy anyway, but as you mention, the voltage curve of LFP is so flat in the middle that if you try to do any balancing there, you are just creating errors.

But, at the end of charge, the voltage of LFP rises quite steeply. At this point, you can easily detect if some cells are fully charged or not. What I did with the "clever" balancing algorithm was simply this: when the BMS detects "pack full" condition - which is any single cell hitting 3.65V - the voltages of all cells are recorded, to calculate required amount of charge removal. Then charge is removed from the cells so that most is removed from highest voltage cells. For example,
V1 = 3.50V -> lowest, do not remove charge
V2 = 3.55V -> remove 1 units of charge
V3 = 3.60V -> remove 2 units of charge
V4 = 3.65V -> remove 3 units of charge

Removal of charge is simply keeping a resistor on for X minutes.

This worked very well. Voltages were sensed during low-current charging. Actual shunting could go on regardless of use (i.e., the vehicle can be driven and the balancing goes on, based on values acquired at the end of charge.)

[paulca]

Yea.  On the question above "How do you get power for each cell?"

I used a floating power supply set to 3.65V and manually moved it between cells.  Individual cell charging/monitoring is only tricky when you are also referenced to the pack ground and VBat.  For that you need some interesting circuitry and the real challenge is measuring the top VBat voltage as it's your max + rail and thus measuring it has challenges.  I considered writing/making a cell monitor to publish those stats to my data system.  However...

I am going to buy a BMS.  Not least for the protection.  Reviews are pointing me to a JDB BMS on AliExpress.  Probably just a 40A one, although given it's source, maybe the 100A one as I have a 40A system.  It has RS485 comms and enough settings to basicly turn it all the way down to minimal/essential.

I find the balancing is most important near the top and near the bottom.  Without cell monitoring a 12V pack pulled down to, say 10V (0%) might on inspection find 3 cells above their minimum and 1 cell well below it.  The opposite happened during charnge.

For LiFEPO4 this isn't such a bad deal, certainly on the charge side as the minimum and maximum voltages are not contained with in the packs normal charge/discharge region but well outside it.  Unlike a Li-ion which is 100% charged at 4.2V a LiFePo4 cell is pretty much fully charged at 3.65V per cell.  However, you can, continue to charge them all the way up to 4.2V per cell.  It's just there is no point.  The sell discharge at 4.2V will drop it back to 3.8V pretty quickly.  Similar thing happens at the other end.  Below about 3.3V there is very little capacitor there to get.  It means your charging regime can cut off long before MAX and your BMS cut it off long before MIN.  Li-Ions when trying to maximise your pack life, usually results in a full absorption charge at 4.2V and a cut out around 3.0V  so you can get the full capacity out of them.  Or if you want them to last, you do the 20%-80% thing.  It's just that the LiFEPO4 cells the 20% 80% based on voltage is about right, though gives you 99% of your capacity.  So easier to manage.

I need to readjust my solar charge controller.  I was sure I set it correctly with an absolute maximum VBatt of 14.40V, but I have seen peak voltages of 15.1V.  Only for a few seconds when the sun pops out from behind some cloud the controller doesn't appear to cap the voltage that fast and produces a slight overshoot.  Again the pack is technically good all the way up to 4*4.2V or 16.8V, though you would never take it there delibrately.

When the pack is nearly fully charged and the panel is doing absorption (which is also set to long at 4 hours), I measured the cells and found an imbalance of 200mV.  It's hear the risk lies.  Without a cell monitor one cell could end up quite high.  At the same time however the self discharge rate rises for all the cells, so, in a normal solar day, assuming the pack full charges, the cells will all look imbalanced at the top.  However within a few hours they will have settled down and put under a little light load they appear with in 10mV of each other.

[paulca]

On this particular BMS you get to set quite a lot.  The preferred balancing method is to do it during or close to absorption.  So when the cell voltages are below say 3.50V no balancing occurs.  When cells are above that they become balance candidates.  I am not sure, but similar at the bottom end might be nice too to prolong the "single low cell" cuts pack.

