I'm designing a battery pack that's on the order of 10s5p. Each 10s stack has its own BMS, with high side driven FETs for charge and discharge. This is based on the Texas Instruments reference design, TIDA-010030.

I need a pack that uses 5 of these stacks in parallel. Usually, in power supply situations, you'd employ an ideal diode to OR them together. BUT - considering that this is several battery packs, all with their own control FETs, I'm not sure if this is necessary and could use a little light shed on the subject from anyone who's done it before.

The only things pro or con that I could think of were this:

  • Packs that are NOT diode ORed together would risk dumping charge into lesser charged packs
    • These packs are self-charge balanced, so this wouldn't be an issue
  • Packs will shut themselves down if there's a problem
  • Packs will be charged in parallel and discharged in parallel

I can't think of any reason why they couldn't be paralleled without ORing them together. Anyone else have any input on this? Has anyone out there made parallel pack?

Edit: an example of "charge and discharge FETs for low side switch can be found on page 10 of this document from TI. I would be paralleling 5 full 10s BMS stacks. The only difference is that I'm using a high side switch and this example is low side. Same issue though.

The problem I'm facing with the ideal diode system is in relation to charging the packs... I would need an awful lot of circuitry to make that happen.

  • \$\begingroup\$ When you say "all with their own control FETs" do you mean that said control FETs could isolate a pack from the rest, in both charging and discharging? Will you be designing the pack so that it's modular, with each series string independently replaceable? What about the case where you pop a mostly-discharged string into a fully-charged pack? \$\endgroup\$
    – TimWescott
    Nov 3, 2021 at 15:00
  • \$\begingroup\$ @TimWescott - (see edits above) - that's the question I'm asking. In theory, the BMS can be commanded to turn on and off at will, and will do so automatically upon discovery of error. all series packs will be designed into the same board, but technically, could be considered independent, but not modular. The only time you'd have charge imbalance across strings is when there's possible cell damage in the pack at which point the individual BMS chips will shut down the string, leaving the other 4 in use. This includes shutdown of charge cycle. But is this sufficient to string 5 packs in parallel? \$\endgroup\$ Nov 3, 2021 at 16:13
  • 2
    \$\begingroup\$ Because each BMS has its own characteristics, and there's an infinite number of ways to build something to your written description above -- why don't you make a simplified yet sufficient schematic -- maybe just the common rail, ground, the BMS with the part number you intend to use, the FET(s) that you believe will provide the same service as an ideal diode, and the battery string? \$\endgroup\$
    – TimWescott
    Nov 3, 2021 at 22:19
  • \$\begingroup\$ @TimWescott, sorry, not following. Can you rephrase the statement? \$\endgroup\$ Nov 9, 2021 at 14:26
  • 2
    \$\begingroup\$ You stated the pack is 10S5P. Why not create a single battery that uses 10 series strings of 5 parallel cells instead of 5 individual packs with 10 cells in series? You would then only need 1 BMS to manage that stack and you have the same energy. For what it's worth, I've built a few packs like this and have taken them to mass production over the last few years. \$\endgroup\$
    – Tim51
    Nov 12, 2021 at 19:32

3 Answers 3


Answer is neither yes, nor no. And, it depends on how much value you consider to add to your battery pack.

There is no problem, once 5P are at near the same voltage, close enough that the Over-current condition do not occur from the cross current.

Otherwise, over-current protection kicks in, depending on the settings (that is in the BMS), the protection can remain open. You need to detect that condition, or provide balancing of parallel elements.

Overtime, imbalance occurs anyway, and degraded parallel elements do not contribute current to the discharging/charging as much as good ones. The imbalance progresses, do not stop. Eventually, a few good ones take major part of the main current, then eventually gives in. It appears like a sudden failure due to the good few. This condition has to be detected.

While 5P are working independently, there is no information exchange among the 5 individual BMS. That can lead to a real dangerous situation. If one of the parallel element cuts off due to over-temperature, the rest still runs the load/charging until they feel the heat. At the time every BMS cuts off, the reactor goes it's own. You need to detect individual BMS status, and be able to force them shut off. I do not have any information how it can happen, but considered as a possible scenario for large power-banks. Especially when we saw a rental power-bank exploded (not gonna tell what company it was).

Imbalance tells many problems, and causes many problems as well. Meantime, paralleled cells, without individual BMS, conveniently appear balanced but can develop larger issue overtime. The main variable of the trouble is the difference of "impedance growth" between cells.
The condition that developing "impedance growth", can accelerate and get into run-away condition. That was one of the difficult and fearful aspect that BMS designers had to deal with. When Sony batteries were exploding in flame, we suspected ti's BMS too, not controlling/protecting until the internal impedance gets overloaded to the melting point - it turned out to be Sony's production problem -.

I don't think I have answered yes neither no. I have not answered about "ideal diod" either. But, I feel I have given enough background idea, as one of the BMS design engineers once.

