Warning! This article is about LiFePO4 cells. Some of it may apply to other chemistries, but YMMV.
Edit: I've also written a short How To Bottom Balance. You should still read this too.
I've talked about this before, but I realised I don't have a single, nice post about the matter that I could refer to when needed. Hence this post. All said only applies to current LiFePO4 cells. Embrace yourself.
Bottom Balancing (BB) is an alternative to Battery Management Systems (BMS). BMS is traditionally a system which includes a little device connected to each individual cell. That little device will monitor the voltage of each cell and also alert the master board, charger or controller if the voltage drops below or exceeds preset values. While charging the device can also shunt the cell effectively forcing it to stay at a preset voltage level while other cells catch up. This is called Top Balancing.
Sounds good, doesn't it? I thought so too. A nice device you can buy, maybe a little expensive, but still solves all battery related problems nice and clean. Keeps good care of your cells, controls charging, protects from over discharge and so on.
Turns out Top Balancing and by extension BMS is based on part fantasy and part lead acid heritage. It assumes that you can tell everything just by looking at the current voltage of the cell at any point in time and every state of charge. It also assumes you can get maximum potential out of your cells by boiling them at the highest allowed voltage and even balance them by doing so. Perhaps even refresh them to full capacity.
The truth is far from all these things. The only way to determine state of charge from the cell voltage is to let it sit for a day or two and then measure the voltage down to accuracy of 0.001 volts. Put any load on the cell and you no longer have a realiable reading. Put any charge on the cell and the same thing happens. The cell voltage will sag under load and rise during charge. If you have Top Balanced the cells, which always have different state of charge, will even react by different amount.
What this leads to is that reading cell voltage in these situations is completely useless. If you can't determine your state of charge (SOC) from the reading how could you control your charging or discharging by looking at these values. You just can't.
Even if you assumed that Top Balancing would work and you could even out the differences in cell capacity at the top you would have a problem. The problem would manifest itself when the first cell, the weakest one in your pack, gets near empty. Due to the nature of the LiFePO4 charge and discharge curve, which is very flat for about 90% of the cell charge and very steep at both ends, you end up with that weak cell plummeting in voltage.
If your BMS is fast enough it will disconnect the pack and save the cell. If not, the single cell will keep going down, kill itself and possibly even it's friends nearby by going into reversal or even swelling rapidly. Even worse there's no way to know when this will happen. The exact moment of weakest cell going empty is unpredictable so you end up using such a big safety margin that you're not even using th pack to it's fullest. Add into account the fact that high load will exaggerate the voltage drop and you end up in a situation where you or your BMS can't know what's really going on. Your vehicle either stops sooner than it should have or it breaks down.
Charging is pretty much the same in reverse. You try to fill up each by thinking that if you get all of them to say 3.65 volts they will be as full as they can be and they don't mind waiting for their buddies. It sort of works, because the cell voltage will start to climb when the cell gets fuller, but trying to top them like this achieves nothing and there is a possibility you might also be hurting your cells if you do this. Charging is less critical because it usually happens much slower than discharging. You can charge at say 20 amps, but you may discharge at 1000 amps. Things happen so much faster at 1000 amps.
Because we know that cell state of charge can be measured when the cell has rested, and especially when it's very near empty, we can try something different. Instead of trying to balance the cells at the top by forcing them to a certain voltage we drain them. We drain them down to about 2.75 volts. That's well in the steep downwards part of the discharge curve. Well enough to be a reliable indication of state of charge, but not too far to be dangerous to the cell.
When draining you need to drain the cells a little below 2.75 V, let them bounce back to a little above 2.75 V and then repeat until the cell doesn't bounce anymore, but stays at about 2.75 V even after a period of rest. Once you have done this on all of your cells and removed any load from them you can be quite sure that they are now all at the same state of charge very near empty. Bottom Balanced, that is. We have taken all the cells, independent of their actualy capacity to the same line and we can now charge them. But not one by one.
In order to charge this Bottom Balanced pack it is vitally important that you connect the cells and only the cells, nothing else, to each other. You cannot put anything else on the cell terminals except the strap or a copper bar that connects it to the next cell. If you do put something else you will ruin the Bottom Balancing and you will either have to start from the beginning or not have balanced pack anymore.
