Active balancer for lithium batteries: what it is and why you need one

I’m often asked about various battery packs, and the questions are about both battery protection and balancing. For a UPS conversion project, I needed a battery pack, a protection board, and a balancer. The battery is still on its way, and there’s no point testing the protection board without it, so today I’ll tell you about the balancer.

Actually, I’ll be covering two balancers, one of them in two versions.
The small ones were ordered for LiFePO4 battery packs: one for a high-power UPS, the other for a smaller one. I bought both boards from the same seller in two separate orders, but the seller has two different listings:
1. $3.59 + $1.31 shipping; I bought it with an 11.11 coupon
2. $3.50 + $4.73 shipping. This listing seems too expensive because of the shipping, but the seller doesn’t combine shipping costs for different items. Since I was ordering a protection board, I added the balancer to the order.

The second board was bought on Taobao about a year and a half ago, but I never got around to testing it. So I decided to cover both boards in one review.


As some extra background, I’ll cover the questions I get most often: what a balancer is, what types there are, and how they differ.
I don’t think there’s any need to explain what it’s for—that’s clear from the name, and I’ve already covered it before.

Passive balancing circuit

So, technically, the simplest type of balancer is a passive one. Its job is to limit the voltage on cells that are more highly charged than the others. It only works during charging, diverting some (or all) of the current through itself so the other cells can charge.
For obvious reasons, this type of balancer can only work while charging—or, more precisely, at the end of the charge cycle.


It’s clear that limiting the voltage with Zener diodes is problematic because of a few technical issues, including a limited choice of voltage ratings and variations in their characteristics.
You can replace the Zener diode with an adjustable one based on a TL431, but it can’t handle much power, so you’ll need to add a transistor to boost its output.


This circuit provides very accurate balancing and can do it in a single charge cycle, but it has at least two problems:
1. Since the current in the circuit never drops to zero, but simply flows either through the battery or through the balancer, the charger can only shut off using a timer.
2. Since all the current then flows through the balancer, a lot of heat is generated at the end of charging. With large battery packs, this means you need substantial cooling. Needless to say, installing this type of balancer inside the battery pack is out of the question.

Here’s an example of a balancer with parallel shunt regulators, rated for currents up to 2 A and packs up to 20S. At maximum power, it dissipates up to 170 W.


But ideally, you’d want to build the balancer into the battery pack itself, which means reducing heat dissipation. This is done by adding resistors.
When the end-of-charge voltage is reached, the control circuit starts shunting the battery through a resistor, diverting some of the current through itself. In this case, that’s resistor R1.


These balancers are often built right into protection boards. They’re easy to spot: usually there are several large, low-resistance resistors placed next to each other, and the number of resistors matches the number of batteries that can be connected to the board.


Photo of a 5S protection board. You can see the 150 Ω resistors connected in pairs, so each balancer can draw around 50–55 mA.


There are also separate boards made for different numbers of channels, usually 4 or 8. If your pack has 5–7 cells, you use an 8-channel board and simply leave the extra channels unused. I happen to have a few of these boards lying around.


But even that doesn’t help much: the resistors can heat up to around 80 °C at the end of charging, and inside a battery enclosure the temperature can be even higher. And there are lithium batteries right next to them...

Overall, the advantages are simplicity and low cost. The drawbacks are low balancing current, which means that if the imbalance is very large, the charger will still “pull the blanket over to its side.” To avoid this, you need to charge at a low current. Also, remember that the balancer draws some current, so if your charger cuts off when the current drops below a set level, you need to take that into account.
For example, the charging current is 1 A, and the cutoff is set to 50 mA. With a 60 mA balancer, it will never shut off. In this case, set the current cutoff to 50 + 60 mA = 110 mA. Then the charger will shut off when the current through the batteries themselves drops below 50 mA.




Active balancers

To get around the problems described above, people came up with circuits that transfer energy from one cell to another. A capacitor balancer is relatively simple. The principle is straightforward: first, a capacitor is charged from the cell with the higher voltage, then switched to the cell with the lower voltage.
As a result, the more highly charged cell gradually transfers some of its charge to the less charged one. In effect, the cells are "virtually" connected in parallel this way.

