4S 40A protection board for LiFePO4 and an example of using it in a UPS

Recently I posted a review of 32700 LiFePO4 batteries, but as you can imagine, batteries can't be used without a protection board, so I ordered one as well. Since I had it in my hands anyway, I put together a short review in case it might be useful to someone.


First, the price. The seller listed it at $3.86 plus $0.88 shipping, which seemed reasonable to me. I was thinking of ordering several boards, but when I tried to order two, the shipping cost went up to $3.79. I could have placed several separate orders, of course, but I thought I'd try one first, since I've had a case before where I was sent a regular board instead of a LiFePO4 one. I might write that one up too.

Confusion between LiFePO4 and Li-Ion boards is pretty common, so you need to be especially careful and check the photos in the reviews. The boards aren't interchangeable, and there's no way to modify one to work in place of the other.


Anyway, after a while I received a small envelope with my board inside.


The product page includes a screenshot from the datasheet stating that the board can handle 40 A continuously and 80 A briefly, with a protection trip delay of 150 ms.


The board is sold in two versions, with or without a balancer. For cyclic use (power tools, for example), the version with a balancer is the better choice. For my purpose—replacing lead-acid batteries in a UPS—the basic version would have worked too, but the balancer didn't get in my way, so I decided to keep it.


The board doesn't have a central protection controller; essentially, it's a higher-power, multichannel equivalent of the usual DW01-based boards. Some people might say that's a bad thing. Maybe, but I personally prefer this kind, since they usually don't have problems recovering after a protection shutdown, when you have to connect the battery to a charger to reset it.

The board also has 10 AOD472 transistors, with an on-resistance of 6–9.5 mΩ. That gives a calculated resistance of 2.4–3.8 mΩ for the power stage, at a maximum current of up to 275 A.
There is no current-sense shunt on the board; the power transistors themselves perform that function, as they do in the DW01. There's nothing unusual about that, but if you want to lower (or raise) the trip current, you'll need to remove (or install) some of the transistors accordingly.


Here's the protection and balancing section.


The underside only has the markings, but you can already see that the traces are not only wide, but also reinforced with vias connecting the two sides of the board. This also helps dissipate heat.


The board was originally designed specifically for power tools. The seller's page even has a fairly decent description, albeit machine-translated by Google.


For the test, I picked four batteries with the lowest resistance. I didn't measure their capacity; I just charged them all the same way.


The board is exactly as wide as two 32700 batteries, so it can be used in packs with the batteries arranged either lengthwise or side by side, if you will.


The wiring is very simple: a standard pack of four batteries connected in series. The seller showed a 4S2P pack, but for the test I decided to stick with 4S1P for now.


First I stuck the batteries together with foam double-sided tape, then secured the whole thing with regular tape.


I folded the tabs so the overlap would land on the negative terminal. That way, even if I overheat it while soldering and melt the heat-shrink tubing, nothing dangerous will happen.


The rest is pretty straightforward: I stuck a piece of cardboard on with double-sided tape, then attached the board—with the power wires already soldered on—with more double-sided tape. It's best to solder the wires beforehand so you don't heat up the board after it's mounted on the batteries.
After that, I soldered the balance wires. At this point, the battery assembly is almost complete; all that's missing is an outer layer of heat-shrink tubing, but I won't use one since this is just a test.


The board went straight into operating mode; I didn't have to force it to start by connecting it to a charger. For the initial tests, I used an EBC-A10H electronic load, rated for up to 10 A and 150 W, which is just right for this battery pack.


The preliminary charge immediately revealed a “quirk” here too: by default, the load’s LiFePo4 charging mode is set to stop when the current drops to 50 mA. Since the balancing current here is 100 mA, it would charge forever in that mode. For it to shut off properly, you need to set the current to 100 mA plus the termination current.


After about 20 minutes, the resistors reached around 80 degrees. That seems a bit too hot to me; I think an extra layer of cardboard between the circuit board and the battery wouldn’t hurt.


For a more thorough balancing, I should have let it run for a while, but I didn’t feel like waiting, so I moved on to the capacity test.
Discharging at 10 A, I got 5685 mAh, or 68 Wh, of which 65 Wh came from the 10–14 V range. The protection board, not the load, cut off the discharge.


The total voltage drop across the protection board was about 50 mV at 10 A. The voltage also tended to rise: over roughly 3 minutes of discharge, it went up by 10 mV.


During the tests, I monitored the voltage of each cell. After the discharge cut off, the third cell had the highest voltage and the fourth had the lowest, so I’ll use the fourth cell for a more accurate measurement of the cutoff voltage.


I started a discharge at 5 A while monitoring the voltage. The tester showed that the board cut off at 2.09 V, which is practically in line with the specs.


After that, I fully charged the battery to check the overcharge cutoff voltage.


At first, the fourth cell had the highest voltage, but then I noticed it rose to about 3.71 V and started to drop, falling to 3.70 V shortly afterward. In other words, we can see the balancing process at work: the balancer holds this channel back when its voltage gets too high, while the other cells continue charging at a low current.


But in practice, to equalize the cell voltages, you need to set a fairly low charging current and leave the pack at the termination voltage for a while. Also, if the battery is out of balance, this can trigger the protection.
In my case, the charging current was 4 A, so the pack still ended up unbalanced by the “finish,” with cell voltages 1-2-3-4 as follows.