[Siwastaja]

I find the balancing is most important near the top and near the bottom.

You can't balance at both top and bottom. You choose one (usually top), and then the pack is out of balance at the other end, because capacities differ. That's life.

As I explained above, you could design a high-power, redistributive, high-efficiency balancer which rebalances the pack during each and every cycle, so that it drifts from top balance to bottom balance. But this would be colossally expensive, for a small gain in added capacity.

Remember that balancing is only for maximizing energy. Cell-level LVC and cell-level HVC is needed anyway. It's hard to guarantee perfect balance (for safety). Hence, safety is always provided by the HVC and LVC; when any cell hits any limit, charging/discharging the pack is stopped. In a perfectly top-balanced pack, HVC happens in every cell nearly at the same time (so it's basically random which cell triggers first).

[paulca]

Dumb idea, but...

A single "auxilary" cell.  It is not part of the main pack.

Under discharge when a cell gets within 10mV of LVC the aux cell is used to support it, using whatever concuction of relays/mosfets to float that aux cell in paralell with the low cell and rotate across them supporting the low cells while we drain the last bit out of the high cells.

Under charge, once the main pack has cells exceeding "Nominal charge" voltage when a passive resistive balancer would cap it, instead the aux cell is placed in parallel with the high cell to pull it down.  Completing the charge when all cells including the aux cell are at nominal charge voltage.

[Siwastaja]

Good idea, but WITHOUT disconnection of the "aux cell". I did exactly that with the LFP EV pack mentioned above; out of the 26 (IIRC) cells, I think two had significantly less capacity than others, bringing the usable capacity of the whole pack down. I simply added some tiny 2Ah cylindrical cells (of the same chemistry, of course - voltage curve needs to be similar enough) in parallel to get their capacity up. Worked fine.

All you need to do is to run the pack to empty once and see which cell(s) caused the LVC, and increase their capacity by adding the tiny aux cell (permanently in parallel).

Dynamic disconnection/connection is a lot of work and expensive components, and if something goes wrong and your logic connects the cells together at wrong time, it's catastrophic, as large currents will flow unless the voltages are equal to maybe +/- tens of mV.

But make sure you have a problem with capacity differences first. If you buy decent cells, they are very close to each other in capacity, so there is not much to gain.

This is a more general comment: make sure you understand the problems you are solving (or if they exist at all) before thinking about how to solve them.

[paulca]

Dynamic disconnection/connection is a lot of work and expensive components, and if something goes wrong and your logic connects the cells together at wrong time, it's catastrophic, as large currents will flow unless the voltages are equal to maybe +/- tens of mV.

Yes.  My thought experiments around cross linking between cells etc.  Tended to include, "I'm a software engineer and no, I don't trust even my own code during development with real cells.", not without safety interlocks.

I was considering using a chain of MCU->74H logical gate array-> Mosfet drivers -> Mosfets.  The logic array hard coded to prevent any invalid mosfet configuration with basic NAND gates.

It is far, far too much effort for an "hobby" class nano gen off grid system.  Might be a different story if you are developing the next bleeding edge multi-cell EV pack and have 1000 cells to balance and a product with a £10-15k market value.

[shapirus]

I don't know. It sounds to me like you are trying to reinvent the wheel for an unclear purpose.

Here is my personal example. Three LiFePO4 batteries. All three use Heltec active capacitor-based balancers which balance the cell voltages continuously regardless of the SoC level, unless disabled via the RUN jumper.


1. 4s, 600 Ah.





2. 4s, 18 Ah.




3. 8s, 18 Ah.

No direct access for multimeter, so in this case cell voltages are taken from the BMS reporting over a serial port.

