I was alarmed by the fact that after just 2 years, all three of these packs had appeared to degrade badly: Of the three, two of them had seen their original 6+ amp-hour capacity diminished to 2 amp hours or less (the third was at about half capacity) and was decreasing every time I checked them.
In response to this post I had several comments/suggestions (see the comments at the end of the original post linked above) about what might explain the problems that I was having, including:
- Temperature variations. Very cold temperatures will dramatically reduce the available cell capacity, but these batteries were all tested at similar temperatures.
- Perhaps I'd charged the cells to too-high a voltage and damaged them? I'd checked and found that the internal circuits prevented the cells from "seeing" voltages in excess to those noted in the manufacturer's recommendations.
- Self discharge? Perhaps the protection circuits inside the battery were causing the packs to bleed down? Maybe the cells themselves were losing charge on their own?
More recently I finally got the time and inclination to disassemble one of my LiFePO4 packs and do a bit of looking.
First, I charged the pack by connecting it to a 14.6 volt, current-limited (to 500 mA) source. After the current dropped to nearly zero (overnight) I disconnected it from the charger and let it sit for a week or so to let the cells "settle out" and attain a steady-state voltage. Upon disassembly to the point that I could measure the voltage across each cell I found that they were fairly close - within 0.1-0.15 volts or so - not radically different, but still a bit more than I would have expected.
I then put the battery pack on my West Mountain Radio CBA-II battery tester and did a capacity test which subjected it to a constant 2 amp load while graphing and integrating the results. As expected, this pack, which had measured at about 6 amp-hours when it was new, now measured less than 2 amp-hours. Upon measuring the voltages across the cells in this discharged state the precise nature of the problem became clearer: Several of the cells were in the 2.8-3.1 volt area but one cell was at about 2.4 volts and it had been lower when it was under the load of the tester.
Did I have a bad cell?
I charged the pack again and when complete - but still while on the charging voltage - I checked the cell voltages and saw that at least one of them was at about 3.9 volts - just at the voltage where the protection circuit would disconnect the charging current for the entire string indicated that charging would no longer proceed. I then set an adjustable voltage supply to precisely 3.65 volts and connected it across each cell, in turn, for several hours (until the current dropped off) to assure that every cell was fully charged.
Interestingly, the cell that I had thought might be bad (the one at about 2.4 volts) took a charge - and then kept taking a charge for several hours before the current tapered off. In contrast, the cell that had been at the highest voltage immediately reached full charge as indicated by the current dropping off quite soon after I'd applied it. I was now ready to re-test the entire battery to see if this was the cause of the problem.
Cell imbalance and how it can cause apparent loss of capacity:
This behavior implied a problem that was different from what I'd expected: Gross cell imbalance.
One vexing problem with any system of series-connected cells (e.g. a battery) is that since not all cells are exactly the same, one will inevitably discharge sooner than the others. On "primary" batteries, this isn't too much of a problem since being non-rechargeable, a few percent of difference between cells won't have too much effect over the pack as a whole. On a "secondary" battery - the types that can be recharged - this can be an issue, especially if the problem accumulates over a number of charges.
Perhaps the most notable of deleterious side-effect of cell imbalance is that experienced in NiCd packs where the premature discharge of one cell results in its effectively being charged backwards and a chemical process occurs in which a dendrite (metallic hair) grows, reducing its capacity and in severe cases, shorts the cell and/or causes it to self-discharge very quickly. In this case, this cell - already weakened - will be the first to discharge the next time it is used and this will happen again, causing even more degradation!
In the case of a Lithium pack, this unintentional "reverse charge" doesn't occur as there are (supposed to be) circuits that prevent any of the cells in the pack from being discharged too deeply by disconnecting the entire battery from the load - something that (in Lithium chemistry) can cause permanent damage. This same circuit also does something else: If, while charging, the voltage on one or more cells exceeds a particular threshold (typically 4.2 volts for most conventional Lithium-Ion cells) the entire battery is, again, disconnected.
For the LiFePO4 battery, the chemistry is a bit different. While the "low voltage disconnect" of these cells is comparable to that of some conventional Li-Ion chemistries (2.0 volts) the high-voltage disconnect on the particular LiFePO4 cells that these packs contain is 3.95 volts maximum.
Note: The nominal "full charge" voltage for these particular LiFePO4 cells is normally 3.65 volts. As is typical, this voltage immediately decreases when the charge voltage is disconnected.
