LFP mobile House bank - xSyP serial vs parallel design thoughts

john61ct

1 TW
Joined
Dec 18, 2018
Messages
8,758
Say you want a nominal 12V bank.

You could just wire a single string of four cells aka 4S, each at the same AH capacity you want for the bank as a whole.

However, there's zero redundancy there, if one cell goes bad, you no longer have 12V.

Next option, 4S2P is pretty standard, first wiring 4x 3.2V cells in series to get to nominal 12V,

then paralleling two of those strings in order to increase AH capacity.

This gives you easy redundancy. Each half of the bank is still 12V, so if one cell dies, you just split the bank,

see which half has the bad cell by checking for lower voltage, and carry on with the better half, which is now a 12V bank at half capacity.

If you wire using 2P4S instead, first wiring 2x 3.2V cells in parallel to get to nominal 12V,

then paralleling two of those strings in order to increase AH capacity,

you'd no longer be at 12V if one cell dies, you'd have more work to break down the bank, maybe far from home, to be able to identify the bad cell(s) and then reassemble to a single 4S bank.

For even greater redundancy, 4S3P or 3P4S means if one cell fails, you still have 2/3 of your AH capacity at 12V. The former has the above advantage over the latter.

However, getting up to 4+ strings, it is generally recommended to go "parallel first", with say 6P4S rather than 4S6P. Apparently lots of **individual cells** in parallel is not a big problem to get to higher total bank AH capacity,

but going past three **strings** in parallel can lead to balancing issues. I believe that is true, even with a proper balanced wiring scheme using buss bars in place, as per "Method 3" here: http://www.smartgauge.co.uk/batt_con.html

In theory of course this issue can be accommodated simply by choosing larger AH capacity cells as your basis to start with, but for the sake of this discussion let's assume there are good reasons for the owner using a greater number of smaller cells.

Another easy solution would be hardware that allowed for an independent per-cell charging regime, so high-accuracy top-balancing with the xSyP bank still assembled, was as easy as just pushing a button. But again, for the sake of this discussion, let's assume that's just a pipe dream.

So finally, first question for the hive mind:

Do **you** think going "parallel first" with 6P4S is better than 4S6P? Why, what factors (other than convenient redundancy as discussed above) are involved?

If a specific context is required, let's posit aluminum-cased prismatic cells, each 20-25AH, with a target of 120-150AH total bank capacity at 12V nominal, so still relatively easy to handle at around 45-50lbs.

Q2: What would be your configuration of choice, using the same total 24-count of those same-sized cells, to go to a 24V bank at half the AH?

Q3: A nominal 48V bank (now getting into propulsion territory), 30-38AH total?
 
john61ct said:
This gives you easy redundancy. Each half of the bank is still 12V, so if one cell dies, you just split the bank
IF you remove that string once the cell dies. But you have to stay on top of it. Letting that string sit at a bizarre VPC while the system is trying to cycle it can be bad news. So in that case make sure you have independent BMSes to catch the failure.
you'd no longer be at 12V if one cell dies, you'd have more work to break down the bank, maybe far from home, to be able to identify the bad cell(s) and then reassemble to a single 4S bank.
You'd still be at 12V, you'd just see an odd VPC distribution (i.e. the parallel string with one bad cell will read lower.)
but going past three **strings** in parallel can lead to balancing issues. I believe that is true, even with a proper balanced wiring scheme using buss bars in place, as per "Method 3" here:
You're applying lead acid rules to LiFePO4 here. Since LiFePO4 does not have an inherent balancing mechanism like lead acid does, the same rules do not apply.
Q3: A nominal 48V bank (now getting into propulsion territory), 30-38AH total?
For home power I would go with 48V. Home power means loads of at LEAST 1000 watts, and often around 3000 watts. For that you really need 48V.
 
My main question was the first one, and afaict was not answered in your reply.

I'd prefer to keep the topic of this thread on the serial vs parallel ordering question, not exactly "just theory" but without getting side tracked into extraneous issues.

I believe that topic is not dependent on "the big picture" use case, and my use cases certainly have nothing to do with home power, as in a S&B dwelling.

