4S Capacitor Balancer and a Little About How It Works

This is my third review of an active balancer and my second review of a capacitor-based one, also known as a capacitive balancer. Last time, I focused more on how it performs, but said almost nothing about how it works. Today, I’ll make up for that omission.


I bought this balancer instead of the inductive one, because it seemed rather weak to me, especially since I planned to use it with high-capacity batteries. But I won’t drag this out—let’s get to the review.

It comes in a small antistatic bag.


From the product description on the store page:

But unfortunately, it’s not all as rosy as it looks. I’ll get to that at the end of the review.

Inside are the balancer board and wires with a plug for connecting to the board. The wires are about 20 cm long, and they’re very stiff; I wouldn’t say they have a particularly large cross-section.


The top side of the board is almost entirely covered by ten 2200 µF, 6.3 V polymer capacitors.
Besides the connector, the board also has holes for soldering wires, and I recommend using those instead so you can use thicker wires.


If you follow the link to the store, attentive readers will notice that the photo on the product page doesn’t quite match the board in this review.
The store shows a board with nine capacitors and no connector, while I received one with ten capacitors and a connector. In the end, I had to find the photo for the heading on AliExpress.


The back of the board is even more crowded: four fuses at the top, 16 transistors below them, then the drivers and pulse generator.


Before moving on to a more detailed description, here’s a comparison with the board I tested previously. That one was for 6S packs, so it had almost one and a half times fewer components.


But note that the difference wasn’t just in the number of components. The transistors on the board on the right are clearly more powerful; it also has a DC-DC converter, and the board is coated with varnish for protection.


And here’s what people asked me about in the comments on the previous review: the schematic. I’ll also try to explain how this whole “kitchen” works.

First, here’s the overall schematic. In reality, the board has about a couple dozen more components; the small ceramic capacitors on the supply rails of each chip and the ones in parallel with the eight polymer capacitors aren’t shown.

To make the schematic easier to understand, I highlighted the circuits for the first and fourth channels in color. As you can see, the pulse generator is the popular NE555 timer. The pulses then go to four MOSFET drivers and finally to the power switches.


Since it’s hard to make out the connections in the full schematic, I’ve left just the oscillator, one driver, and one set of switches as an example.


Don’t be confused by the driver inputs being connected together. On the EG2131 used here, they’re inverted relative to each other: a logic high turns on the upper switch, while a logic low turns on the lower one. It’s a simple and convenient solution; apparently, the driver itself provides the dead time.


The power switches are IRF7455 MOSFETs with an on-resistance of 7.5 mΩ.


Now for how it works.
If we remove everything unnecessary from the schematic and leave only the power stage, then replace that with switches to simplify things, it looks something like this.
The upper switches turn on in sync and in antiphase with the lower ones.


The operating principle is very simple and consists of two phases. Because of the drivers, the lower switch actually turns on first, but for clarity I’ll show it in the more intuitive order.
1. The upper and middle capacitors are connected in parallel through the switches, while each channel is connected in series. The voltage across the capacitors is equal to the voltage across the cells.
2. The upper switches open and the lower ones close, connecting all the capacitors in parallel. Their voltages equalize, so the more highly charged capacitors transfer energy to the less charged ones.
3. The lower switches open and the upper ones close. The capacitors that gave up some of their energy are charged again by the higher-voltage cells, while those that received energy transfer it to the lower-voltage cells.

The diagrams show the first and second phases; after that, they simply alternate. The circuit runs continuously as long as it has power. Unlike an inductive balancer, it doesn’t transfer charge from cell 1 to cell 2, then from cell 2 to cell 3, from cell 3 to cell 4, and so on; instead, it balances the entire battery at once.


The operating frequency is about 7 kHz, and the board makes a quiet but unpleasant squealing noise while running.
You can improve its efficiency by increasing the frequency, but then you’ll also need to use transistors with a lower RDS(on). The load on the capacitors will also increase.
In any case, as you can see, you can build a similar balancer yourself, since it doesn’t contain any specialized chips—just an oscillator and sets of a driver, four transistors, and capacitors. The number of sets equals the number of battery cells connected.

A natural question might be: can you build this circuit for 100,500 cells? No, probably not 100,500, but 100–120 is quite achievable. The limiting factor is the maximum isolation voltage of the transistor drivers, which is usually 600 V. Also, above 15–17 V, you need a voltage regulator or converter to power the drivers and oscillator.


I tested the balancer the same way as the inductive one: I connected it to a 4S LiFePO4 battery pack with the protection board disconnected.
Nothing burned out when I connected it, and the operation indicator LED on the board lit up. It was so bright that you could easily use it to light a room. And since it was tiny, it really hurt your eyes.