What seems to have happened was this:
- When the battery pack runs down: One cell reaches 2.0 volts, the internal circuit disconnects the load. Other cells have probably not reached 2.0 volts.
- When the battery pack is charged. One cell reaches 3.95 volts, the internal circuit disconnects the charging current. Inevitably, there will be one or more other cells that did not reach this voltage and were thus not as fully-charged.
- When the battery is discharged the next time. The cell(s) that didn't get fully charged the last time will likely be the one(s) that discharge(s) first: When that happens, the load is disconnected, but the other cells will still have charge on them.
- When the battery is charged again. Those cells that had the most charge will reach the maximum voltage first, leaving the more-discharged cells with a less-then-full charge.
- As this process repeats, the disparity between those cells that are fully charged and those that were not will increase. Because the overall capacity of the pack is dictated by the weakest cell - in this case, the one that has the least amount of charge - the apparent capacity of the battery pack will gradually decrease.
A LiFePO4 battery pack when nearly new (black) and after 2 years (red). There
is very clearly a loss of available capacity! (2 amp discharge rate.)
Click on the graph for a larger version.
It should be noted that the load test current in the above graph was 2 amps and this is much higher than the "20 hour rate" at which battery manufacturers typically rate batteries. Because of this, the tested capacity was somewhat lower than the pack's official 6+ amp-hour rating.In each case you can see that near the end of the discharge curve, the battery voltage starts to droop - and then is abruptly cut off by the protection circuit when one or more cells drops below 2.0 volts. What I should have noticed was that the overall voltage before the curve started to get steep was actually lower when the pack was "good" than it was when its capacity had diminished. What this indicated was that when the battery pack was newer and performing better (black trace) all of the cells' voltage were dropping more nearly in unison while on the red trace, we don't see that gradual drop to nearly the same extent implying that, perhaps, just one cell was dropping out sooner.
Equalizing the cells - the result:
As mentioned above, I individually charged the cells using a regulated, current-limited power supply that was set to 3.65 volts and then individually tested each cell's amp-hour capacity. I was pleased to note that every cell seemed to be fairly close to its rated specifications and, as I now suspected, the cells that had been noted to have discharged first when the pack as a whole was tested were, in fact, those that had slightly lower capacity than the others by 5% or so.
As described abive, I then recharged the entire pack and then topped each cell off, individually, to 3.65 volts, and then re-tested the entire pack again:
Figure 2 shows the result of that test. As can be seen, the capacity of the battery pack after having equalized the cells is slightly better than it was when it was brand new!
Clearly, there was a problem to be solved: How do I keep this - and other - LiFePO4 battery packs "equalized" and still have them be practical and safe to use. Here are the points to consider:
- I did not want to remove the battery's built-in protection circuit as it takes very few cycles of gross over/under voltage before one permanently wrecks Lithium-Ion cells. While LiFePO4 cells are considered to be (more or less) "intrinsically safe" when compared with conventional Lithium-Ion cells, that just means that they are less like to catch fire/explode if mistreated, not that they won't degrade quickly!
- A practical method of charging the battery pack and maintaining equalibrium of the cells within. What this meant was that I was not going to disassemble the pack every time I wanted to charge it so that I could get at each cell!
- Charging/equalizing should be relatively simple and convenient. This goes with the above and it would imply a simple box that I could connect to the battery while charging, when I wanted to equalize the cells.
Instead, I decided to modify each pack, bringing out connections from each individual cell so that the voltages could be monitored and controlled during "normal" charging of the pack. To do this, I added a short cable with a 5-pin DIN connector which provided the points needed to which I could connect an external box with which the cells could be equalized when charging. What this also meant was that I needed to construct only one external device to monitor and control the voltage on three LiFePO4 batteries that I had!
A charge monitoring and control circuit:
Please note that the diagram in Figure 3 represents what would be required for just a 2-cell (nominally "6 volt") LiFePO4 battery: One would simply replicate the circuits as needed for as many cells as the battery pack contained, connecting them in series as shown in the diagram. For the discussion below, we'll generally refer to the parts in the circuit connected to Cell 1 which are numbered with parts designations in the "100's".