I just meant "House bank" as in "general auxiliary" loads, or "not critical to travel", not Starter engine cranking, nor dedicated to specific uses like winching.

I will be building banks with all three of those voltages for various scenarios, mostly in a mobile context, but with a wide range of daily loads, including some with well under a kWh per day, often less than half that. Also at higher voltages for future high-amp propulsion needs.

And yes, systems will notify of LVD / HVD imbalance and temp issues, or isolate the bank if needed.

Back to the topic:

billvon said:
john61ct said:
but going past three **strings** in parallel can lead to balancing issues. I believe that is true, even with a proper balanced wiring scheme using buss bars in place, as per "Method 3" here:
You're applying lead acid rules to LiFePO4 here. Since LiFePO4 does not have an inherent balancing mechanism like lead acid does, the same rules do not apply.
Please clarify what you mean by "inherent balancing mechanism" and which chemistry does / does not do it, using which wiring scheme?

Do you mean (guessing here) that LFP units at different SoC do balance out over time, but only when connected in serial, but

packs connected in parallel, if starting out at different voltages,

those differences get maintained as the whole-bank gets cycled, until the bank gets broken out and explicitly re-balanced?

But that lead batts on the other hand, do tend to balance their SoC state, between units connected in parallel as well as in series?




 
john61ct said:
billvon said:
john61ct said:
but going past three **strings** in parallel can lead to balancing issues. I believe that is true, even with a proper balanced wiring scheme using buss bars in place, as per "Method 3" here:
You're applying lead acid rules to LiFePO4 here. Since LiFePO4 does not have an inherent balancing mechanism like lead acid does, the same rules do not apply.
Please clarify what you mean by "inherent balancing mechanism" and which chemistry does / does not do it, using which wiring scheme?
Lead acid batteries balance themselves during overcharge. The cells being overcharged lose a little electrolyte while staying at 100% charge, and the lower charged batteries 'catch up.' (That's why you will occasionally see reference to an 'equalizing charge' which forces this to happen rapidly.)

This works well in a single series string, because if you set charge a voltage that guarantees a little overcharge (i.e. equalize) then all the cells see the same current and get fully charged. But if you have, say, four cell strings in parallel, then the "strong" (low ESR) string sees most of the current, and the cells that are getting a little worn (i.e. higher ESR) see less current. The cells that are getting a little worn, though, are the cells that need the balancing. So after a while charge current gets grossly out of balance - one string does most of the work while the other string(s) see less charge current and never get quite balanced, which ages them more rapidly.

You can never overcharge a LiFePO4 or Li-ion cell without damaging it, so the above does not apply. Forcing all cells to be 3.6 volts (in the case of LiFePO4) ensures they are fully charged; the current isn't as important. So a (say) 4S4P array of LiFePO4 is less problematic than a similar array for lead acid.

However, that also means that there's a requirement to balance, or at least monitor, the LiFePO4's. Which usually needs to be done with an additional device.
 
billvon said:
Lead acid batteries balance themselves during overcharge. The cells being overcharged lose a little electrolyte while staying at 100% charge, and the lower charged batteries 'catch up.'
So I assume you mean a series'd unit, say 6V, rather than separate 2V cells like Rolls Surrette, which could be put in parallel before serial.

> But if you have, say, four cell strings in parallel, then the "strong" (low ESR) string sees most of the current, and the cells that are getting a little worn (i.e. higher ESR) see less current. The cells that are getting a little worn, though, are the cells that need the balancing. So after a while charge current gets grossly out of balance - one string does most of the work while the other string(s) see less charge current and never get quite balanced, which ages them more rapidly.

Yes. Are you saying that works differently for lead than it does for LFP?

> You can never overcharge a LiFePO4 or Li-ion cell without damaging it, so the above does not apply. Forcing all cells to be 3.6 volts (in the case of LiFePO4) ensures they are fully charged; the current isn't as important. So a (say) 4S4P array of LiFePO4 is less problematic than a similar array for lead acid.

Let's take both the charging profile issue and protective infrastructure out of the picture for the purpose of my main question, IOW assume all is well on both fronts.

In that context, are you saying that a LFP 6P4S setup is not preferable to 4S6P?