Ultimately, the circuit's job is to equalize the potentials at the cell terminals. And here I'll answer another common question—or rather, two:
1. Can this kind of balancer overcharge the battery? No, it equalizes the potentials, just as connecting the cells in parallel would. Roughly speaking, if two cells are at 3.5 and 3.7 V, after balancing they'll both be at 3.6 V.
2. Do different batteries need different balancers? No. Since it simply equalizes the voltage, it doesn't matter what that voltage is, as long as the controller itself can operate. That's why these balancers are usually universal, working with LTO, LiFePO4, and "regular" Li-Ion batteries alike.


With the previous circuit, the batteries can simply be connected in parallel. But if you need to balance cells connected in series, you just add another switch; the basic principle stays the same.


A few years ago, I published a review in which I made a 2S battery charging board and put an active balancer based on the 7660 chip on the same board.


Strictly speaking, this IC isn't a balancer; it's just a negative-voltage generator. But in this case, it can also be used for this somewhat unconventional application.


The balancer has low power and works slowly, but it has one advantage: it runs all the time, day and night, for months on end.
That's also a drawback to some extent, since the circuit constantly consumes power, albeit not much. In my case, this wasn't critical because the batteries had individual protection, so overdischarge wasn't a concern.

In the end, the balancing table over two days looked like this.


A natural question is how to balance more than two cells. The same way: you just use more balancers. The number of balancers is always one less than the number of cells.

The first balancer equalizes the voltage of cells 1 and 2
The second, cells 2 and 3
The third, cells 3 and 4
The fourth, cells 4 and 5.

As you can see from the circuits, no matter how the voltages are distributed among the cells, the balancers will still bring them to roughly the same level. The biggest challenge is when the largest voltage difference is between cells 1 and 5, but even then the voltages will equalize.


Modern capacitor balancers are, of course, much more sophisticated: dedicated controllers that put the circuit into sleep mode, polymer capacitors, and balancing currents of up to 5 A. But the prices are impressive too: an 8S balancer can easily cost around 25 dollars, and I'm afraid to even think about the price of the monster shown below.

The advantages are that it works all the time and provides a high balancing current. The downside is the price.


The second type of active balancer is the inductive type. It's essentially the same as a capacitive balancer, but the energy transfer works a little differently: inductors are used as intermediate energy storage elements.

The advantages are much the same as with a capacitive balancer, but the current is usually lower, around 1–1.5 A. On the other hand, it's noticeably cheaper.


And of course, the question is: which one is actually better? In my view, the active type, naturally.
The thing is, with an active balancer, you effectively get the battery's total capacity without being limited by differences between the cells. With a passive balancer, even after balancing, you still only get the capacity of the weakest cell. There is one caveat, though: the result depends directly on the balancer's power and the discharge current.

To simplify, let's take a series-connected pack of three cells rated at 1, 2, and 3 Ah.
With a passive balancer, you'll get 1 Ah, because even after equalization, the 1 Ah cell will be the first to discharge, and the protection board will disconnect the load.
With an active balancer, charge is constantly taken from the more highly charged cells and transferred to the weakest one. In theory, you could get the average capacity—in this example, 2 Ah—but the balancer's efficiency will of course reduce that figure.

What does this look like in practice? You use a tool, then take a break. While the battery "rests," the balancer transfers energy to the weakest cell, and then you get back to work.
There are drawbacks, too. At high power consumption (for example, in a UPS), it won't help much. Also, the battery wears out more in this setup because it effectively goes through charge/discharge cycles. But it's up to you to decide what you're willing to sacrifice.


Alternative balancing circuits
1. The option of disconnecting charged cells and removing them from the circuit has been mentioned, but apparently the implementation difficulties and limited benefits killed the idea—especially since the switching circuits also carry the operating current.
2. Charging each cell with an independent charger. Essentially, the result is the same as with a transistor-based passive balancer: good efficiency, but the same drawbacks in terms of reduced capacity and the need for a multichannel power supply. One example is the ImaxRC B3 PRO charger.
3. Balancing using an isolated DC-DC converter. It is similar to an active balancer, but technically more complex, so there is little point in using it. This type of balancer is also called a two-level balancer, since it often works together with a passive balancer.



As you’ve probably guessed by now, this is about an inductive balancer: two boards for 2-4S packs and one for up to 10S.

The number of cells a balancer is designed for is the maximum; you can connect fewer and it will work just the same. Boards with more channels simply cost more.


The package includes connection wires. The smaller boards come with two wire colors: a common black wire and red wires for the cells. The larger board has multicolored wires, which is a little more convenient.