Once you have all the cells connected to each other forming a pack, all cells still at 2.75 volts, undisturbed, you can start charging them. But only as a whole. You can never ever charge or discharge only a part of the pack or you must start over again (and do not pass go). Since we know that the manufacturers recommend that most LiFePO4 cells not be charged over 3.65 volts to prevent damage we leave a little headroom and charge to a total voltage which is 3.5-3.55 volts times cell count.
Now you may think your giving up something when you're not charging to 3.65 volts. And you are correct. But it's not nearly as much as you'd think. The charge curve is mostly almost flat and only starts to rise at the very end. And it does rise fast. So fast even that you're putting very little energy into the cells when it starts to rise above 3.4 V. What that means is that the amount of miles you are losing is insignificant. Instead it gives our pack a little room to breathe and a safety margin which is needed, because the weakest cells will rise in voltage first. Just like they would die first if you were Top Balancing.
So how do we know when to end the charge? It's very easy. You either pick a charger that has a maximum charge voltage of 3.55 times cell count or choose you cell count based on a charger you can get. I prefer the latter and use a 116.8 volt charger for 33 cells and a 87.6 volt charger for 25 cells. Thats 3.54 or 3.50 volts per cell, respectively. While charging each cell will receive exactly as much charge as the next. At the end of the charge the weakest cells will rise in voltage a bit more than the best, but nowhere near enough to damage them. When enough cells reach a higher voltage the charger will cut the charge as the total charge voltage has been reached.
The chargers currently used do a CC/CV charge, which means that the charger charges at full amps until a set voltage is reached and the holds that voltage by lowering the current until the current drops to about 5% of full current. The last CV phase accounts for 10-20% of the charge depending on cell type. There is now talk of dropping the CV phase completely with the latest cells and only doing the CC phase, which would only give us a 90% charge, but it would happen fast and might save the cells even more than is generally believed. Thus keeping that 90% range for a longer time instead of dropping to say 80% over time.
This charging that I've described here is just a procedure. It's a way to charge the cells. Once the charge has been cut and you have let the cells rest you will observe that they dropped in voltage to about 3.4 volts each. This is the voltage of a fully charged cell. We simply used a slightly higher average voltage to get them there and did this in series to make sure they all got the same amount of current.
We haven't reached the best part of Bottom Balancing yet. It's so obvious to me I almost forgot to write about it even though it's the main reason why we do it.
The magic of Bottom Balancing happens when you run out of juice. As we drained all cells in the pack to the same voltage and have only charged them together they will also reach empty at the same time. There are two benefits in this. First, since they do it simultaneously and since we are at the steep end of the curve there will be hardly any energy left in them so eventually the vehicle will not move anymore. You start to feel this in you accelerator pedal or twist grip before it happens because you start losing power so you can drive to the side of the road safely instead of the car stopping completely and unexpectedly as would happen with a BMS.
Secondly, and quite importantly, the cells, in harmony, will protect each other. Since none of them have much energy left in them they don't have the power to kill their buddies either. In the BMS based Top Balancing scenario the weakest cell will be violently attacked by the other cells, which still have a lot of power left, by trying to extract as much energy from that weak cell as they are able to produce themselves.
In order to prevent a careless driver from eventually harming the cells by keeping the pedal pressed to the floor even though the vehicle isn't even moving anymore, we program a low voltage limit into the controller. This can be as low as 2 volts times cell count, but you can define it higher if you wish to leave more margin. I have personally tested a 2 volt setting (25 cells, 50 volt limit) on a motorcycle and the cells suffered no harm although I repeatedly hit that 50 volt limit. In fact they bounced back to safe territory some time after the load had been removed.
So there. Didn't come out very short, but that's my view of Bottom Balancing and why it works. YMMV.
ps. The so-called "Cell Drift" has not been observed at all when Bottom Balancing. Therefore it has been concluded that these cells do not drift by themselves. Any observed drift has been caused by the BMS themselves or cell level monitoring which loads the cells unevenly. There is no reason to fight "Cell Drift" by installing equipment which will in fact cause it, unnecessarily wear out the cells and even cause a fire hazard thanks to a mess of spaghetti wiring.
Disclaimer: All of my battery ramblings are based on my own experience and Jack Rickard's original work on the subject. They are applicable to CALB SE- and CA-series cells. Other cells and chemistries may at least require different voltages. I take no responsibility for any problems or damage caused by anyone.