Warning: bad BMS with almost-fake balancing

[anpaza]

Thought it would be good to warn anybody who's going to buy a BMS.

Recently a friend brought his battery (16S LiFePO4 A123 "20Ah" cells) because it lost some capacity. With the nominal 18-19Ah it supplies only 15Ah after several months of everyday usage. He was blaming the battery to have a bad cell.
After inspecting the battery I discovered that it was simply disbalanced, and all the cells were in good state. The bottom four cells were sligthly discharged (that's because the logic part of BMS draws current from them) and the 10th cell also was a bit discharged, compared to the rest.

But how that happened, he used a BMS with balancing circuits?

Upon inspection of the circuit I came to the conclusion that the "balancing" function of the BMS is purely decorative and doesn't really work.
So I thought it would be good to warn everyone about it so that bad quality BMS producers will get what they deserve.

Here's the BMS picture:


Here's another (green) variant of it, the reverse side:


The technical details, if anybody is interested.

Most cheap BMSes use cheap battery monitoring ICs used in notebook batteries and alike. Those ICs usually contain two voltage comparators, one for "overvoltage" and another for "undervoltage". The BMS that was in my friend's battery used Epson S-2861 ICs, with the "overvoltage" threshold of 3.9V (for LiFePO4) and "undervoltage" of 2.0V. Both limits have a hysteresis, in our case 0.1V for "overvoltage" (e.g. it disables charger at 3.9V and re-enables it when voltage drops to 3.8V) and 0.3V for "undervoltage" (e.g. disables load at 2.0V and enables when voltage rises to 2.3V).

Now this IC doesn't supply a third comparator that is needed for "balancing". So most cheap BMSes will contain yet another small IC (usually a TL431) that compares cell voltage to a relatively high, but lower than "overvoltage" limit (usually something like 3.6V) and when the cell voltage is over that limit, it will enable a small load on the cell (the large black rectangle 100 ohm resistors you see on the picture above, labeled "101"). This should allow all overcharged cells to be slowly discharged (all at a time) until the voltage on the most charged cells drop below 3.6V, allowing less discharged cells to catch up.

So I was interested to see why the balancing resistors are present on the above BMS, but they don't work as expected. And guess what... they use the "overvoltage" signal from S-8261 to enable the balancing load! Yes, saving a few cents per every BMS (on TL431 and two resistors), hurray!

So, the BMS "work" like that: suppose you charge the battery and some cell reaches 3.9V (ouch!). When this happens, the S-8261 will signal to disable the charging channel (one of the two groups of power MOSFETs). And at the same time enable the balancing resistor. Then the cell will slowly discharge to 3.8V (which took about 30 minutes in my case) and after that the "balancing" will be finished. Got it? Only one channel at a time - the first cell to reach 3.9V will be discharged to 3.8V and that's all the "balancing" that the BMS does. Then charging process resumes again, some other cell reaches 3.9V and the cycle repeats.

If you happened to be one of those unfortunates with such a BMS I would suggest to rise the charger voltage to 3.8V * number_of_cells (60.8 volts in my case) instead of the "standard" 3.65*number_of_cells for LiFePO4. But I won't suggest to do this for lithium-polymer, because charging them to 4.3V is dangerous (and the Li-Po variant of this BMS uses an "overvoltage" of 4.3V, ouch again!).

Also another thing that may help a little bit is to solder additional resistors in parallel to balancing resistors. I have soldered a 100 Ohm resistor on top of every balancing resistor and thus shortened the length of the "charge to 3.9V - discharge to 3.8V - enable charger again" cycle about twice. So, if you sometime leave the battery on charge for all night, that may compensate the accumulated disbalance, if it is not too large. This won't work with every charger, though - some models will turn off and never turn on again until you pull the plug out of the wall outlet.

 

[dogman dan]

This is why we say leave it on the charger overnight. If really unbalanced, leave it on the charger a week or more.

Also why we say charge it when you can, don't wait till it's completely drained to charge causing it to get unbalanced worse than if you charged more often.

