How closely do cells need capacity matched?

williaty

New Member
Joined
May 14, 2021
Messages
23
I am struggling with my first ever DIY-from-cells battery. It's a 4S LiFePO4 bank using 280Ah cells from SolarSupplyHouse. After being top balanced according to the protocol suggested on this forum, the cells won't stay even close to balanced for more than a few days of light usage. I use this cell in a camper and the goal is to be able to keep my tiny 12V fridge and CPAP on for a long period of clouds, not to support crazy high loads for heating or AC or something. So a normal day sees me pull about 300-350Wh out of the bank and then put it back in the next morning with solar. The most I've ever done in a single day was 750Wh (according to the Victron app) after a couple of days of clouds. Yet by the 5-6th day, the Overkill BMS will disconnect the battery from charging because one cell will go above 3.650V even while the bank voltage is under 13.500V. Most of the time, at least two of the cells are still under 3.400V when the overvoltage protection kicks in.

I labeled the 4 cells A through D so I could track what cell is doing what as I tested things. Looking at the cell voltages through the Overkill BMS, Cell A will always spike during charging first. During the 2 full discharges I did just for testing, Cell A was also the first to hit the low voltage cutoff. To me, that sounded like Cell A must have a smaller capacity than the other cells if it hit both the top and bottom limits first. In general, Cell C is right on A's heels, so I assumed A and C must be very close in capacity. Cells B and D seemed to stay pretty near each other as well but, of course, since they weren't hitting the steep low/high parts of the curve, it was harder to tell for sure.

In frustration, I bought one of the cheap battery capacity testers Will has shown in his videos. I charged each cell up to 3.600V by itself, then discharged it via the capacity tester. I was really shocked at how close the cells are! The biggest difference is between Cells A and B and that's only 2%!

How close do cells have to be capacity matched? Could just the 2% difference I found let the battery bank drift out of balance in under a week?
 

mikefitz

Solar Addict
Joined
May 28, 2020
Messages
887
Could just the 2% difference I found let the battery bank drift out of balance in under a week?
I would not think so, unless there is abnormal unequal self discharge.
You have most likely checked the following:
Cell volts measured with a meter matching BMS reported readings.
Correct tightening of buss bar fixings with clean connection surfaces, ( no bolts bottoming out).
BMS lead terminations making good contact with cable and cell fixings.
There is no additional load on the cells other than the BMS balance leads.
A charged cell does not suffer from high self discharge.

I suggest, since you have the facilities, charge the cells to 3.60 individually, assemble the battery and see how things turn out. If there is imbalance, use the 3.6 charger to 'pull up' the low cells.

Mike
 

kenryan

Solar Enthusiast
Joined
Jun 29, 2021
Messages
153
I'm not sure I understand. Did you do a battery (not cell) capacity test (charge the battery, discharge it, and note how many amp hours you got out of it) when you first started using it?

Did you then do another battery (not cell) capacity test after having used it for awhile?

Was there a difference between the two capacity tests? If so, how much (in amp hours)?
 

williaty

New Member
Joined
May 14, 2021
Messages
23
When this all started, I did not have a capacity tester. I have no idea where the bank started at. I top balanced the pack and put it in service on vacation, which is how I first discovered I had a problem.

I just finished the last per-cell discharge/capacity test tonight. It'll take me about 3 days per cell (so 12 days total) to charge them up to 3.45V/cell so I can put them in parallel and then top balance them up to 3.600V/cell. After that, the little capacity tester will take about 3 more days to run the pack down to 0% SoC. Even at that point, I'll only have a first number for pack capacity and nothing to compare it to.
 

Short_Shot

Photon Sorcerer
Joined
Jul 13, 2021
Messages
1,820
Charge to 3.65v rather than 3.6 and try again?

How low is the charge current dropping to before you're calling each cell charged?
 

williaty

New Member
Joined
May 14, 2021
Messages
23
Charge to 3.65v rather than 3.6 and try again?

How low is the charge current dropping to before you're calling each cell charged?
Less than an amp. Usually, it slows significantly above 3.4V/cell and then drops like someone flipped a switch once the cell voltage hits the power supply voltage.
 

time2roll

Photon Sorcerer
Joined
Mar 20, 2021
Messages
2,419
I suggest, since you have the facilities, charge the cells to 3.60 individually, assemble the battery and see how things turn out. If there is imbalance, use the 3.6 charger to 'pull up' the low cells.
+1 for this. I would just boost charge the low cells with the the pack still assembled a few times and see what goes.
 