How it works:
As noted above, wires are connected to the cells inside the battery pack and brought out through a cable that is separate from the main power connector. In my case, I used a 5-pin DIN connector as it had precisely the number of connections needed: Battery - (minus) and the + (positive) terminals of each of the four individual cells. You will note that in the diagram there are devices marked F101 and F201 which are 1 amp, self-resetting thermal fuses located at the point where the wires connect to the cells (or protection circuit) inside the pack. These are extremely important as an accidental short-circuit of one of the wires emerging from the pack could result in burns or a fire without its presence as the cells will easily supply enough current to burn the wires upen! Adding these devices is a cheap way to make the modification safe and should a short occur, simply fixing the problem will cause the fuse to automatically reset when it cools down - more on these devices later.
The heart of the circuit is U101, a TL431 shunt regulator - a part that may be found in nearly every PC power supply. This device is essentially a programmable Zener diode: If the "adjust" terminal is more than 2.495 volts (nominal) above the anode terminal, it will conduct via the cathode. The values of R103 and R104 are chosen so that if more than 3.65 volts appears between the "top" of R103 and the bottom of R104 (which is also connected to the anode of U101) more than 2.495 volts will appear on its adjust terminal and the device will turn on.
When U101 turns on, it will start to "pull down" on the base of Q101, a PNP power transistor, which will also conduct, amplifying the effects of U101 and start drawing current through R101 (R102 keeps Q101 turned off otherwise.) Were it not for the presence of R101, a 0.25 ohm resistor, the "knee" of this conduction would be extremely steep, but this resistor gives the circuit a bit of elasticity (a bit of range over which the current will go from nearly zero to maximum) and also assure that the circuit will not oscillate. With the parts in the diagram, a current of about 200 milliamps through the shunt regulator - plus the finite gain of U101 - will result in a clamping voltage on the cell of around 3.75 volts, not including the effects of the resistance of the wire connecting to the cell within the battery.
As a convenient indicator that the U101 circuit is actively shunting current from the cell, Q102 detects when U101 is "pulling down" on the base of Q101. When this happens, Q102 is turned on and LED101 is illuminated: Q102 and associated components (LED101, R105, R106) are optional if you don't need an indication of a particular cell reaching full charge.
Finally, there is also a simple circuit consisting of R1 and LED1 that is across the entire battery. The sole purpose of this is that of an indicator to show that this circuit is connected to something and serve as a reminder that it should be disconnected when charging is complete!
Comments on parts and construction:
By cascading multiple copies of this circuit, one can apply them to as many cells as needed whether it is just a pair of cells or many more. Because my LiFePO4 packs were "12 volt" types with 4 cells in series, I cascaded 4 of the above circuits together, building it on a small piece of perforated prototype board as shown in Figure 4.
The type of prototyping board that I used has a small copper ring around each hole making construction much easier and more durable.
I do not currently have plans to produce a circuit board for this project, but let me know if you do as that might be of interest to others.
A 4-cell Cell equalizer circuit built on perforated
Click on the image for a larger version.
The TIP32A PNP power transistors are of the TO-220 type (with metal tabs) and were chosen simply because I had a lot of them onhand, but practically any PNP power transistor could have been used - but I would not recommend Darlington types due to possible voltage limitations. As can be seen in Figure 4, pieces of copper, bent to fit within the box and to increase surface area - were soldered to the tabs of these transistors to improve power dissipation.
With the heat sinks shown and with the box put together, the unit can easily withstand 200 milliamps of shunt current on each of the four sections without causing excessive thermal stress of the components inside from the 3-4 watts heat generated. If more shunt current capacity is desired, larger heatsinks and/or better ventilation - or perhaps connecting the transistor's tabs to a metal box (making sure that their heat sinks were electrically insulated) - would be recommended along with the decrease of the value of R101. Practically speaking, by proper choice of the transistor, the value of R101 and heat sinking one could design this sort of circuit to accommodate practically any amount of current!
In the lid of the box I drilled some holes and used cyanoacrylate ("Super") glue to to hold the LEDs in place temporarily (overnight) and then used clear epoxy to more-permanently secure them into position afterwards. Because this shunt regulator was intended for 4 cells, there is a red LED for each of the shunt regulators to indicate its "full charge" status plus a fifth, bright blue LED as an indicator to show that the unit is, in fact, connected to a battery - which serves as a reminder to disconnect it when charging is complete.
An external view of the cell equalizer unit built to acommodate
up to 4 cells. There are separate indicators that show which
cell(s) are at full charge - plus a "power" indicator.
Click on the image for a larger version.