Remaining Qs:
john61ct said:
Why, what factors (other than convenient redundancy as discussed above) are involved?

Q2: What would be your configuration of choice, using the same total 24-count of those same-sized cells, to go to a 24V bank at half the AH?

Q3: A nominal 48V bank?



 
Almost always the preferred layout is parallel at the cell level first to build capacity, then put these guys in series to achieve your desired voltage (4s for 12 V nominal).

Sure, if one cell is a dud, it will drag all cells in parallel down with it, but that's a fairly uncommon thing. If you instead put two 4s1p strings in parallel, and a cell started to fail, the next string will dump current into it in a big way - 4 times higher current than had it happened in a lower voltage 1sYp arrangement.
 
Another issue with series first then parallel is that if you have nothing to disconnect strings from each other automatically upon any cell imbalance, and if you are not monitoring all individual cell voltages, is that if a cell begins to become imbalanced, either high or low, the other cells in the string will be forced into imbalance as well. If the cell gets low, say, 2.8v, while the rest are at 3.3v, then that's up to half a volt to be spread among those three cells. As the imbalance becomes worse, you may end up severely overcharging one or more cells at the top end, and severely overdischarging, potentially even reversing, the low cells.

All this can happen while all the other strings are perfectly balanced.

The more cells in series, the less the imbalance will affect each individual cell, if there is only one or two that are imbalanced. The fewer cells, the more it will affect them all.

If all the cells are paralleled first, this kind of problem will take longer to happen, and it is much easier to monitor to prevent it (and generally cheaper to keep balanced if using an automated system).

The main thing that will help minimize this is to ensure all the cells you use are well-matched, so that they all have the same capacity and same internal resistance at all the states of charge, etc. Just having cells from the same batch will help, but doesn't guarantee anything--you'd have to test each one to ensure they are all the same, and they ahve to be tested under teh conditions they'll be used at.
 
Related post on wiring up the BC168 per-cell balancing charger, or perhaps a more recent better version of the same type?

https://endless-sphere.com/forums/viewtopic.php?f=31&t=41107&p=1439025#p1439025
 
john61ct said:
So I assume you mean a series'd unit, say 6V, rather than separate 2V cells like Rolls Surrette, which could be put in parallel before serial.
Yes, a large string in series, not parallel.
Are you saying that works differently for lead than it does for LFP?
Yes. LFP don't need to be fully charged to last a long time. Lead acids do. Lead acids are OK with overcharge; LFP's are not.

In that context, are you saying that a LFP 6P4S setup is not preferable to 4S6P?
If you can afford to build 6 individual batteries (i.e. 6 BMSes, 6 disconnects) then 6 strings will give you more flexibility and better protection. If not, go with a single BMS on a string of paralleled cells.

Q3: A nominal 48V bank?
For a 48V LiFePO4 I'd either use two strings of 16 each with separate BMSes or one string of paralleled (2P) cells.
 
billvon said:
Lead acid batteries balance themselves during overcharge
Sorry, I want to clarify this "chemistry vs self-balancing" issue independently of SoC / voltage / charging.

Two sets of battery cells

#1. 6x lead cells
vs
#2. 4x LFP cells

within each set, each cell resting at slightly different voltages / SoC, small enough any "self balancing flow" current rate issues are minor.

Two separate intra-pack wiring configurations:

A. the set connected in series, both sets at ~12V

B. in parallel, 1B @2V, 2B @3.2V nominal.

Sitting isolated overnight, which of the four sets (if any) will approach getting to balance on their own?

 
jonescg said:
Almost always the preferred layout is parallel at the cell level first to build capacity, then put these guys in series to achieve your desired voltage (4s for 12 V nominal).

Sure, if one cell is a dud, it will drag all cells in parallel down with it, but that's a fairly uncommon thing. If you instead put two 4s strings in parallel, and a cell started to fail, the next string will dump current into it in a big way - 4 times higher current than had it happened in a lower voltage 1sYp arrangement.

OK, my paraphrase, more general than the original context:

Units in a string are **much** more vulnerable to a single unit failing.

You would need to separate the "bad string" from the paralleled "still good" strings ASAP, ideally automatically via BMS-type functionality.

correct?