Short_Shot

Photon Sorcerer
Joined
Jul 13, 2021
Messages
1,820
Less than an amp. Usually, it slows significantly above 3.4V/cell and then drops like someone flipped a switch once the cell voltage hits the power supply voltage.
Less than an amp isn't going to be a very good balance in my opinion, especially since you aren't going to the full charge voltage.

I take mine to less than 0.1a for 4 cells and they're pretty bang on.
 

williaty

New Member
Joined
May 14, 2021
Messages
23
Less than an amp isn't going to be a very good balance in my opinion, especially since you aren't going to the full charge voltage.
It drops to 900-ish mA and then never drops any lower. There's nothing else happening to wait for.
 

Short_Shot

Photon Sorcerer
Joined
Jul 13, 2021
Messages
1,820
It drops to 900-ish mA and then never drops any lower. There's nothing else happening to wait for.

How long is "never"?
"Never" is a very long time to put .9a into a cell.

Where is this energy going exactly? .9a for 5 hours is 4.5ah.

For 24 hours that's 21.6ah that's going.... where?

Unless you've got something else drawing power, it means the cell is still charging.
 

fafrd

Photon Sorcerer
Joined
Aug 11, 2020
Messages
2,445
I am struggling with my first ever DIY-from-cells battery. It's a 4S LiFePO4 bank using 280Ah cells from SolarSupplyHouse. After being top balanced according to the protocol suggested on this forum, the cells won't stay even close to balanced for more than a few days of light usage. I use this cell in a camper and the goal is to be able to keep my tiny 12V fridge and CPAP on for a long period of clouds, not to support crazy high loads for heating or AC or something. So a normal day sees me pull about 300-350Wh out of the bank and then put it back in the next morning with solar. The most I've ever done in a single day was 750Wh (according to the Victron app) after a couple of days of clouds. Yet by the 5-6th day, the Overkill BMS will disconnect the battery from charging because one cell will go above 3.650V even while the bank voltage is under 13.500V. Most of the time, at least two of the cells are still under 3.400V when the overvoltage protection kicks in.

I labeled the 4 cells A through D so I could track what cell is doing what as I tested things. Looking at the cell voltages through the Overkill BMS, Cell A will always spike during charging first. During the 2 full discharges I did just for testing, Cell A was also the first to hit the low voltage cutoff. To me, that sounded like Cell A must have a smaller capacity than the other cells if it hit both the top and bottom limits first. In general, Cell C is right on A's heels, so I assumed A and C must be very close in capacity. Cells B and D seemed to stay pretty near each other as well but, of course, since they weren't hitting the steep low/high parts of the curve, it was harder to tell for sure.
If the same cell hits LVD and LVD, that is the cell with the lowest capacity and your battery has been balanced to deliver that full capacity as well as possible.
In frustration, I bought one of the cheap battery capacity testers Will has shown in his videos. I charged each cell up to 3.600V by itself, then discharged it via the capacity tester. I was really shocked at how close the cells are! The biggest difference is between Cells A and B and that's only 2%!

How close do cells have to be capacity matched? Could just the 2% difference I found let the battery bank drift out of balance in under a week?
A cell at ~98% SOC will be at approximately 3.371V when a cell at 100% SOC reaches 3.650V:

E7A5C18D-D8AD-450C-899B-7F4982FEDB2E.jpeg
 

Short_Shot

Photon Sorcerer
Joined
Jul 13, 2021
Messages
1,820
If the same cell hits LVD and LVD, that is the cell with the lowest capacity and your battery has been balanced to deliver that full capacity as well as possible.



A cell at ~98% SOC will be at approximately 3.371V when a cell at 100% SOC reaches 3.650V:

View attachment 70333
While this is a good visual, I will reiterate that tables and charts only really apply to the specific cells tested and that specific measurement equipment. Relying on this millivolt precision of these tables to make a determination can be off by double-digit percentage for different cells when comparing them.

It's not good practice to post tables with such granular level of detail because others will find their results don't match and become frustrated, especially when talking about a 2% difference which is dramatically less than the margin of error.

The graphs (without the data labels) are usually pretty useful to illustrate the matter however since the accuracy of interpretation falls outside the inaccuracies of the equipment and cells originally used to make them.
 
Last edited:

fafrd

Photon Sorcerer
Joined
Aug 11, 2020
Messages
2,445
While this is a good visual, I will reiterate that tables and charts only really apply to the specific cells tested and that specific measurement equipment. Relying on this millivolt precision of these tables to make a determination can be off by double-digit percentage for different cells when comparing them.

It's not good practice to post tables with such granular level of detail because others will find their results don't match and become frustrated, especially when talking about a 2% difference which is dramatically less than the margin of error.

