Parallel LiPo Battery System

[Kuusou]

I'm seeking feedback on the design of a parallel lithium polymer battery system shown in the attached schematic. The primary goal is to safely combine two individual battery packs to increase system capacity while preventing any negative side effects, such as mutual influence between cells.



Design Details:
Battery Management System (BMS): BRCL3130ME
Datasheet: https://wmsc.lcsc.com/wmsc/upload/file/pdf/v2/lcsc/1811141355_Foshan-Blue-Rocket-Elec-BRCL3130ME_C328560.pdf

Charging IC: PJ4054B
Datasheet: https://wmsc.lcsc.com/wmsc/upload/file/pdf/v2/lcsc/2110272230_PJSEMI-PJ4054B_C2911335.pdf

Low Voltage Cutoff: APX803S-31SA-7
Datasheet: https://wmsc.lcsc.com/wmsc/upload/file/pdf/v2/lcsc/1811141622_Diodes-Incorporated-APX803S-31SA-7_C129757.pdf

P-Channel MOSFET: AS3423B
Datasheet: https://wmsc.lcsc.com/wmsc/upload/file/pdf/v2/lcsc/1912111437_FMS-Formosa-Microsemi-AS3423B_C341542.pdf

Questions:
Do I need to add Schottky Diodes between the APX803S-31SA-7 and each LiPo cell before they are connected in parallel to prevent mutual influence?

Is the current layout needlessly complex? Can the individual battery packs (same manufacturer and rating) be combined in parallel safely without all these extra components? For example, could they be connected together in parallel with each cell only having it's own BMS and use one charging IC? Are my concerns with the batteries influencing each other and equalizing valid?

Any insights or recommendations for improving this design would be greatly appreciated.

[Kuusou]

I've updated the schematic to include the diodes I was asking about above. Can someone please confirm whether this approach I'm taking is valid or not? It would be greatly appreciated!

[wasedadoc]

All that extra instead of using a 2000mAh cell?

[Kuusou]

All that extra instead of using a 2000mAh cell?

Space constrained and I want to keep weight equal.

[Peabody]

If you connect two batteries in parallel, they should have the same voltage to begin with, and then never be disconnected after that.  That prevents large currents from flowing when parallel batteries have different voltages.  I think that also means that the batteries should not have protection built in.  That's because a protection trigger might turn off one, but not the other, and then the voltages may be different when the normal circuit it restored.  Lithium polymer rectangular packs typically do have protection built in.  Do yours?  If they do, then you don't need the protection ICs in your circuit, but I think connecting them in parallel would be problematic.

Your original circuit already has the batteries connected in parallel, and I don't think the separate chargers with outputs connected will work.  They will not necessarily be trying to do the same thing at the same time.

I don't know about adding the diodes.  It's not clear exactly what is accomplished doing that.  At first I thought it would let you use the two chargers properly, but I'm not sure.

I don't understand the low voltage chip.  It seems you turn on its mosfet when battery voltage is very low.  Why would you want to do that?  Wouldn't you want to turn it off?

I can't quite make out the orientation of Q1 and Q2.  Well actually the body diodes.  Are they drawn correctly?  In any case, you might want to consider having the mosfets oriented in opposite directions, so both have to be turned on before any current can flow.  Edit:  I see them clearly in the first post.  Q1 looks right to me.  I think I would flip Q2.

Basically, I think you should have one charger, one protection chip, and parallel unprotected cells.  Then you don't need the diodes or the voltage drop they produce.

[Siwastaja]

Separate chargers -> good! But you need separate load switches, and absolute certainty they never switch on together (unless voltages are qualified first within some +/- dozen of mV). Diodes are the easiest way to achieve this, but lose a lot of power. (Beware of very low Vf schottkys - while datasheet Vf is appealing, already at not-so-high temperatures they reverse leak so much they enter positive feedback and self-destroy.)