These bms's can take a long time to balance a pack, so you should add some balancing plugs to the wiring. Same thing you see on an RC lipo pack, they allow the use of cell checkers to monitor how the pack is doing, and also the use of a wide variety of methods for balancing the pack a lot faster.

 

[anpaza]

These BMS have two big problems: the first is that it "balances" one cell at a time (vs multiple cells at a time like most other BMSes do), so it will take "years" to balance a slightly disbalanced battery.
And second is that the balancing voltage threshold is way too high, so if your charger uses standard output voltage, the battery may get disbalanced (say, some cells 3,8V and some 3,3V) and the BMS won't even try to do anything about it. And if you rise the charger output voltage, you will constantly stress your cells with high voltages, it varies from manufacturer to manufacturer how cells will react to this, some people are reporting A123 20Ah cells to inflate when charged to high voltages.

 

[dogman dan]

I agree, you got a problem with that bms. That's why I think your buddy needs some balance plugs on the pack. Then he could just plug in a 12v lightbulb into each high cell one by one and manually balance that thing from time to time, or use an RC charger to run a balance charge from time to time.

For example, with 8s balance plugs, and a way to connect to half of the battery at a time, he could balance charge with an 8s RC charger. If the pack is used carefully, this would need doing at most once a month, or less. Or you can just bring up one or two low cells one at a time. Instead of waiting forever for discharge, simply charge the low ones.

Stuck with nothing but that crappy bms would be a pain. The other obvious option is of course, dump that shitty bms for a better one. But even the "good" ones still take a long time to balance a pack. Just like the RC charger will.

 

[cwah]

I'm not sure but I think my bms from bestechpower behave the same way

 

[dnmun]

i did not see any evidence that you presented to make your case that only one channel is balanced at a time.

i did not see any evidence that there was anything wrong with the balancing function.

i did not see evidence that you actually measured the voltages across the shunt resistors in order to determine if the shunt transistor had turned on.

i do not think you have proved your thesis that the BMS does not work.

balancing does not start at 3.9V. the HVC function is separate from the balancing and it is handed by a different circuit.

i can see signal circuits for both HVC and LVC on the front and back and the shunt transistors also. it is not missing any functional parts.

i think your argument is inadequate to prove your thesis.

that BMS works and you just were not able to determine how it works.

 

[anpaza]

i did not see any evidence that you presented to make your case that only one channel is balanced at a time.

That's evident from the BMS schematics, and also is evident in practice: I hooked the charger (well, a 0-60V bench top power supply in fact) and left it for all the night, then from time to time inspected the BMS. At any given time only one balancing resistor was hot; that's the one connected to S-2861 that triggered first. The rest of S-2861 weren't given a chance to trigger, because the charging channel is cut off by the same S-2861 that triggered, so they will just silently wait till the triggered S-2861 will sense that the cell was discharged below 3.8V.

i did not see any evidence that there was anything wrong with the balancing function.

The evidence is that the battery got disbalanced by about 3 Ah in a couple of month of usage.
A well balancing BMS won't allow the battery to disbalance even if the cell quality is not so good.
For example, for four years I use a BMS bought together with a cheap LiFePO4 battery from ebay and the battery is still perfectly balanced, despite the fact that it's built of rejected A123 cells (those yellow ones with the 'F' (failed) leter on them).

i did not see evidence that you actually measured the voltages across the shunt resistors in order to determine if the shunt transistor had turned on.

Actually I did it, but its just easier to touch shunt resistors. The one that is enabled is pretty hot (its dissipating 0.3W of heat), the temperature of all others is same as the temperature of the board.

i do not think you have proved your thesis that the BMS does not work.

In fact, I didn't meant to prove anything. I just wanted to warn those who's going to buy a BMS to avoid these ones.
If you plan to buy one like this and don't feel my arguments convincing, just go ahead and buy.

balancing does not start at 3.9V. the HVC function is separate from the balancing and it is handed by a different circuit.