The graphs (without the data labels) are usually pretty useful to illustrate the matter however since the accuracy of interpretation falls outside the inaccuracies of the equipment and cells originally used to make them.
My point was that cells charged to only 3.4V when one or two cells charge up close to 3.6 or 3.65V can easily be explained by a capacity difference of as little as 2%…

And I’ll repeat the central point that if the same cell is the first to discharge triggering LVD and the first to charge up enough to trigger HVD, that’s the cell limiting your overall battery capacity and you are doing as well as possible (no benefit to rebalancing).

It’s when one cell hits LVD and another cell hits HVD that you can gain some additional capacity through careful rebalancing..,
 

Short_Shot

Photon Sorcerer
Joined
Jul 13, 2021
Messages
1,820
My point was that cells charged to only 3.4V when one or two cells charge up close to 3.6 or 3.65V can easily be explained by a capacity difference of as little as 2%…

And I’ll repeat the central point that if the same cell is the first to discharge triggering LVD and the first to charge up enough to trigger HVD, that’s the cell limiting your overall battery capacity and you are doing as well as possible (no benefit to rebalancing).

It’s when one cell hits LVD and another cell hits HVD that you can gain some additional capacity through careful rebalancing..,
I agree.

However the major complaint was one cell hitting hvd while the whole pack is only at 13.5v.

This suggests the other cells are nowhere near close to balanced at the top. This isn't hard to do when you're skipping out on the last 50mv and cutting the charge when the current is at 0.9a still.

Just because they are all at 3.6v while on charge doesn't mean they're balanced by SoC. I don't know what the resting voltage is of these cells are after this 3.6v balance or even if they truly are all at 3.600v. There could be a significant margin of error.

If one cell is at 3.65v HVD and the other 3 cells are exactly equal in voltage then they must be at 3.283v each.

3.283*3 is 9.85v. Plus 3.65 is 13.5.

So the balance is way out despite the fact that your conditions of a single cell being the hvd and lvd is met.

While I agree with that point (to a limited extent, it's not as valuable as you think), it's not the sole determining factor about a good balance.

If that one cell is so wildly different in capacity that the other cells are nowhere near full charge or full discharge when it is, then you've got a much bigger problem than balance.

Based on what op stated that seems to be a possibility. Hitting hvd with 13.5 pack voltage is a big red flag.
 
Last edited:

fafrd

Photon Sorcerer
Joined
Aug 11, 2020
Messages
2,445
I agree.

However the major complaint was one cell hitting hvd while the whole pack is only at 13.5v.

This suggests the other cells are nowhere near close to balanced at the top. This isn't hard to do when you're skipping out on the last 50mv and cutting the charge when the current is at 0.9a still.
You are correct - I thought he said the other cells were at 3.4V but you are right, he said the pack was at 13.5V.

If one cell was truly charged up above 3.6V, that only leaves 9.9V for the other 3 cells, meaning the lowest cell can’t be higher than 3.3V (meaning potentially no more than ~70% full).
Just because they are all at 3.6v while on charge doesn't mean they're balanced by SoC. I don't know what the resting voltage is of these cells are after this 3.6v balance or even if they truly are all at 3.600v. There could be a significant margin of error.

If one cell is at 3.65v HVD and the other 3 cells are exactly equal in voltage then they must be at 3.283v each.

3.283*3 is 9.85v. Plus 3.65 is 13.5.

So the balance is way out despite the fact that your conditions of a single cell being the hvd and lvd is met.
Yes, I agree, the top-balance is way, way off. There may also be a higher-resistance connection to that one cell which would result in it showing a higher voltage while charging which can trigger HVD and the disappear the moment after charging stops.

The OP should connect a multimeter to the runner cell and watch what happens to the voltage as it charges to HVD and triggers an end to charging. If it’s voltage drops by much more than the others on the seconds after charging, he may have a connection problem.

While I agree with that point (to a limited extent, it's not as valuable as you think), it's not the sole determining factor about a good balance.

If that one cell is so wildly different in capacity that the other cells are nowhere near full charge or full discharge when it is, then you've got a much bigger problem than balance.

Based on what op stated that seems to be a possibility. Hitting hvd with 13.5 pack voltage is a big red flag.
Yes, if that same cell is triggering LVD, he got a much weaker / lower-capacity cell that is limiting overall pack capacity (and no amount of rebalancing will change things).

It’s not clear to me whether he measured cell capacitor at 150W at the individual cell level or the full battery level, but even if measuring at individual cell level using a full 10A, that is a discharge rate of less than 0.04C.

He has not said anything about discharge current under actual load, but if his fridge consumes anything like mine when it is running, it can easily draw over 20A from a 4S battery. CPAP machine, I have no idea, but it sounds as though peak discharge current can easily be 3 or 4 times the 10A he used when capacity testing.