As before, I first fully charged the battery, then discharged it to create a large imbalance. As before, the first cell discharged for 38 minutes at 5.5 A, the third for 19 minutes, and the fourth for 57 minutes. At the end of the preparation, the charge levels of the cells were as follows:
1. 35%
2. 100%
3. 70%
4. 5%


Since the voltage of LiFePO4 batteries doesn't vary much with their state of charge, the difference between the fully charged cell (2) and the almost fully discharged one (4) was just 0.5 V.


I didn't run a detailed test at short intervals, since essentially the same test was covered in previous reviews, but I did take occasional measurements to see what was happening.
Here are the results at about 2, 3, and 7 hours after the start. You can see that the voltages quickly converged toward an average. The biggest change was, of course, in the most discharged cell (4), though a 20 mV increase is hardly a major change.


I checked the temperatures during the test and found that the hottest spot wasn't around the transistors, but near the LED...
Temperature almost immediately after the test began and a couple of hours later.


After investigating, I found that the LED's resistor was heating up (I drew the schematic later). It's rated at 220 Ω, which, with about 10 V across it and a 50% duty cycle, gives an average current of around 22 mA and a power dissipation of 0.22 W.

Of course, I replaced the resistor. I first installed a 1.5 kΩ one, then a 3 kΩ one, but I think it could actually be increased to 10 kΩ. I used an 0805 resistor, which fit perfectly in place of the original.


1. The heating in this spot practically disappeared, and the NE555 turned out to be the hottest chip.
2. Interestingly, throughout all the measurements, both at the very beginning and at the end (after the charging cycle), the transistors handling the negative side of cells three and four got the hottest.


After about 15 hours, I noticed that the cell voltages were dropping. They were dropping even on the fourth cell, the most discharged one.
Results at 14, 26, and 36 hours after the start of the test.


As last time, after about 34–36 hours of balancing, the battery was charged until the most charged cell (2) reached 3.65 V. But whereas 1309 mAh was added last time, this time it was 3932 mAh, which points to two things:
1. More active balancing
2. Higher power consumption by the board itself, as well as the lack of a standby mode.


After the intermediate charge, I disconnected the balancer and measured the cell voltages. The maximum difference was 5 mV between cells two and four. Note that at the start of the test, cell two had the highest charge and cell four the lowest; now their positions had reversed.


As mentioned above, I didn't stick to any precise time intervals, but I decided to sketch out a small graph anyway. Below are the readings over about 50 hours, starting before the balancer was connected. The intermediate charging cycle is visible near the end.


I'm sure many people are wondering how much power all this consumes. I took the measurements after replacing the LED resistor, so you can add about 20 mA to the results.
1, 2. First, I checked indirectly: the battery was connected to a charger, and the protection board with its passive balancer was connected too. The battery was fully charged, but the current was below the set cutoff current (50 mA), so it would keep charging indefinitely. The whole setup draws 160 mA. When I disconnect the active balancer, the current drops to 90 mA, so the balancer draws 70 mA.
3, 4. I connected the board to a power supply without a battery and found that its current draw depends quite a bit on the supply voltage: at 12 V it drew 33 mA, but at 15 V it was already drawing 65 mA.

The description claimed:

But as we can see, that's not the case at all. The actual figure is four times higher, not to mention that it also says:


This basically answers the question of why so much charge was needed afterward. The 5700 mAh battery was charged to about 50% (5 + 33 + 66 + 100%), or around 2800–2900 mAh. After 36 hours of testing at a current draw of, say, 40 mA, about 1500 mAh had been drained from it. The calculation is rough, of course, but it still explains why the voltage on the fourth cell started dropping after a while, even though it had been rising before.


Well, the conclusions are basically the same as in the previous review, but I'll say a little more anyway.
The balancer works, though not quite as well as the 6S model I tested earlier. There's also a difference in build quality: at least this board isn't coated with lacquer.
I was surprised by the designers' decision to drive the LED at 20–25 mA, even though the board is always on and powered by the battery.
It also turned out that the balancer doesn't meet its stated specifications for its own current consumption.

Compared with a passive balancer, this one also has its own advantages and disadvantages.
The one reviewed here balances faster and more accurately, since it is always working.
The inductive balancer is less accurate and slower, but its dedicated controller puts it into sleep mode after balancing, so it consumes very little power—and it costs noticeably less.

Which one to use also depends on the battery type. For a "regular" lithium-ion battery, an inductive balancer will do; you won't lose much. But for LiFePO4, a capacitor-based balancer is definitely better. This is due to the characteristics of LiFePO4 batteries, specifically their charge/discharge curve.

Update
During the discussion, it turned out that the drivers and transistors actually impose a strict limit on the battery's maximum voltage. This is probably partly why the current draw depends so heavily on the supply voltage. Based on the test results, I can say that this board can't be used with "regular" lithium-ion batteries—unless perhaps with LTO, or as a last resort with LiFePO4, and even then only because I've already tested it in that configuration.
One solution would be to install a regulator that lowers the supply voltage to the drivers and NE555 to a safe level.