_____
Now a comment / question:

At a low voltage like 4S, variations in voltage would be easy to spot at the string-unit level.
 
parallel than series Lifepo4 is great. This way you can ez monitor. Why not LTO cheap last a long time. The new stuff. Look up the threads here on E.S. Buy quality very little failure rate. quality. You will be using a bms.
 
amberwolf said:
The main thing that will help minimize this is to ensure all the cells you use are well-matched, so that they all have the same capacity and same internal resistance at all the states of charge, etc.
Yes, in fact the use of routinely using per-cell charging as with BC168 seems attractive, as well as of course checking for balance at bottom of the cycle before charging. https://endless-sphere.com/forums/viewtopic.php?f=31&t=41107&p=1439025#p1439025

Weak cells (and sub-bank packs) should thus reveal themselves and get pulled out, usually long before any catastrophic failures.

> If all the cells are paralleled first, this kind of problem will take longer to happen, and it is much easier to monitor to prevent it (and generally cheaper to keep balanced if using an automated system)

OK, so say the lowest unit is 6x 25Ah cells paralleled into 150Ah blocks @3.2V.

One cell failing does not hold as much immediate danger to its blockmates, but doesn't the fact that that failure

only brings down that block's AH capacity, but doesn't have much impact on voltage,

make it **harder** to immediately detect at the block level than with the string approach?

In other words, the parallel approach relies **more** on per-cell monitoring in order to detect and fix the problem.

And if we're talking about a single 12V bank, rather than just pulling a 12V string out and carrying on, I'd need to break down all three of the paralleled "good" blocks and remove a cell from each, in order to return the bank to a balanced state?
 
Outlined in OP, in general.

Otherwise not germane to the specific topic at hand, want to avoid derails please.
 
john61ct said:
Two separate intra-pack wiring configurations:

A. the set connected in series, both sets at ~12V

B. in parallel, 1B @2V, 2B @3.2V nominal.

Sitting isolated overnight, which of the four sets (if any) will approach getting to balance on their own?
None.

LFP will not self-balance, period.

If in "B." you mean that the individual cells are all completely disconnected from all of their series connections, and then reconnected so all teh cells are now in parallel, that will balance them, but I seriously doubt you'd want to do that every time you charge it up (or ever, really).

Building an automated way to do that is going to be potentially larger than your battery, depending on the current the battery must supply, because the relays/contactors you need to do it get very very large, and it would require at least two for every cell, probably more, depending on how many poles and contacts each relay has. And if any of the relays fails to switch, you have a serious risk of fire due to shorting across cells, etc.

Flooded Lead Acid (FLA) will only self balance while charge current is flowing. It does so by boiling off electrolyte, which is why you have to keep topping off your car battery water.

The cells within a 6-cell FLA (or SLA,etc) can't be hooked up to other paralleled FLA to balance because there are no external connection points for them. So you can't do version B with those.

SLA (sealed lead acid) can't be topped off, so when you let it self-balance, it slowly destroys it. But if you don't let it self-balance, eventually it is unusable becuase some cells don't charge enough to prevent sulfation, and it's destroyed that way.
 
john61ct said:
At a low voltage like 4S, variations in voltage would be easy to spot at the string-unit level.
No, because the strings are all in parallel, so all the *string* voltages are exactly the same.

If a string has a cell that drops in voltage then current pours out of the *other* strings *into* this string, and then the cell resistances cause the voltage of the string total to maintain the same voltage as the others.

Over time the low cell gets lower, the high cells get higher, and things go bad from there to worse to potentially fires.


The ONLY way you can tell by voltage what is going on is to separately monitor every single cell in a pack made of series strings with no cell-level parallel connections.

Way harder to do than just keeping them all in parallel.


The only chemistry you'd ever do this kind of thing with is NiCd or NiMH, because you can't parallel those to charge without serious risk of fire, and unhooking the series strings to charge is a lot easier than unhooking every single cell's parallel connections.
 
john61ct said:
amberwolf said:
The main thing that will help minimize this is to ensure all the cells you use are well-matched, so that they all have the same capacity and same internal resistance at all the states of charge, etc.
Yes, in fact the use of routinely using per-cell charging as with BC168 seems attractive, as well as of course checking for balance at bottom of the cycle before charging. https://endless-sphere.com/forums/viewtopic.php?f=31&t=41107&p=1439025#p1439025

Thatis completely different from making sure the cells you use are well-matched, so that they all have the same capacity and same internal resistance at all the states of charge, etc.