Reconsider if you REALLY need all this. Permanently connecting the two cells in parallel, manufacture-time, would reduce the number of chager ICs, protection ICs, and load switches to just one each. The space and cost savings would probably get you that another cell, and you'd still get manufacture-time configurability of supplying either one or two cells.

[Kuusou]

If you connect two batteries in parallel, they should have the same voltage to begin with, and then never be disconnected after that.  That prevents large currents from flowing when parallel batteries have different voltages.  I think that also means that the batteries should not have protection built in.  That's because a protection trigger might turn off one, but not the other, and then the voltages may be different when the normal circuit it restored.  Lithium polymer rectangular packs typically do have protection built in.  Do yours?  If they do, then you don't need the protection ICs in your circuit, but I think connecting them in parallel would be problematic.

I've read that if you try to connect two packs together, even if they're the same manufacturer and capacity that it can cause them to fail:
"When you put LiPos with mismatched discharge curves in parallel, any number of things can happen. The cells can self-limit to keep their output voltages more or less in sync, or they can go into any number of conditions that lead to immediate failure or degradation of the polymer over time. The usual mode of degradation is dendritic (tree-like) growth of metallic lithium on the anode and cathode, creating low-resistance paths that will eventually meet and short the whole cell internally.

That's what happened with the Samsung Galaxy 7: between the battery architecture and load pattern on the cell, there was a pattern that could lead to an internal short. The actual frequency of that happening was way below 0.1% of devices actually in the field, but there was no way to know how many more devices would fail over time. Scrapping the whole product line and destroying billions of dollars worth of existing phones ended up being the most cost-effective way to deal with the problem safely. Kudos to them for doing it."

So my goal was to connect the pack output only and prevent them from influencing each other. I wanted them to be charged individually and have their own bms to make it as safe as possible. That's why I wanted to get a confirmation on whether this statement is overblown or not.

Your original circuit already has the batteries connected in parallel, and I don't think the separate chargers with outputs connected will work.  They will not necessarily be trying to do the same thing at the same time.

I thought I had it setup where the packs were being charged as individual packs but I was still getting their combined output because the diodes would prevent current traveling from one battery to another.

I don't understand the low voltage chip.  It seems you turn on its mosfet when battery voltage is very low.  Why would you want to do that?  Wouldn't you want to turn it off?

The lAPX803S-31SA-7 reset pin is normally low, when it's triggered it turns high cutting off battery power to the rest of the system. The goal in this is because I have a buck converter that is set to run at 3.0V and I wanted the batteries to be cutoff at an amount slightly higher than this to prevent any issues.

Basically, I think you should have one charger, one protection chip, and parallel unprotected cells.  Then you don't need the diodes or the voltage drop they produce.

This would be preferable, I'm just trying to be as safe as possible with the batteries. If my concerns are moot then I will definitely take that approach. Thank you for your feedback.

[Kuusou]

Separate chargers -> good! But you need separate load switches, and absolute certainty they never switch on together (unless voltages are qualified first within some +/- dozen of mV). Diodes are the easiest way to achieve this, but lose a lot of power. (Beware of very low Vf schottkys - while datasheet Vf is appealing, already at not-so-high temperatures they reverse leak so much they enter positive feedback and self-destroy.)

Reconsider if you REALLY need all this. Permanently connecting the two cells in parallel, manufacture-time, would reduce the number of chager ICs, protection ICs, and load switches to just one each. The space and cost savings would probably get you that another cell, and you'd still get manufacture-time configurability of supplying either one or two cells.

Thank you for taking the time to take a look and provide feedback, it's very much appreciated!

I read elsewhere on the Arduino forum that connecting the individual packs together in parallel can have some seriously bad consequences and so the entire purpose of this circuit was to have my cake and eat it to.
"When you put LiPos with mismatched discharge curves in parallel, any number of things can happen. The cells can self-limit to keep their output voltages more or less in sync, or they can go into any number of conditions that lead to immediate failure or degradation of the polymer over time. The usual mode of degradation is dendritic (tree-like) growth of metallic lithium on the anode and cathode, creating low-resistance paths that will eventually meet and short the whole cell internally.