Code: [Select]

Command: cellVoltages - Cell Voltages Information
--------------------------------------------------------------------------------
Parameter       Value           Unit
cell_01_voltage 3.409           V   
cell_02_voltage 3.411           V   
cell_03_voltage 3.411           V   
cell_04_voltage 3.411           V   
cell_05_voltage 3.411           V   
cell_06_voltage 3.411           V   
cell_07_voltage 3.411           V   
cell_08_voltage 3.409           V   


The only situation where this wouldn't be sufficient that I can think of is very high-capacity batteries made of poorly matched cells where the balancing current achievable by these balancers won't be high enough to catch up. But they can be parallelled, if necessary.

[Siwastaja]

Heltec active capacitor-based balancers

Fun to see how the world goes on, and it seems Chinese are now producing cheap redistributive balancers and people are apparently buying them. Didn't know that. And I'm sure they work better than the overly expensive and complicated products I have seen years ago.

... but then again:

The quiescent current is about 12 mA

So an energy saving device the only purpose of which is to possibly maybe save 1mA is wasting 12mA by just doing the saving.

If it works, then it could be usable with semi-faulty packs which have high-leakage cells. Although such cells could be a fire risk anyway.

[shapirus]

The quiescent current is about 12 mA

...a significant part of which is the indicator LED. Can be removed if necessary, and then again, there's the on/off jumper which can be shorted, for example, only when the charger is connected.

So an energy saving device the only purpose of which is to possibly maybe save 1mA is wasting 12mA by just doing the saving.

It's not an energy-saving device, at least, it's not its primary purpose. Having a balancer is essential to prevent an uneven degradation of the cells and, in case there is no per-cell voltage monitoring, to prevent cell damage.

When there is no balancer, even with well-matched cells, disbalance, even if tiny in the beginning, increases over time, cycle after cycle. When the battery is used in buffer mode (e.g., a typical UPS), battery without a balancer will slowly drift out of balance even if no switching to battery occurs, because the self-dicharge rate of the cells is inevitably unequal while the charger will keep the total voltage of the pack constant.

Different types of balancers exist. Some require the user to make sure that the cell voltage at which they start balancing is actually reached at the target charging voltage of the entire pack (so as not to end up with one cell being at 3.65V and others at 3.35V or so).

For LiFePO4 packs, especially in buffer mode (float charging) and/or with final charge voltage well below the typical 3.60V/cell (say 3.35-3.45V/cell) these capacitive active balancers are what I find to work best for me. It's as plug-and-play as it gets, no worries about the acceptable voltage range or anything else. I think they also go into a soft power-off state when the cell voltages go below a certain threshold to prevent overdischarge (such as when the battery is kept unattended for a long time), but I'm not 100% sure about this.

For LiPo packs used in the cyclic mode, those typical for the RC hobby use, I'm satisfied with the built-in resistor-based balancer in my Turnigy Accucell 6.

[Siwastaja]

It's not an energy-saving device

Compared to a classic dissipative balancer, it is. Both perform the job of keeping cells balanced just as well.

There is, of course, a corollary to saving energy: larger balancing currents also become possible because power dissipation is not a problem. Although I highly doubt the need for such high balancing current. My experiments on aged, damaged, second-life 240Ah LFP pack showed that 40mA balancing current is sufficient, with a proper algorithm. Other products of the time used dumber algorithm and worked well with 200mA-500mA. I find it extremely unlikely that 5A balancing current is truly needed.

Voltage ranges, user experience, plug/play or not, is totally independent of if the underlying type is dissipative or redistributive.

OP is considering designing their own BMS, for whatever reason. "Why do that when you can buy an off-the-shelf product" is useless advice, although one can and should look at off-the-shelf products for motivation.

[Siwastaja]

If a redistributive balancer with high current capability continuously balances the pack (at any voltage), there will be two sides to this coin:

1) if the cell capacities differ a lot, usable pack capacity will increase. Classic top or bottom balanced pack gets the capacity of smallest cell. Continously balanced pack gets the average capacity of the cells (assuming 100% efficiency for redistribution)

2) if the cell capacities do not differ a lot, but ESRs do differ, usable pack capacity will decrease, as the balancer misidentifies small voltage differences as the cells being at different SoC even if they aren't, and redistributes charge incorrectly, having to then later correct itself. Energy is wasted in this back-and-forth redistribution.