The smaller boards are identical, unsurprisingly. The underside of the larger board has a matte finish, which feels a little unusual; the smaller ones are glossy.


You can see three balancers on the smaller board, along with fairly large inductors. The specified balancing current is 1.2 A. The maximum channel voltage is limited to 4.3 V by small Zener diodes, so it’s best not to exceed that.


The larger board has the same specified current of 1.2 A, but the controllers have a different package, and the inductors are clearly smaller. The board is universal, supporting up to 11S, with one position left empty. It also has nine LEDs indicating the balancing process for the corresponding pairs of cells.
On top of that, this board is coated with a decent layer of protective varnish.


Both boards are built around the dedicated ETA3000 controller.


The datasheet includes a typical application circuit and notes that the chip comes in two package variants, which is also evident on the boards shown here. I already knew about the first type, with the square chip, and the smaller board even has the controller’s name printed on it.


And in this case, it really is a chip originally designed for balancing circuits. It can detect voltage differences between cells, enter sleep mode to reduce power consumption, and even indicate that balancing is in progress.


There’s also an example showing how to connect several controllers for larger packs, but the principle is the same as the one I showed above: each controller handles a pair of cells—1-2, 2-3, 3-4, and so on.


The balancing current can be set anywhere from 0.1 to 2 A; a table of component values is provided for this purpose.


Let’s move on to the tests.
For testing, I used a pack of four 5700 mAh LiFePO4 cells from this review. It had a protection board with a passive balancer, which I had to disconnect so it wouldn’t interfere. Obviously, you shouldn’t do this, but everything was under constant supervision.

Before disconnecting the board, I fully charged the battery.
Then I disconnected the board and desoldered its wires.
To simulate an imbalance, I partially discharged the cells. Since I knew their capacity, I simply discharged them at 5.5 A for a set time: 38 minutes for the first cell, left the second alone, 19 minutes for the third, and 57 minutes for the fourth. This gave me an approximate state of charge for each cell:
1. 35%
2. 100%
3. 70%
4. 5%

The last two photos show the discharge time and the capacity “taken out.”


Cell voltages, from the first to the fourth. The order in the photos will be the same throughout.


I connected the balancer using the supplied wires, although for faster operation it would be better to either shorten them or replace them with thicker-gauge wires.


The board gets warm during operation, but I wouldn’t say it gets very hot. Below are three thermal images: taken about 5-10 minutes after startup, then after an hour, and another hour later. The temperature peaked at 47 degrees. The components responsible for cells 4–3 and 2–3 were heating up, clearly because charge was actively being “transferred” from the second cell (fully charged) to the third, and then to the fourth (almost completely discharged).
I stopped checking at 4 a.m.; at half past nine in the morning, the board was cold.


The test ran for a long time, although, as it turned out later, there wasn’t really any need for that—and you’ll see why from the graph, too.
Thirty-four hours after the test began, the cell voltages looked like this.


Next, I ran two experiments. First I connected the smaller board; after an hour, nothing had changed, so I disconnected it and connected the larger one instead.


Since I was desoldering the wires, I also checked the board’s own current consumption. The negative rail drew 1.77 µA, with occasional peaks up to 6 µA; that’s the board’s automatic circuitry at work. The current on the B4 rail was slightly higher, at 2.14 µA, with similar spikes up to 6 µA.


I let it run for another hour and a half; still, no changes. The multimeter was a big help here, as it could track changes with a resolution of 0.1 mV.


For the next test, I connected the setup to a charger but left the protection board disconnected. For safety, I monitored the voltage on the second cell, since it was the least discharged.
When the battery voltage started rising sharply near the end of charging, the first LED on the balancer lit up, and it began “transferring” charge to the first cell.
A short while later, three LEDs were lit. Energy was now being drawn to charge the other two cells as well—the third and fourth. This is because LiFePO4 has very flat discharge and charge curves, followed by a sharp drop or rise at the end. So the battery charged up and its voltage began rising sharply, while the voltages on the other cells lagged behind, and the board started “feeding” them.
It took 1309 mAh to fully charge the second cell. Remember, it was fully charged at the start of the test, so that’s the capacity the board “transferred” to the other cells. But keep in mind that the average state of charge was even lower, since some energy was also drawn from the third cell, which was at 70%.