That's what happened with the Samsung Galaxy 7: between the battery architecture and load pattern on the cell, there was a pattern that could lead to an internal short. The actual frequency of that happening was way below 0.1% of devices actually in the field, but there was no way to know how many more devices would fail over time. Scrapping the whole product line and destroying billions of dollars worth of existing phones ended up being the most cost-effective way to deal with the problem safely. Kudos to them for doing it."

Is the statement above overblown? When you mention connecting them together at manufacture-time are you referring to the battery packs being matched at the factory or when I'm building my system? Because I'd much prefer to do away with all the extra components and just connect them in parallel, I just want to be certain that it's safe.

[Peabody]

The datasheet I found for the lAPX803S-31SA-7 said its output is active low.

I've never done anything other than 18650 cells in parallel, and have had no problems.  I just don't know about the issues you uncovered.

I don't think the Galaxy 7 problem involved parallel cells.  My understanding is it was simply the size of the cells that was the issue.

In any case, your original circuit, without the diodes, had the positive cell terminals shorted together, and being charged by two separate chargers.  That doesn't work.  Perhaps the diodes makes that work, but at the cost of the diode voltage drops.

[Kuusou]

The datasheet I found for the lAPX803S-31SA-7 said its output is active low.

It looks like I read the datasheet wrong, thank you for catching this!

[Kuusou]

Ok how does this design look? I've removed all the extra components, the Batt+ & Batt- labels are for the second pattern that connects in parallel to the first from another pcb. Q2 has been replaced with an N-Channel MOSFET. When the RESET pin on the APX8035-31SA-7 is triggered it'll go low, closing the gate and cutting off the batteries from ground and the rest of the system. I've included a pull down resistor to make sure it goes to ground. Will this layout work properly?

[Siwastaja]

"When you put LiPos with mismatched discharge curves in parallel, any number of things can happen. The cells can self-limit to keep their output voltages more or less in sync, or they can go into any number of conditions that lead to immediate failure or degradation of the polymer over time. The usual mode of degradation is dendritic (tree-like) growth of metallic lithium on the anode and cathode, creating low-resistance paths that will eventually meet and short the whole cell internally.

That's what happened with the Samsung Galaxy 7: between the battery architecture and load pattern on the cell, there was a pattern that could lead to an internal short. The actual frequency of that happening was way below 0.1% of devices actually in the field, but there was no way to know how many more devices would fail over time. Scrapping the whole product line and destroying billions of dollars worth of existing phones ended up being the most cost-effective way to deal with the problem safely. Kudos to them for doing it."

Just FYI, this is total pseudo-information word salad bullshit, maybe written by AI, or AI-like human waste. Ignore completely.

[Siwastaja]

And note the difference of paralleling battery packs (of many series cells) vs. just cells. If you parallel battery packs, so that internal cell taps are not paralleled, then every paralleled string need their own management (cell level voltage measurements, load switches). OTOH, if you parallel all the cells first, then make the series pack, each cell combo acts like one bigger cell, so only one management system is needed for the whole.

BTW, even though you are using 1-cell "packs", it's a similar case: your desire for user-connectable parallel packs brings you the requirement to duplicate all management for each pack. This management includes limiting charge and discharge currents. And in this specific case, preventing large currents that would happen by paralleling two packs even with slightly different SoC. If you react reliably and super quickly, you can first connect, then disconnect based on overcurrent detection, but that seems pretty iffy to me; better first qualify voltages and only allow parallel connection once they match. And this requires steering current first to your loads from the fuller pack.

A lot of effort. You avoid all this by permanent manufacturing-time cell paralleling.

[Kuusou]

And note the difference of paralleling battery packs (of many series cells) vs. just cells. If you parallel battery packs, so that internal cell taps are not paralleled, then every paralleled string need their own management (cell level voltage measurements, load switches). OTOH, if you parallel all the cells first, then make the series pack, each cell combo acts like one bigger cell, so only one management system is needed for the whole.