Figure 5 gives a view of the completed shunt regulator showing the short cable and the male DIN connector that is connected to the mating connector on the battery. As the label indicates, the five LEDs on the front show the full-charge status and whether or not it is connected to power. It is worth noting that when they were installed, the LEDs weren't fully pushed to protrude through the front panel - this to reduce the likelihood that they might accidentally be broken loose and pushed back into the box during handling.
- In theory, one could parallel a fairly low-value resistance across each cell - say, a 22 ohm, 1 watt device - and achieve equalization. This value of resistance would equate to about 165 milliamps of current at 3.65 volts. Eventually - after, perhaps 36-48 hours for a pack of this size - the cells would gradually equalize their charges: The lower the resistance, the more quickly equalization will occur. If this is done, resistors with the appropriate power rating should be used and they should all be of equal value and, of course, they should be disconnected immediately after charge has been completed, just before the charging voltage is disconnected
- For different types of chemistry one would choose the values of R103/R104 as appropriate (see the data sheet for the TL431 for more information.) Note, however, that this circuit will simply not work for voltages much less than 3.2 volts! If, for example, one wanted to regulate 2 volt lead-acid cells, this circuit would not work, directly, but multiple regulators could be placed across pairs of cells in an interleaved manner and achieve the same effect.
- Once constructed, the clamping voltage of the individual cells' circuits should be checked at the same current, say 200 mA and if necessary, adjust the value of R103 or R104 to make all of them equal: All cells being within 50 millivolts (0.05 volts) should be more than adequate. It's worth noting that normal variations in the resistor values (particularly if 5% tolerance parts are used) as well as variations among TL431's can cause voltages to differ slightly.
Modifying the battery packs:
One of the more critical aspects of this projects was safely modifying the battery pack. As noted above, the wires connecting to the outboard charge equalizer have current limiting on them in the form of (approximately) 1 amp, self-resetting thermal ("PPTC") fuses that are located inside the battery's case.
For a Wikipedia article on these devices, click on this link.
The fuses are necessary safety components to prevent burns/fire should the added connection(s) be accidentally shorted out!Some builders may not be familiar with these self-resetting fuses so a few words of explanation are in order. Typically resembling small, yellow disk-ceramic capacitors, these devices are really specially-designed thermistors: If the current through them exceeds the design threshold, they will get hot (but not too hot!) and their resistance will skyrocket. When in this "tripped" state, enough current will flow through them (typically 5%-15% of the rated trip current) to keep them warm and in the "tripped" condition. Once the fault has been removed, they will cool down and reset themselves when their internal resistance returns to normal.
These devices are readily available from electronics suppliers like Digi-Key, Mouser and Jameco (to name but a few) and aren't all that much more expensive than normal "one-time" fuses. Available in current ratings from below 100 mA to 10's of amps, the 1-amp rating noted is a suggestion: The devices that I used were actually rated for 1.1 amps of "holding" current, but anything in the range of 0.6-1.5 amps would have been just fine in this application as that current range will keep the wire from burning up in the event of a short, yet be enough to handle the expected shunt current.
When "blown" these self-resetting fuses do not completely stop the current flow, but reduce it to a "safe" level. What this means is that should the wire be shorted out indefinitely, the affected cell(s) will eventually be discharged, but you aren't likely to set anything on fire - and this is particularly important since our connection to the cells must bypass the protection circuits within the battery pack!
The other pack (Figure 7) was constructed in the traditional way: Series-parallel connected cells (four pairs in series in a "4S2P" configuration) to get the desired voltage and capacity and as such, the modification was much easier to accomplish.
It is worth noting that on the battery packs, female DIN connectors are used to prevent accidental contact with the voltages that are always present on these pins.
Using the Shunt Regulator/Equalizer:
The three modified LiFePO4 packs, the extra wire/connector
being clearly visible.
Click on the image for a larger version.
Through experimentation, I determined the precise voltage at which all four shunt circuits would conduct at 150-200 milliamps. When the current from the power supply settled down this amount, I knew that the voltages on the cells are equalized and that the battery could be removed from the charging voltage.
Using it for the first time:
The first time I used the device on a battery pack that was known to be badly out of equalibrium for which I had NOT already charged each cell individually, it took 24-36 hours for all of the cells to come to equalibrium. What happened was that once the "most charged" cell achieved its target voltage, the shunt regulator took over and the other cells were charged by the "excess" current being passed around the fully charged cell(s). Because this current was only on the order of 100-200 milliamps, it took quite a while for the other cells - some of which were at only 20-25% charge - to be fully "topped off" at this rather low current.