I disconnected the charger. All three LEDs stayed lit for a while; after a few minutes, two went out, and a couple of minutes later, the last one went out too.


And now, here it is as a graph.
Red arrow: two hours after the test began, with 30-minute intervals.
Green arrow: nine hours after the test began, then hourly intervals. After seven hours, I switched to two-hour intervals.
Blue arrow: additional charging of the battery until the second cell was fully charged, followed by another 16 hours. The intervals were half an hour at first, then one to two hours, and six hours for the final interval.

As you can see, there’s no point in monitoring it for long. After just a few hours, even with a large imbalance, the voltages “settle down,” and there’s very little change after that.


More or less the same thing was shown in the datasheet, which includes two graphs: one in standalone mode and one during charging.
It says balancing takes 3 hours, but as you can imagine, that depends on the balancer current and the battery capacity.


Next, I planned to expand the experiment. I had a hoverboard battery lying around for this. It’s built in a 10S2P configuration and has a capacity of 4 Ah. But it came to me pretty worn out, and a test I ran about a year ago showed a capacity of 2.2 Ah—that’s what’s written on the battery.
I connected the battery to the charger, but it didn’t charge for long before shutting off by itself; the protection kicked in.



The experiment was planned as follows:
1. Fully charge the battery
2. Fully discharge it and measure its capacity
3. Fully charge it again.
4. Connect the balancer and wait a few hours.
5. Put it on charge
6. Discharge it and compare its capacity with the result from step 2.

The battery discharged pretty quickly again, delivering just 724 mAh at 4.4 A. Oh well, maybe that makes it more interesting.
I charged it again until the protection board shut off; it took almost 800 mAh.


I chose this battery for a reason. First, it’s 10S, which is just right for the balancer. Second, it has a protection board inside, and the cells themselves are connected through a connector. There is one catch, though: the contact layout on the battery and the balancer is mirrored, even though the connectors are the same size. By the way, a friend who works with batteries like these said they’re all wired that way, but batteries for Segways have the opposite contact order, i.e. just like a balancer.

Quite a few of these batteries have passed through my hands, and the insides were more or less the same.


But apparently today wasn't my day, because this battery has a completely different protection board, with no connector installed at all. The cell leads are soldered directly to the board.


Well, I thought, maybe that's actually for the best. I'll solder the wires directly to the connecting strip from the batteries. But first I had to figure out the connection order and where to solder, and also measure the battery voltages. That's when I hit the second snag: one of the branches was hard-shorted. I decided not to give up and tried to "force it through" with a high charging current, even though you absolutely must not do that. Unfortunately, even at 18 A, nothing changed; the pair remained shorted.

I had to end the experiment there; I only had one battery.


Before moving on to the conclusions, I'll try to explain what I actually got from these experiments and discuss some specifics of using this type of balancer.
As advertised, the boards keep the difference within a pair to around 30 mV, but as always, "there's a catch."
The thing is, a 30 mV difference means one thing for "regular" lithium batteries and something quite different for LiFePO4. Below is a comparison graph from tests of two battery types with the same capacity under identical conditions.

You can see that the LiFePO4 curve is almost flat, so for these batteries, 30 mV represents a large difference in capacity—roughly 10–20%, depending on where you are on the curve.
With regular lithium-ion batteries, the voltage drops almost linearly, so the same 30 mV corresponds to a smaller difference in state of charge.

Put simply, this balancer will work much better with regular batteries than with LiFePO4, because a small voltage difference doesn't always mean a small capacity difference for this type of battery. I think the same applies to LTO as well.


Now for the conclusions.
I can say that this balancer is an excellent fit for "regular" lithium-ion batteries; you can tell just from how it works, even without testing. It works quickly, runs cool, and also draws very little current in standby mode. But it performs noticeably worse with LiFePO4, due not so much to the quality of its operation as to the characteristics of the batteries themselves. To be fair, this was also a very "brutal" test—in real life, such a wide spread (5–100%) is extremely rare.
I'll probably continue testing it, and I'll definitely test it with LiFePO4 batteries too, once the ones I ordered arrive.

The balancer should also work well alongside a passive balancer installed on the protection board. By the way, I haven't come across a protection board with a built-in active balancer yet.
Now I'm really keen to play around with a capacitor balancer, but the price is a bit discouraging: the four-channel version costs around $15.


That's all from me. I hope you found this useful.