BTW, even though you are using 1-cell "packs", it's a similar case: your desire for user-connectable parallel packs brings you the requirement to duplicate all management for each pack. This management includes limiting charge and discharge currents. And in this specific case, preventing large currents that would happen by paralleling two packs even with slightly different SoC. If you react reliably and super quickly, you can first connect, then disconnect based on overcurrent detection, but that seems pretty iffy to me; better first qualify voltages and only allow parallel connection once they match. And this requires steering current first to your loads from the fuller pack.

A lot of effort. You avoid all this by permanent manufacturing-time cell paralleling.

I'm a little confused by your statement. Are you operating under the assumption that these packs will be connected and disconnected many times? Because the user won't be able to connect or disconnect the battery packs. They will be connected once when I'm building it and that's it.

From your previous responses it sounds like my concern with the pack influencing each other was a moot point but are you saying that I still need all the extra circuity, such as a BMS for each pack and a charging IC as well???

If I get the battery packs made without a BMS and instead use the one in the design above, will connecting them in parallel like I've done work if they're never disconnected again? The goal of the circuit is to allow the user to plug in a 5V source and have it charge the batteries while powering the system. If the 5v source is plugged in it's suppose to cutoff the batteries from the load. If the batteries get below 3.08V it's suppose to cut them off from the load also.

[Siwastaja]

I'm a little confused by your statement. Are you operating under the assumption that these packs will be connected and disconnected many times? Because the user won't be able to connect or disconnect the battery packs. They will be connected once when I'm building it and that's it.

OK, I missed that, because it changes everything. Just parallel bare cells. Double-check that voltages are within say +/- 50mV or so (they are from the factory, if you are buying decent cells), then connect the cells in parallel. This way you don't need to duplicate protection circuitry etc. And in fact, it's better that you don't. Unnecessary protection circuits can fail. They can disconnect cells for any reason, and then reconnect them when the cells are in different voltages, causing massive currents, which, hopefully, trigger overcurrent protection of the modules. But such events are, even if not dangerous when everything goes right, risky, and definitely unwanted.

With permanently paralleled bare cells, such large currents surges caused by connection of different voltage cells simply can never happen, and everybody's happy. Of course, an internal failure on one cell escalates to the whole combination, but it's no different of a larger cell developing internal failure.

If you absolutely have to parallel complete packs that already have some protection, then the devil is in the details; how that protection is exactly designed to work. It's a can of worms. In best case, it works fine. In worst case, you have a problem nest, possibly dangerous. Two separate chargers plus load ORring through series diodes is the safest bet, but the power loss of the diodes is awkward.

[Peabody]

I think the output of the new schematic is messed up.  You've added a 5V output which you don't need, D11 is in the wrong place, and the P-channel mosfet is oriented wrong.  The circuit below shows what I think it should be.

The N-channel mosfet circuit doesn't make sense.  When turned on, it would ground the battery output, which of course you don't want to do.  I think you'll need to put the second P-channel back in, and control its gate with the N-channel.  Or maybe it would be better to control the gate of the single P-channel from either VBUS or the reset IC.  I would need to think about that.

Edit:  Can I persuade you to use the APX810S reset IC?  It is push/pull, and its output is active high.  So it should be able to drive the second P-channel mosfet directly, just like the original schematic you posted.

I'll ask again about the batteries.  Do they have built-in protection?

[Kuusou]

I think the output of the new schematic is messed up.  You've added a 5V output which you don't need, D11 is in the wrong place, and the P-channel mosfet is oriented wrong.  The circuit below shows what I think it should be.

The N-channel mosfet circuit doesn't make sense.  When turned on, it would ground the battery output, which of course you don't want to do.  I think you'll need to put the second P-channel back in, and control its gate with the N-channel.  Or maybe it would be better to control the gate of the single P-channel from either VBUS or the reset IC.  I would need to think about that.