What you're talking about is a band-aid used because the cells are *not* well-matched, or not good quality. Very common problem, but completely avoidable.

Bottom-balancing isn't a good way to do it, unless you're literally going to use the cells down to zero percent of their capacity. If you stay within the middle 80% of the capacity of a pack, it's less likely to become imbalanced anyway, assuming you start with well-matched quality cells.

Also, something like the BC168 could take days, or even weeks, to balance a pack the size you're wanting, if there was any kind of serious imbalance. None of these little RC chargers have more than a few dozen mA of balancing capability, just like most of the common ebike-pack BMS units. They're all meant to keep balanced packs that are already well-matched, so that any variations are very small to nearly nonexistent.


Weak cells (and sub-bank packs) should thus reveal themselves and get pulled out, usually long before any catastrophic failures.
While this is good to do as the pack ages, you should not need to do this while it is new and healthy, if you start with well-matched quality cells.


If you're not using well-matched quality cells, you may see all sorts of problems, including cells that have such different internal resistances that they are balanced while sitting there, but whenever under load or charging they are wildly imbalanced.


OK, so say the lowest unit is 6x 25Ah cells paralleled into 150Ah blocks @3.2V.

One cell failing does not hold as much immediate danger to its blockmates, but doesn't the fact that that failure only brings down that block's AH capacity, but doesn't have much impact on voltage, make it **harder** to immediately detect at the block level than with the string approach?

In other words, the parallel approach relies **more** on per-cell monitoring in order to detect and fix the problem.
No, if you are using an automated system (common BMS), as it keeps them balanced at the top of charge, where variations in voltage make it easy to see what state of charge it's in.

If you are trying to balance them down in the middle of their SoC curve, where voltage variation is minimal for any particular point in capacity, then sure, it'd be harder--but that's not how it's done, simply because it's harder. ;)


If you have well-matched quality cells then until the pack ages significantly and cells begin to drift from their specifications, then it is not going to be a problem. By the time the pack ages taht much, it's going to be increasing in resistance and decreasing in capacity anyway, so you will notice that and can then test the cells to weed out bad ones to give the pack a little more life.




And if we're talking about a single 12V bank, rather than just pulling a 12V string out and carrying on, I'd need to break down all three of the paralleled "good" blocks and remove a cell from each, in order to return the bank to a balanced state?

No, because the automated balancer (any common BMS) would keep the entire battery balanced *for you*.
 
john61ct said:
Two sets of battery cells

#1. 6x lead cells vs #2. 4x LFP cells

within each set, each cell resting at slightly different voltages / SoC, small enough any "self balancing flow" current rate issues are minor.

Two separate intra-pack wiring configurations:

A. the set connected in series, both sets at ~12V
B. in parallel, 1B @2V, 2B @3.2V nominal.

Sitting isolated overnight, which of the four sets (if any) will approach getting to balance on their own?

The paralleled LiFePO4 batteries will balance overnight, since voltage is the primary factor that balances those cells (equal voltage roughly means equal states of charge.)

The paralleled lead acid batteries will not balance, since current is the primary factor that balances those cells (charge current brings all cells to 100% eventually.)

In the series no-load case, neither will balance. No cell will see any specific voltage or current so nothing will happen.
 
john61ct said:
Two sets of battery cells

#1. 6x lead cells
vs
#2. 4x LFP cells

within each set, each cell resting at slightly different voltages / SoC, small enough any "self balancing flow" current rate issues are minor.

Two separate intra-pack wiring configurations:

A. the set connected in series, both sets at ~12V

B. in parallel, 1B @2V, 2B @3.2V nominal.

Sitting isolated overnight, which of the four sets (if any) will approach getting to balance on their own?
Sorry if unclear, purely hypothetical questions to isolate the variables.