This uncertainty is why I'm not a big fan of continuous redistribution. It looks nice when the voltages are seemingly equal, but equal voltage at every moment is not exactly the right metric to look at.

The big case for this continuous strategy would be packs where cell capacities are unstable and large differences emerge. With current li-ion not having such a problem, this future chemistry with said problem should be competitive in some other way.

[shapirus]

OP is considering designing their own BMS, for whatever reason. "Why do that when you can buy an off-the-shelf product" is useless advice, although one can and should look at off-the-shelf products for motivation.

It depends on the goal. If it's only to get a working product in the end at a reasonable price, then an off-the-shelf component is often the best solution. If it's also for educational purpose, then yes, DIY may be the way to go, however, it's necessary to understand that in this particular case implementing a DIY device may be too complicated (if it is to be done properly) to prefer it over a ready-made one. It very much depends on personal preference, so I'm giving food for thought rather than an advice.

[Siwastaja]

It depends on the goal. If it's only to get a working product in the end at a reasonable price

Looking at paulca's threads, it seems to me he wants to explore all kind of ideas, some totally impractical. So DIY designing a complicated active balancer is not a surprising idea from him. Of course, if one just wants to install a battery pack, they use off-the-shelf components. And if one just wants to develop a reliable and usable BMS in least amount of time, while still learning the relevant concepts, they should use the dissipative balancer design because it achieves the same, but is simplest to develop. On the other hand, if one wants to explore as much design space as possible, prototyping active balancers is great fun. My design over a decade ago, designed for 300-400V packs, used a flyback converter per cell, bidirectionally delivering to/from a common 30-40V "balancing DC bus". I hand-wound the transformers, and got finally to some 70% charge transfer efficiency, with synchronous rectification, before losing interest and just develop a simple distributed BMS with dissipative balancing, which ended up working great and being used in a few projects.

[paulca]

The reality in my case...  today, battery 80-90% full, taking 20 amps at 14.xV and the min-max balance was 0.009V.

Now the panel has fully absorbed them up to 14.40+V I expect a different story.

[rteodor]

I am considering building a battery and while enumerating the available options of BMS-es and balancers I ran out of paper.

My current understanding is that active balancing is not really needed because its more expensive and then is excess energy anyway. Still there might be cases for active balancing:

1. bottom balancing (I did not see anybody doing that). Probably done during a polar night when conserving every bit of energy is important.
2. large and aged capacities might need active top balancing. Even more so if charging is done at lower voltage levels (like 3.45 bulk / 3.35 float for LFP). That is because the charging current is bigger at lower voltages and aged cells will have larger differences in capacity.

Fun to see how the world goes on, and it seems Chinese are now producing cheap redistributive balancers and people are apparently buying them. Didn't know that. And I'm sure they work better than the overly expensive and complicated products I have seen years ago.

Andy from Off Grid Garage (yt channel) noticed the following about Heltech capacitive active balancer: it is terrible if its used continuously (unbalancing the cells actually) but it does a wonderful job if its only used at top in float. This balancer seems to work great with a BMS capable of activating an external balancer (some JBD variant in his setup). The balancer is also unable to balance advertised [5 or 10] Amps even in a test setup with fully charged and fully discharged cells in the same pack. The voltages are just not enough to drive such currents.

And there are also inductive balancers (Heltech has one too).... And there is JK BMS with integrated active balancer....

[shapirus]

Andy from Off Grid Garage (yt channel) noticed the following about Heltech capacitive active balancer: it is terrible if its used continuously (unbalancing the cells actually)

I wonder what the application conditions were. I've never noticed anything like this with mine (which I have three of). Whatever happens, it always tries its best to equalize the voltages. Yes, it may fail to balance fast enough, when the charge/discharge current is too high and the cells are too unequal (and the advertised balancing current is not reached, yes). But it doesn't mean that it makes things worse.