Edit:  Can I persuade you to use the APX810S reset IC?  It is push/pull, and its output is active high.  So it should be able to drive the second P-channel mosfet directly, just like the original schematic you posted.

I'll ask again about the batteries.  Do they have built-in protection?

Thank you for your feedback and taking the time to provide suggestions! So the reason for the 5V line is because that is the power from the USB-C connector, the idea is to use that to power the system and skip the boost converter elsewhere in my system. Otherwise Vbatt supplies power to the boost converter and everything else. Are you sure about the P-Channel MOSFET being in the wrong direction? From my reading the current is suppose to flow from source to drain, which is from pin 2 to pin 3, whereas an N-Channel is the opposite, with current flowing from drain to source.

I've been reading up on P&N Channel fets and was looking at this solution:

The idea is that while the reset pin is high, it turns on the N-Channel FET so that the Gate of the P-Channel is connected to ground and pulled low, thus turning on the P-Channel Fet and allowing current to flow. When the voltage drops to 3.08V the reset pin goes low and turns off the N-channel FET, which causes the pull up resistor to drive the gate of the P-Channel high thus closing the gate and cutting off the battery. I'm not sure if this will work like I think though, for example is the connection between the drain and source on the N-Channel going to be enough to drive the P-Channel gate low while it has the pull up resistor connected?

Another approach I was considering was using an inverter like the SN74LVC1G04DCKR https://wmsc.lcsc.com/wmsc/upload/file/pdf/v2/lcsc/2405141142_UMW-Youtai-Semiconductor-Co---Ltd--SN74LVC1G04DCKR_C2983744.pdf on the reset pin. It's a few cents cheaper per part vs the N-Channel FET I'm looking at so that's a plus. I also think if I went that route I wouldn't need a pull up resistor for the P-Channel anymore would I?

The APX810S series looks perfect but the issue is I'm using JLCPCB to build my boards so I'm limited on what's in stock, unless I want to buy an entire reel. The APX810S-31SR-7 looks like what I'd need so I will definitely keep it in mind, thank you for the suggestion!

I'm going to order the batteries without a BMS per your previous suggestion.

[Peabody]

The 5V and Vbatt nets make sense if they are connected together downstream at the input of the boost converter.  Then it's equivalent to your original circuit.  Then the 4.7V after D11 will be boosted back to 5V if VBUS is present.  Edit: If you try to connect the 4.7V to the circuit *after* the boost converter, then you may get current flowing backward through the boost - its behavior in that situation is undefined.

Current flows equally well in either direction through a mosfet.  So the decision here is which way you want the body diode to point.  In this circuit, you want it to prevent current from the 5V source from flowing back to the battery, which could damage the battery.  So it should point the same way as D11 - to block that.  Here's  a Microchip app note that has examples of this type of circuit, and you'll see that the drain is toward the battery.

http://ww1.microchip.com/downloads/en/AppNotes/01149c.pdf

Actually, I would put the two mosfets in opposite orientations, with the drains connected together.

Q2 is going to need a pullup resistor too.  The /RESET output of the voltage monitor chip is open collector, so without a pullup resistor, the gate of Q2 would be floating when the voltage is ok.  However, that pullup and R5 can be much higher values, like 100K or even higher.

I think the LVC1G04 would work, but its input would need a pullup resistor for the same reason Q2 would need one.  I don't know about eliminating the pullup on Q3.  I think you probably could do that because as the battery voltage drops, the protection circuit would shut down the battery before the LVC would stop operating.  So Q3's gate would never float unless there is no battery current at all, and then it wouldn't matter.

[Kuusou]

The 5V and Vbatt nets make sense if they are connected together downstream at the input of the boost converter.  Then it's equivalent to your original circuit.  Then the 4.7V after D11 will be boosted back to 5V if VBUS is present.  Edit: If you try to connect the 4.7V to the circuit *after* the boost converter, then you may get current flowing backward through the boost - its behavior in that situation is undefined.