  • \$\begingroup\$ You have valid points, and I failed to mention that I have a microcontroller interfacing all BMSs. I presume that ORed packs would have this issue as well, but would just be unable to backfeed/recharge eachother. The problem I have with the 5P10S Idea is that the application in question won't allow for the risk associated with it. If one cell bursts, it could cause a far more rapid chain reaction, even with anti-thermal runaway measures in place. Since I need the 10S voltage, I'm kind of stuck with the 10S5P approach. And I'm most likely going to need 21700 cells for my 1kWh pack. \$\endgroup\$ Nov 17, 2021 at 15:50
  • \$\begingroup\$ Can you add your thoughts on the ideal diode? TI's response was to use them, but then again I've been given misleading info from their "tech support" in the past, which is why I'm reaching out to you kind folk in the working community. \$\endgroup\$ Nov 17, 2021 at 15:55
  • \$\begingroup\$ I would learn about what I did not know, but I wouldn't take as the whole, including jay's. :) I know you knew what I said already. Just you wanted to build some confidence or feeling of consensus. ... But, something happens, then you are alone. H h.. I did not say ideal diode itself addresses everything. Since you have 10S, you just need to combine them in parallel, put together (assembly) when they are balanced, monitor the individual 10S status, control the individual series string. If you do not have access to the control & shutdown, then you need to add something to do that. Problem still \$\endgroup\$
    – jay
    Nov 17, 2021 at 16:45
  • \$\begingroup\$ Problem still remains, if the pack was to be modular, replaceable/hot-swapable 10S modules. \$\endgroup\$
    – jay
    Nov 17, 2021 at 16:48
  • \$\begingroup\$ So it appears that what I'm doing really hasn't been done before, and there are equally concerning risks to the two methods discussed. Thank you for your time, it's really appreciated to have a pseudo discussion about it rather than being dismissed or treated as if it's a "stupid question". You've given me enough info to make my decision - thank you! \$\endgroup\$ Nov 17, 2021 at 17:10

The only reason to parallel stacks is to spread the load between them. But if the stacks are not balanced then low voltage stacks will not share the load. In fact, if you do not have OR-ing diodes then they will increase the load on the other stacks. This alone makes your design decision highly questionable.

Reversing the build to have one serial string of 5 parallel blocks is much better alternative. Of course it has the usual inherent danger of one failing cell shorting out the other four in a block. While you can find several ways of dealing with this (up to going overboard with diodes on each cell), the best solution IMHO would be to look for higher rated cells and using less of them in parallel.

  • \$\begingroup\$ Correct - my cells are individually balanced, see TIDA-010030 cell balancing section. It's passive, and voltage controlled versus computer controlled, but it does work. \$\endgroup\$ Nov 17, 2021 at 16:44
  • \$\begingroup\$ the stacks are Diode ORed, and will share the load. If one stack dips in voltage, the others will compensate, and we're talking minute loads. The problem with the parallel cells in series is that if one cell goes down, they all go down and usually hard. Even if fusible links are used, if a cell goes critical and ruptures, it will take the others with it, even with the use of anti-thermal runaway measures. They protect but not that much. I've figured it out though, thanks! \$\endgroup\$ Feb 14, 2022 at 16:59

I was just thinking the same as @Tim51, that it is better with a 5P10S rather than 10S5P. It seems @Maple is in agreement.

Then you only need 10 cell BMS channels but with 5x higher dissipation ratings*

Since Parallel is self-balancing by default, minimizing the serial packs to balance is ideal to reduce complexity. There may be other tradeoffs.

Albeit, this eliminates the redundancy of adding N packs to dynamically swap packs on the fly with a good one, that is if the design needed hot swap.

The degree of battery mismatch is a high order exponential without a BMS. I would expect the matching of cells within a given batch is like any capacitor with << 1 % tolerance or less on process variance, but batch to batch being a stackup of all the materials and process much higher like 10% for C equivalent and ESR.

For this reason , a larger serial bank from a reputable OEM will always be more balanced initially than your efforts unless you have large quantities, very small bins and allow for yield loss on mismatch. That does not mean you should try, just saying the odds are stacked against you. No Pun Intended.


Fused wired OR ought to be possibly with added protection.


I thought that active bidirectional balancing over 10 years ago would be best and it seems to be a reality today.


There is also a patent

  • \$\begingroup\$ Good point on cell matching. I don't know if battery manufacturers doing that. At their volumes it is probably cheaper to pay damage settlements than check every cell. But for custom build packs there are no excuses. \$\endgroup\$
    – Maple
    Nov 13, 2021 at 20:35
  • \$\begingroup\$ As a Test Engineer , I can't imagine any company not testing to tight tolerances for each cell with Process control limits. FWIW @Maple Loss of performance results from a weak cell in loss of brand loyalty. Not everyone uses a BMS and any mismatch I can prove on Falstad how long it takes to damage the weakest cell assuming 10kF and say 50 mohm ESR with one cell low by 0.1% without a BMS and the difference with 1%. Meanwhile e-caps are typ 10% \$\endgroup\$ Nov 13, 2021 at 20:40
  • \$\begingroup\$ @Maple most companies have a low level of testing the cells at a minimum. I've purchased many cells in the past, and have found that they are fairly well matched prior to assembly. Also, a lot of chargers these days can now re-balance unbalanced cells, but the worst I've ever encountered at purchase was an imbalance of 0.1V or a completely dead cell. I won't even try to recharge a dead cell (<2.8V) - I'll just wrap it and recycle it. \$\endgroup\$ Nov 17, 2021 at 15:53
  • \$\begingroup\$ I hope you are right about companies doing at least basic cell test. But @TonyStewartEE75 and I were talking about cell matching during battery assembly. Even thoroughly tested individual cells do not guarantee the match, as Tony pointed out. This task falls upon battery manufacturer. Specifically in this case upon yourself. And even more specifically if you follow an advice and use parallel blocks of 5 cells (or, as in my answer, less than 5 but with higher current rating) \$\endgroup\$
    – Maple
    Nov 17, 2021 at 16:12
  • \$\begingroup\$ @Maple I will be attempting to match cells, but not at the first iteration - I need to prove concept first. \$\endgroup\$ Nov 17, 2021 at 16:59

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.