2 series scenarios
1A: 6S lead
2A: 4S LFP

2 parallel scenarios
1B: 6P lead
2B: 4P LFP
amberwolf said:

amberwolf said:
If in "B." you mean that the individual cells are all completely disconnected from all of their series connections, and then reconnected so all the cells are now in parallel,
Yes that is what I mean, as in a controlled experiment.

> that will balance them

So in parallel. With both lead and LFP?


> LFP will not self-balance, period.
Except in Parallel, right?

Again, isolated blocks, no charging no loads other than each other.

Also, no 4/6/12V lead here, only single cell 2V units, as is common for large banks in marine & industrial use cases.

So, if one lead cell is at 1.85V, and the other is at 2.05V, hooking them up **in parallel** (nominal 2V) will cause no current to flow?

And if one LFP cell is at 2.8V, and the other is at 3.3V, hooking them up in **series** (nominal 6.4V) will cause no current to flow?

So far I believe you're saying (please correct, and of course looking from input from anyone who knows here, not just amberwolf)

1A: 6S lead - UNKNOWN
2A: 4S LFP - NO
1B: 6P lead - YES
2B: 4P LFP - YES
 
billvon said:
The paralleled LiFePO4 batteries will balance overnight

confirms
2B: 4P LFP - YES

> The paralleled lead acid batteries will not balance
So
1B: 6P lead - NO

Even if now **larger** Voltage / SoC differences exist between the 2V cells, (as above, no charge source), no current will flow when paralleled?


> In the series no-load case, neither will balance.
So
1A: 6S lead - NO
2A: 4S LFP - NO



 
john61ct said:
Even if now **larger** Voltage / SoC differences exist between the 2V cells, (as above, no charge source), no current will flow when paralleled?
When cells are paralleled, current will flow until the voltages in each cell are identical. This will do a good job of balancing paralleled LiFePO4 batteries. It will not do a good job of balancing lead acid batteries.
1A: 6S lead - NO
2A: 4S LFP - NO
Right. If they are in series, and nothing is connected to the end of string (i.e. at night no load) then no current will flow no matter what.
 
amberwolf said:
That is completely different from making sure the cells you use are well-matched
I agree.

> BC168 could take days, or even weeks

At .3C, a couple of hours. Remember, not talking multiple cells at a time here.

Each cell is a tiny 20-25Ah only, and a cycle may often draw less than 30-40%. Other charge sources may also be available, so bulk-charge the big bank at its voltage to say 85% Soc, finish with the per-cell balance charger.

And maybe just as a maintenance routine, like regularly conditioning / equalizing lead.

The BC168 and its ilk only charge each cell individually, no bleeding at a lower rate, and charges To the target resting setpoint Voltage. IOW the charge rate (6-8A) **is** the balance rate.

And you can set it to a storage SoC, and the endpoints will be just as effortlessly balanced there as at a high SoC.

 
And yes, well-matched quality prismatic cells are a given, about $1.50 - $2 per AH delivered, likes of Winston/Thundersky/Voltronix, CALB, GBS, A123 & Sinopoly - any other makers suggested in the 60+AH cell sizes? Reliable sources for the US market?
 
And if we're talking about a single 12V bank, rather than just pulling a 12V string out and carrying on, I'd need to break down all three of the paralleled "good" blocks and remove a cell from each, in order to return the bank to a balanced state?
> No, because the automated balancer (any common BMS) would keep the entire battery balanced *for you*.

I'm talking about disaster recovery, e.g. in the middle of the Pacific, need to get my nav system going.

_______
I realize about per-cell BMS monitoring being ideal. But this question here is asking about the S/P factor only, making that facility more or less necessary:
OK, so say the bottom-most buliding block unit is 6x 25Ah cells paralleled into 150Ah blocks @3.2V.

One cell failing does not hold as much immediate danger to its blockmates, but doesn't the fact that that failure only brings down that block's AH capacity, but doesn't have much impact on voltage, make it **harder** to immediately detect at the block level than with the string approach?

In other words, the parallel approach relies **more** on per-cell monitoring in order to detect and fix the problem.
Just answering "you will have a per-cell BMS" avoids the question, rather than furthering my understanding.
 
Back
Top