Current flows equally well in either direction through a mosfet.  So the decision here is which way you want the body diode to point.  In this circuit, you want it to prevent current from the 5V source from flowing back to the battery, which could damage the battery.  So it should point the same way as D11 - to block that.  Here's  a Microchip app note that has examples of this type of circuit, and you'll see that the drain is toward the battery.

http://ww1.microchip.com/downloads/en/AppNotes/01149c.pdf

Actually, I would put the two mosfets in opposite orientations, with the drains connected together.

Q2 is going to need a pullup resistor too.  The /RESET output of the voltage monitor chip is open collector, so without a pullup resistor, the gate of Q2 would be floating when the voltage is ok.  However, that pullup and R5 can be much higher values, like 100K or even higher.

I think the LVC1G04 would work, but its input would need a pullup resistor for the same reason Q2 would need one.  I don't know about eliminating the pullup on Q3.  I think you probably could do that because as the battery voltage drops, the protection circuit would shut down the battery before the LVC would stop operating.  So Q3's gate would never float unless there is no battery current at all, and then it wouldn't matter.

The reason I've broken out the 5V from Vbatt is because I have multiple voltages in my design. The boost converter I'm using is an MT3608 and it has a diode on the output so I don't see why combining the 5V after the converter would cause any issues. I've also changed the R15 value from 10k to 100k.

I've decided to ditch the N-Channel and use the APX810S-31SR-7. Apparently getting items restocked at JLCPCB isn't as hard as I had thought. I've connected it directly to the P-Channel FET and I've flipped them as you've suggested. I also added a pull-up resistor on the reset pin to ensure it's pulled high at cutoff. Thank you for linking that white paper by the way, it was a great read.

[Peabody]

In the boost converter, there is probably also a connection from the output diode's cathode to the converter chip.  If you apply 5V there when the input to the converter is floating, will any current flow back through the chip to ground?  Can you test that ahead of time?  The datasheet is unlikely to help.  I agree that it will not flow back through the diode, but will it flow back through the chip?

So are  you saying you have part of the circuit that is powered by either 5V or the battery, and another part that's powered by 5V if present, but otherwise shut down?

The APX810S is supposed to have a push/pull output.  So in theory it doesn't need a pullup resistor. The datasheet says "the outputs are guaranteed to be in the correct logic state for VCC down to 1V", and your protection circuit will have long since shut down the battery at something like 2.5V, so the logic should always be right at any possible Vcc above zero volts.

[Kuusou]

In the boost converter, there is probably also a connection from the output diode's cathode to the converter chip.  If you apply 5V there when the input to the converter is floating, will any current flow back through the chip to ground?  Can you test that ahead of time?  The datasheet is unlikely to help.  I agree that it will not flow back through the diode, but will it flow back through the chip?

So are  you saying you have part of the circuit that is powered by either 5V or the battery, and another part that's powered by 5V if present, but otherwise shut down?

The APX810S is supposed to have a push/pull output.  So in theory it doesn't need a pullup resistor. The datasheet says "the outputs are guaranteed to be in the correct logic state for VCC down to 1V", and your protection circuit will have long since shut down the battery at something like 2.5V, so the logic should always be right at any possible Vcc above zero volts.

I'm going to have to test it, which is the plan. I'll be adding lots of test points for every component before finalizing the design. If it does go through the chip and to ground in testing then I'll just add a second diode and adjust the converter to account for the voltage drop from it.

I have 1 part powered by 5V and the MCU powered by the Vbatt and a buck/boost converter and LDO combo. LDO powers the chip during deep sleep and then switches to the buck/boost converter during normal operations.

I didn't catch that, I'll remove the R5 resistor, one less component is always welcome!

[Peabody]

Switching between a buck/boost converter and an LDO sounds pretty slick.  If it's not a State secret, I'd like to see how you did that.