Charge & Discharge Lithium Cells in Series

How do you charge and discharge lithium cells connected in series? It's not easy. If you choose to design the circuitry to monitor each cell's voltage, charge current, and discharge current, the circuitry would be quite complex and difficult to optimize for best performance.

However, with an integrated solution, the needed functionality is all on one analog IC, so the overall design becomes much easier — and you can focus on added value elsewhere.

So, why is it difficult to charge and discharge these cells? At first glance, the question may seem simplistic. The answer seems to be contained in the question. You connect them in series and charge them or use them to power the load. And that is absolutely all there is to the problem — as long as you make sure that the cells are absolutely identical, perfectly matched in ampere-hour rating with absolutely identical internal chemistry, composition, and construction.

Why would that matter? Because if the cells are not identical in charge capacity, as you draw energy from them, first one and then another will become completely discharged. Besides the obvious effect of reduced terminal voltage from the battery pack, any cell that is discharged will become reverse biased from the voltage of the remaining cells and the connected load. You can verify this by sketching a simple schematic of cells in series plus a load resister. Assume the voltage across one cell drops to zero and that all that remains is some internal resistance. You will note a polarity reversal at that cell. Pushing current the wrong way through a discharged cell will cause rapid deterioration.

In consideration of the possibility that you can't find perfect cells, the best thing to do is to use cells from the same production lot. The next technique is to switch an additional load onto the stronger cells to bleed them down to match the energy level at the weakest cell when you detect a mismatch in terminal voltage. That will minimize damage, as explained above, but at the cost of making your battery pack as bad as its weakest cell.

There is a better way to do this. Rather than throwing the excessive energy away, using it to boost the weaker cells is the smart thing to do. If you had a special bidirectional circuit that could pull energy from the better cells and supply it to the weaker cells, that would be the ideal solution.

Linear Technology has introduced just such a device in a complete integrated package. The LTC3300-1 monitors up to six cells and has a unique integrated controller that drives a bidirectional flyback switcher with external FETs that draws from the strong cells and charges the weaker cells. For more than six cells, you just stack up another LTC3300 IC.

By controlling the timing of the on-board low-side FETs, one side of the transformer becomes the primary and one side the secondary. Energy can be transferred from right to left or left to right, so any cell can be the current source or sink.

The IC would be useful if you're designing electric vehicles or backup battery systems (e.g., a UPS for a server). The IC has a SPI interface, so it can communicate with your system and report battery status. It's spec'd at a maximum temperature of 125°C, so it's clearly intended for automotive applications. Note that the IC is intended for use only with Li-Ion or LiFePO4 batteries, not NiCd or nickel-metal-hydride.

22 comments on “Charge & Discharge Lithium Cells in Series

  1. eafpres
    March 11, 2013

    Hi Brad–I remember reading about this problem when people started proposing to use large numbers of small capacity, off the shelf cells to power EVs or hybrids.  This seems like a good solution.  If I understand the schematic (not a great assumption!) you need 1 IC per 2 cells in the series group?

    For some designs I've heard of this would mean a lot of these ICs per vehicle.  Having worked in automotive electronics in the past, these babies better be cheap.  Any thoughts on that or am I off in my interpretation?

  2. SunitaT
    March 12, 2013

    For more than six cells, you just stack up another LTC3300 IC

    Brad, thanks for the post.  Whether stacking up LTC3300 ICs allow communication between the batteries connected to one IC with the batteries of other IC? Whether the communication control between them will be in the same pattern as it is within batteries of one IC?

  3. SunitaT
    March 12, 2013

    If I understand the schematic (not a great assumption!) you need 1 IC per 2 cells in the series group?

    @eafpres, if you look at the schematic you can see battery symbols from CELL1 to CELL6 so my assumption is that we can connect 6 cells per 1 IC.

  4. eafpres
    March 12, 2013

    Yes–I see what they are illustrating now. Would be nice to include a pin out of the IC. So my concern is divided by six but I have heard of proposed set ups with very high cell counts.

  5. Brad Albing
    March 12, 2013

    Clearly, if you have a lot of cells (typical in an HEV or EV) you would nee a bunch of these ICs – no getting around it. Maybe that's good, tho'. The voltage the IC is sunjected to is just the span of those 6 cells. The communications interface to the next group up or down would of course be twice that, but the IC's interface cktry is designed w/ that in mind. A very large [span] IC would need to tolerate 100s or even a thousand Volts. Possible, of course, but expensive.

  6. Brad Albing
    March 12, 2013

    Wih respect to any one IC, it can communicate w/ the IC above it and the IC below it. So information can be passed along to subsequent or supersequent ICs in a daisy-chain arrangement.

  7. eafpres
    March 12, 2013

    @Brad–in large systems, especially if there was a likelihood of multiple cells draining down, I wonder if a more complex network topology would be more efficient, as in a mesh setup like you can do in a wireless network.  The issue with the setup of these ICs is in a long series, what happens if one IC fails?  In a mesh topology you could work around that.