Please point your finger at it. Which one? There's just one IC per channel, two PNP transistors (used to pass the CO and DO signal down) and one PMOS switch that enables the shunt resistor.
Besides, I just see with my own eyes that balancing starts at exactly 3.9V (connected a voltmeter to the most charged cell and followed the voltage until the charger cuts off). Other channels were showing 3.6-3.8V and the voltage was steady (my voltmeter shows four digits after decimal point, e.g. like 3.7864). The balancing channel was showing a constantly decreasing voltage, as expected.
The exact model of the used IC is S-8261ACEMD-G4ET2G. You may look in the datasheet what thresholds it uses.

i can see signal circuits for both HVC and LVC on the front and back and the shunt transistors also. it is not missing any functional parts.

As I told, it's missing a third comparator per channel. Look at the trace coming down from the pin 3 of IC. It's directly connected to the gate of the MOSFET. That's the CO pin.

that BMS works and you just were not able to determine how it works.

omg
Are you selling them or what?

 

[dnmun]

no, i don't sell them, i just work on them. i know how to analyze them to determine what is wrong and you did not do anything to determine there was anything not working.

you could still go back and test it if you are interested, but you kinda already decided so i doubt if you will.

 

[dnmun]

ping uses the G4EN as the 6 pin comparator also.

if you wanna figure out if it works, which it does from what you said, then put the battery on the charger and measure the cell voltages and list them here so we can give you advice on how to get the pack in balance again. measure while charging.

find a 5ohm 5W cement power resistor and solder alligator clips on jumpers to each end to use to manually force the pack back into balance after you post up the cell voltages.

we will be able to get the pack balanced and working at full capacity again for you.

 

[silviasol]

I found a bms will only charge for a certain time if the cells are unbalanced and some already at max voltage. Using a lower amp charger will let the bms keep up with the power at full charge. Most chargers will have a lower charge state at the final voltage so the lower that voltage is the longer it will charge at the end of the charging cycle. Do short runs and keep charging until it is max and it should keep up. If you have one cell at say 3v and the others are full charged there is no way for it to get to full charge unless you bring that cell to a .2v less difference based on most specifications of bms boards. This can take time since a bms will only eat up so much power while charging, a fraction of the power the charger will charge with.

 

[dnmun]

that is why you have to manually balance the pack when it gets so far outa balance. coulda helped that guy but he knows it all.

 

[anpaza]

ping uses the G4EN as the 6 pin comparator also.

That's not true.

Here you can find the schematics for the SIGNALAB BMS that Ping sells.
Now look at this part:

A traditional more or less working balancer (using TL431 to feed ~33mA current to balance resistor directly... hmm, it's on the edge, but should work).
Which will work in parallel with same circuits from other channels, which has no hysteresis and which does not disable the charging channel.

 

[dnmun]

he doesn't use that circuit anymore. that was only for the v1 signalab. the V2 signalab uses the the G4EN for the HVC/LVC comparator and another chip, R2RP for the balancing comparator/ gate driver.

if you decide you wanna balance that battery we can show you how. put it on the charger and post up the cell voltages.

make a discharging load from a 5ohm 5W cement power resistor and solder some jumpers on the ends with alligator clips so you can drain charge off the high cell that is shutting down the charger.

 

[anpaza]

he doesn't use that circuit anymore. that was only for the v1 signalab. the V2 signalab uses the the G4EN for the HVC/LVC comparator and another chip, R2RP for the balancing comparator/ gate driver.

Okay, so it anyway has one extra chip for balancing, no matter how it's named - TL431 or R2RP or whatever.
The BMS in the first message of this topic doesn't have anything, that's the point.
And the battery was balanced long ago, before I posted the first message. It shows 18Ah now, rather than 15.

 

[dnmun]

the comparator has 6 pins. Vdd and ground is two on the lower side in that picture, you can see the leg between them that goes to the via that goes to the gate of the shunt mosfet, the upper three legs drive the HVC and LVC signal transistors, and there is one that is gonna be a NC, not connected. it should be tied through a resistor to ground if i am not mistaken. see if there is a 100 ohm surface mount to ground. i cannot see anything when i increase the magnification because there is not enuff definition in the picture.

so all the functions are covered by that one comparator.

my point was that you did not prove your point and show it does not balance by establishing that there is no shunt current when the cell voltage exceeds the 3.6V level. you can still do that.

you did establish that the HVC signal line is working because it cut off the charging. you can also test the LVC by discharging the pack until it hits LVC and the output discharge mosfets are turned off.

so that was the point, you did not prove that it does not work. you decided it doesn't work because the pack was out of balance and hit HVC. that was all.

you stated it requires two different comparators to handle all these functions and that is also wrong because that single 6 pin comparator does handle all three functions and you can see the traces on the pcb.