One does not need to use this device every time a battery is charged!
It takes several cycles (and time!) for the cells within the battery to "drift" apart and fall out of equilibrium so one could probably get away with using it only every 6-12 charges (or 3-6 months, whichever is sooner) and still maintain good charge capacity on the battery pack.
Once the charging has been completed, the battery should be disconnected from the power supply used to charge it and the equalizer circuit should be disconnected. This equalizer circuit, when the voltage(s) on the cell(s) are below the clamping threshold, doesn't draw much current - somewhere in the 3-5 milliamp range - so it is not likely to discharge a pack of this size very quickly. If this were left on a battery connected across a small solar panel, for example, its contribution to the discharge of the battery would be negligible.
Final analysis of the problem:
One of the ways to most-quickly "kill" any Lithium-Ion based cell is to maintain it at a full charge voltage all of the time: Best longevity is generally obtained if it is kept it in a relative state of discharge (30%-60%) and charging it fully only just before you need it. Keeping it in this state reduces the rate of inevitable degradation that occurs over time. With this in mind, I generally keep my Lithium-Ion power packs in that state and have had good luck over the past 15 years or so that I have been applying that principal, typically getting at least 5 years out of the pack before its capacity drops below about 60% of its design.
So was the intent when I got these LiFePO4 packs: I would top them off occasionally - usually just before using them - and when I was done, I'd typically leave them in their partially-charged state. Apparently, this strategy won't work. What to do, then, other than either leave them on a float voltage all of the time or build a circuit like this?
As noted, I'm reluctant to put these batteries on a charger, set to 14.6 volts (3.65 volts/cell) and just walk away. In theory, if I did do that, they would eventually reach equilibrium as the fully-charged cells, having gotten above its maximum voltage threshold, discharged just below it and allowed charge current to be applied across all the cells in the pack, if even occasionally, for a brief moment. Again, doing this will inevitably reduce the overall lifetime of the battery - but perhaps this is an acceptable tradeoff for you.
What the "protection" circuits current found in batteries should really have is a more intelligent charge management capability in that if one cell was at a higher potential than one or more of the others, current would be shunted around it and allowing the other cells to equalize. Alas, that's not the way they work, hence this need/purpose of this project!
Why don't other LiIon battery types experience this "inequality" problem more readily? Part of this has to do with the fact that many of them (e.g. batteries in laptops) are typically "stored" in the device and kept at full-charge voltage all of the time and will eventually equalize.
Lead-acid (e.g. car) along with NiCd and NiMH cells are less-often subject to this problem (inequality related to charge/discharge rather than cell degradation) mostly because they can be subjected to a degree of abuse (overcharging) without worrying too much about damage or, and this allows all of the cells to equalize. (Note: There are equalization techniques applied to various types of batteries - namely lead-acid - but those will not be covered here.)
How do the batteries test now, after more than 2 years of ownership, before equalization, new and after equalization? The graph in Figure 9 tells the story:
As you can see in Figure 9 there was significant degradation of capacity as the cells drifted apart from each other, rapidly decaying in just a few months (5/12 to 9/12). Packs #1 and #2 are identical, originally consisting of a pair of LiFePO4 packs connected in parallel at the outputs of the protection board (one of these was depicted in Figure 6) while the "Metal" pack had just one protection board with the cells connected in series-parallel format. All three of these packs were used approximately equally, but it's interesting to note that the metal-cased pack seems to have degraded far less: Was this due to differences in usage, or because of the way the cells were configured? It could be that with more cells in parallel, the rate-of-change of their respective states of charge caused a slower rate of degradation.
It's also interesting to note that in the case of the metal pack and #2, the capacity after equalization was actually higher than it was when new, indicating that when it was "new", the cells were already drifting apart!
So, what do I think about LiFePO4 cells now?
I'm more than willing to concede that, perhaps, there is something to this longevity of LiFePO4 cells after all. As can be seen above, once the packs were equalized the cells seem none the worse for wear after several years have passed along with a degree of abuse.
What about this (apparent) propensity of the cells to drift apart? It might be interesting to have another pack that is left "on charge" for long periods of time to see if it will equalize itself and/or if it degrades at an accelerated rate. It could also be that newer protection circuits and cells are more likely to maintain a closer charge state than these (now) older packs. Perhaps other brands of batteries would be less susceptible to this problem.
Would I get more LiFePO4 batteries in the future? Yes, now that I know that I should be on the lookout for a problem like this!