    March 12, 2013

    Looks similar to what they add to PV array strings. To help a panel that gets partially shaded, a circuit for that panel would boost the voltage to keep the PV string up in voltage and functional. This appears to boost the dead cell, or, at least, work around it without damaging the cell. Now to just get the system to tell you which cell to replace instead of the whole battery array.

  9. Brad Albing
    March 12, 2013

    I believe that in UPS batteries, there is an archetecture that if not mesh is at least suitably redundant with standard hot-swap controllers. So in effect there is a series-parallel arrangement. Multiple series stacks charged and monitored w/ devices like the part I described; then the series stacks are in parallel.

    So if there were a charge-balancing IC failure, that stack could be isolated and swpped out for a good one while repairs are made.

  10. Brad Albing
    March 12, 2013

    Pretty sure the system can be set up to report which cell is dead via the I2C bus that can send data back to whatever uP or uC is running the system.

  11. MarvA
    March 13, 2013

    The first concern is not dealing with a dead cell, but keeping the charging (and therefore discharging) balanced between the cells.  This keeps from creating dead cells.

  12. WKetel
    March 13, 2013

    @Brad, I really wonder why it would be worthwhile to know which cell is weak in a computer power pack. If you open one it becomes obvious that they are not intended to be repaired at all. Totally welded shut cases and all of the cells welded togather with no extra lead length. What I see is a real effort to assure a market for only one source of replacement battery pack, and every possible effort to assure that no other pack will function in the computer.

    Battery packs in any type of electrical vehicle will probably be similar, although there may be some attempts to repair them. But I don't believe that any repairs other than module swapping will ever be done at a dealers location. So if your EV battery fails, give the man $10K for a replacement, plus turning in the failed battery pack.

  13. MarvA
    March 13, 2013

    Again, I say that the philosophy is to balance charging so you don't tend to get a damaged cell with multiple cycles.  Probably a smarter circuit is to come up with a scheme to balance both charging and discharging.

    When you look at the discharging of a battery pack as a single voltage, you can't know that one cell might be at a lower voltage than the others and in a position of being abused when discharging continues.

  14. paulsky
    March 13, 2013

    Why not just wire the calls in parallel?

    No problems with weak cells pulling the pack down, Charge redistribution works automatically.

    To get the higher pack terminal voltage use a switch mode boost circuit. You would still likely have a less complicated and costly solution than what's being proposed here.


  15. Brad Albing
    March 13, 2013

    I suppose you could do that. Of course, you'd need a boost switcher whose input current might be 100s or 1000s of Amps. Then you'd need power FETs, inductors, and busbars w/ comparable current ratings. And the FETs would need extremely tiny on resistance. Even small amounts of resistance in interconnections would throw away a bunch of power and cause efficiciency to plummet. And the capacitors at the input to the switcher would need huge ripple current ratings.

    But otherwise I suppose it would work.

  16. Brad Albing
    March 14, 2013

    It's tougher to balance the discharge  of the cells – you'd need a pass transistor (probably a FET) in series which each cell and suitable level shifting cktry to drive the gate (or a galvanic isolator ckt to drive the gate). And the pass transistor which need to be sized gig enough (drain current rating) for the load current. Likely 100s of Amps. Technically feasible but economically not so much.

  17. WKetel
    March 14, 2013

    OK, knowing what cells are doing in order to provide balancing makes sense. Yes, balancing the energy available from each cell in a string is quite important, and that is one reason that cell manufacturing is a challenge, trying to produce all of the cells identical is not a simple task. And the closer you get them the harder the task becomes. That could be the “Less Law”, running hard against the “Moores Law”.

  18. eafpres
    March 15, 2013

    Hi Brad–your design sounds pretty good.  Suggest you crank out a schematic, get somebody to simulate some failure modes etc., and bingo you can publish in the SAE Journal.

    Seriously, though, do you think that EV use series setups to keep the current reasonable given the power they need? 

  19. Brad Albing
    March 15, 2013

    I'm sure they use series stacks to keep the overall current draw tolerable. Easier to deal w/ a few 100V than 1k to 10kA.

    And no SAE paper in the works. I'll leave that to the guys at Maxim.

  20. Kufman01
    March 20, 2013

    Parallel cells are not very useful for anything over a few amps.  When you get into systems that need hundreds of amps at a couple hundred volts, a boost system is impractical and extremely inefficient. 

  21. jnissen
    March 20, 2013

    But in an application that needs the higher voltages this is impractical. EV vehicles are such an application. To date the focus has been on charging to ensure matching then just let the discharge be fairly uncontrolled and in series. The weakest cell will obvioulsy sufferthe most and will likely be the first to fail. The charge circuits will rebalance it and it's a constant back and forth. For the time being the trade offs are better than putting in bypass switches around cells. The current carrying capacity of the switches would have to be enomorous and highly expensive for a large pack. Not easy to solve.

  22. StephanWeber
    March 21, 2013

    I think all the nmos give a lot of additional loss . A good method to avoid them completely would be to leave all the cells fix in series, and to have one aux cell which you can just connect to the weakest one in the big set, so that it gets stronger. After some time of course another cell can become the weakest, then just switch only that aux cell!! Much less effort!!

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.