 

[Dlogic]

For my custom built 6S*4P 18 cell Turnigy Nano Tech 8000 mAh cell lipo pack i use 3 i-chargers in parallel. A 42 pin plug is used to establish all balance leads and battery + and - connections at once. Then all i do is push 3 start buttons on the i-chargers and the 3 big packs, made up of 4 individual packs each, get a safe balance charge.

For riding another plug is inserted that makes the serial connections amongst the 3 packs giving me the desired 72 volts at 32 Ah. It does take long to figure it all out and solder the plugs. After that though it´s just a matter of plugging in and charging up. With this setup there´s no need for a BMS as i´ve integrated 3 battery medics into the cnc machined dashboard, giving me a precise readout of every single cells voltage.

Prior to assembly all cells in those packs where charged one by one with a single cell charger. This is important as the i-chargers will have a hard time if the packs are to much out of balance. Not with one pack, but since there´s 4 in parallel i thought that it might take too long for the resistor banks in those chargers to balance them.

Bulk charging is another option, but i don´t mind a bit more work to obtain safety. There´s just to many horror stories out there when referring to lipo.

So far my packs stay in perfect balance. At a charge current of 18 A, it takes just 2 hours for the juice to return.

The charger disconnected.

The charger connected.

To prevent the medics from draining the packs, a 24 pole switch was installed. With a single turn all balance connections to the medics are cut. This switch also acts as my ignition for the motor controller. So far this setup has given me 187 cycles with no problems at all. For the next bike that is just being built the pack will be 6S 8P. This will make the whole setup even simpler and requires only 2 i-chargers.

 

[anpaza]

the comparator has 6 pins. Vdd and ground is two on the lower side in that picture, you can see the leg between them that goes to the via that goes to the gate of the shunt mosfet, the upper three legs drive the HVC and LVC signal transistors, and there is one that is gonna be a NC, not connected. it should be tied through a resistor to ground if i am not mistaken. see if there is a 100 ohm surface mount to ground. i cannot see anything when i increase the magnification because there is not enuff definition in the picture.

S-2861 is not a comparator, it is a battery management IC. Unlike plain comparators, it has hysteresis, time delays and other things.
There are two unconnected pins: pin 2 (the upper middle on the top photo) and pin 4. They aren't connected to anything.
Pin 1 and 3 (on the top side on the top photo) are DO (discharge output) and CO (charge output) respectively.
Well, the datasheet is not secret.

so all the functions are covered by that one comparator.

Only overcharge+hysteresis and overdischarge+hysteresis. Balancing function imposes a third comparator, which will compare cell voltage with some intermediate value (e.g. 3.5-3.6V).
S-2861 doesn't have a third comparator, so it has to be provided separately.

my point was that you did not prove your point and show it does not balance by establishing that there is no shunt current when the cell voltage exceeds the 3.6V level. you can still do that.

Of course I did that. I measured the voltage across all balancing resistors, as I said. I have seen the cell voltage only over one of them, the rest shows zero. The voltage on most cells was above 3.7V (except 5 disbalanced cells and one that was actively balancing), and across the balancing resistor was 0. Also, only one resistor was heating. Isn't that enough for proving?

You say you use these BMSes (exactly these, I don't mean PING or whatever else). Have you ever seen it is balancing more than one cell at a time? Did you measure the voltage across the balancing resistors? Touched them to see which are heating?

you stated it requires two different comparators to handle all these functions and that is also wrong because that single 6 pin comparator does handle all three functions and you can see the traces on the pcb.

I stated that you need three comparators, not two. You have to compare with discharge voltage limit (2.0-2.5V for LiFePO4 usually), overcharge voltage limit (3.7-3.9V usually) and balancing voltage limit (3.5-3.6V usually). You can't achieve this with two comparators in any way.