1. When I stopped charging—that is, when the charging current had dropped to 110 mA and practically all the energy was being dissipated by the balancers—the first cell was at 3.73 V.
2. After that, I triggered the protection by setting the charge termination voltage to 15 V instead of the required 14.6 V. The board cut off at 3.76 V instead of the specified 3.75 V, which is within the stated specs.


Next, I tried to determine the overcurrent protection trip point. To do that, I connected the pack to a load capable of drawing up to 30 A.
Everything went fine for the first 20 seconds or so, but then I heard a small click and the load shut off because the voltage had dropped below the set limit.


It turned out the solder joint had acted as a thermal fuse. The solder melted, and one of the tabs pulled away under its own spring tension.


All right, I resoldered it, adding some more solder in the process, and ran the test again. This time the pack handled 30 A just fine, although the tabs got noticeably hot.
But I figured 30 A was a bit low, so I connected a second load and set it to 10 A. Then I started the first load at 30 A, for a total of about 40 A. After that, I started increasing the current on the second load (the first was already at its maximum).
Before long, the first load shut off again because the voltage had dropped below the set minimum, and smoke started coming from one of the tabs. I had to stop the test. At the time it shut off, the total current was around 50–55 A.


1. After the emergency shutdown and inspection, it turned out the tabs had gotten so hot that they showed temper colors—the metal had darkened where it overheated.
2. A closer look. You can clearly see where the heating was most intense.
3. Interestingly, there’s no sign of this on the other side of the pack, so two of the three connections overheated.
4. The heat was so intense that it melted both the special tape and part of the battery’s heat-shrink tubing.


And this is where I started to get a little concerned, because that kind of heat is definitely not normal. So I decided to check the resistance of the connections and got some interesting results:
1, 2, 3. Resistance across joints 1-2, 2-3, and 3-4. You can see that the 2-3 joint has lower resistance. As I mentioned above, it also ran cooler than the other two, so that’s clearly no coincidence.
4. Total battery resistance, excluding the protection board. The cells themselves are about 7 mΩ, and the connections between cells add another 12 mΩ or so.
5. Resistance of the protection board, excluding the connecting wires—that is, the power transistors plus the traces on the board. In my 10 A test, I got roughly the same result (51 mV / 10 A = 5.1 mΩ).
6. Total battery resistance, including the cells, connections, wires, and protection board.


Interestingly, the product page even had a table explaining what to do if something didn’t work. I saved it just in case.


Later, I left the pack at 100 mA for about an hour and got these results. The results from earlier are on the left, and those after an hour of “rest” are on the right:
3.709 - 3.682
3.614 - 3.636
3.595 - 3.654
3.678 - 3.639

You can see that the voltages are gradually equalizing.

Now for a more illustrative experiment. For this, we’ll use an old UPS.


I actually have two of them, and you may recognize these fairly popular UPS models. The batteries in both died quite a while ago, and the case on one of them even cracked.


If you thought the photo showed two identical models, you’re mistaken: the one on the left is 400 VA, and the one on the right is 600 VA.


The only real difference between them is the battery capacity: the 400 VA model came with a 4 Ah battery, while the 600 VA model had a 7 Ah battery. Yes, there are also a few minor differences in the circuit boards, but for the most part they’re much the same, and the transformers are the same size. I got the less powerful model by chance—someone gave it to me. When I opened it up and compared the two, I realized I could easily fit a 7 Ah battery in it as well. In practice, it works great.


Disconnect the original battery and connect the LiFePO4 pack in its place. Press the magic button, and after the beep and relay click, the UPS switches to battery power.


To test it, I connect a 150 W lamp, although I think it’s actually 125 W. That doesn’t really matter for this experiment. What matters is that everything works. There could still be a problem if the battery discharges far enough to trigger the protection board’s undervoltage cutoff rather than the UPS’s own protection. With a “smart” board, you may have to open up the UPS, since it might not turn back on. A “dumb” board will probably restore power and wait for charging current to appear, but this all depends on the UPS, so you’ll need to test each one individually.

I ran it like this for a little while, since the lamp gets quite hot. The battery jumpers also got noticeably warm because the current was around 15 A.


I connect the UPS to the mains, which starts the charging process. The initial voltage was about 13.27 V; after a while it rose to 13.63 V and stayed there. The charging current was around 250–300 mA at first, but I think that’s because the battery voltage was already close to the charge termination voltage.


So in this UPS model, the charge termination voltage works out to 3.40 V per cell. That’s a bit low, but let’s look at the graph where I compared capacity at different charge termination voltages; I ran this test there as well.
The graph shows that I’ll get about 5700 mAh on discharge. The only downside is that charging will take a very long time, partly because the UPS doesn’t use a four-wire battery connection.


Conclusions.
Overall, I have no complaints about the board. The only thing I couldn’t check was the protection trip current, which is a shame. Everything else works, and the measured values match those given in the description.

I do have an issue with the batteries—or, more precisely, their tabs. Yes, my pack isn’t really designed for currents this high, and ideally I should use at least a 4S2P configuration, or better yet 4S3P. But I honestly didn’t expect the tabs to have such high resistance. In practice, when the battery itself has a resistance of 7 mΩ, the 4–4.5 mΩ resistance of the tabs looks enormous, and it definitely needs to be fixed—either by replacing the tabs or doubling them and soldering on a wire.

But the experiments are still ongoing, and I’m currently waiting for another 10 of these batteries to be shipped, this time from Litokala. I also have a small 2S LiFePO4 board at home that I’d like to test while I’m at it.

That’s all from me. I hope this was useful :)