It’s the capacity under actual load that matters most and just because he got over 270Ah discharging his runner cell at 10A doesn’t mean it’ll deliver close to that when discharged over 0.1C (28A).

I agree hitting HVD with pack voltage at 3.5V is a huge red flag but that same cell hitting LVD with all the other cell are all safely above that level is an even bigger red flag.

My first LiFePO battery (90Ah) had a weak cell in the pack. It only delivered ~70% of the capacity of the other cells (this was before I had any idea what I was doing and bought from a Noname reseller).

I just had to purchase a replacement cell to get closer to full 90Ah capacity of that battery.

So possibly a bad connection but more likely a bad cell is my prognosis (assuming the OP is correct about the same cell triggering both LVD and HDV).
 

Short_Shot

Photon Sorcerer
Joined
Jul 13, 2021
Messages
1,820
You must have an absorption fridge. A straight 12v dc compressor fridge will only pull like 60-80w tops. Certainly not 20a+

Either that or a big old residential boi with defrost.
 

fafrd

Photon Sorcerer
Joined
Aug 11, 2020
Messages
2,445
You must have an absorption fridge. A straight 12v dc compressor fridge will only pull like 60-80w tops. Certainly not 20a+

Either that or a big old residential boi with defrost.
Yeah, it’s the biggest 110VAC fridge Samsung makes. Duty cycle is pretty low but when it decides it’s time to cool things down, power draw is over 500W…

My system is load-shaving and time-shifting standard home consumption.
 

williaty

New Member
Joined
May 14, 2021
Messages
23
OK, after extensive testing, I am back.

First, the fact that I never saw the current drop to zero was because I had too much resistance in the connection between the power supply and the cells. With a direct wired/bolted connection, things behave as I'd expect. Here's where I'm at now:

I labeled the cells A through D so that I can clearly identify the cells independent of their position within the pack.

Due to having suspicions about what I was seeing, I took the time to order a battery capacity tester of the kind that Will Prowse recommends. With this tester, I was able to run synthetic "days" that mimic my real-world use of the battery while camping since the tester allows me to vary the load and it counts total energy dissipated which allows me to know when to stop discharging to mimic a day's worth of use. Additionally, I have run every cell individually from 3.65V down to 2.60V twice and have twice run the bank from 3.65V/cell down to the first cell hitting 2.50V/cell in order to measure total capacity. To remove a variable from the testing, I have been using only the Victron solar charge controller for charing (powering it via a DC benchtop power supply rather than solar panels so I'm not dependent on the sun being out). With all of this, I have come up with the following problems.

1) The bank cannot make its rated capacity. Discharged at the 20-hour rate with the cells being at 70F, the bank capacity is only 272Ah. I repeated the test twice to confirm this number.

2) The bulging problem got worse, not better, with deep discharges. Cells A and C have gotten so bad that I had to file a longer slot into the bus bars because they wouldn't reach between the studs.

3) I have been consistently torquing the nuts on the studs to 30in-lbs using a CDI Torque Products torque screwdriver. The driver was last calibrated in 2019 and has never required adjustment any of the other times I sent it out for calibration over the last 15 years. Even with being at the lowest torque setting I could find recommended in any of the datasheets for prismatic cells like this, I still ripped one of the studs out of Cell A. I drilled and tapped the terminal on the mill to accept a Helicoil but even that feels soft now due to the fact that the pocket in the terminal is so shallow that the helicoil bottoms out against the pocket with only about 3 threads of engagement. I am not confident this repair will last if I have to keep taking the bank apart to put it back into parallel and top balancing it every time I take a trip.

4) The problems follow the cells regardless of which position they are in the pack. Cell A is the worst. It will always hit the high voltage cutoff first and always hit the low voltage cutoff first. It's clearly lower capacity than the other cells. However, Cell C is close on its heels. Cells A and C are never more than 10mV different when Cell A hits cutoff. Cells B and D are significantly larger capacity than A or C as they will be 250-350mV away from the low/high cutoff when Cell A triggers a shutdown.

5) The pack will slip out of balance in 4-6 synthetic "days" of testing. My real world use pulls 300-750Wh per day out of the pack. For this testing, I rebuild the bank into parallel, top balance to 3.650V, reassemble into series, and cycle the pack down 300-750Wh using the capacity tester as a load and then the Victron SCC to recharge. I alternate between "days" at the higher end and lower end of that range to try to make a more realistic test. The pack gains 60-90mV of imbalance per "day" when doing this. Eventually, Cell A will trigger the cell overvoltage protection on charging, usually with the bank voltage around 13.5V resulting in an imbalance between the highest and lowest cells of around 350mV.

With those results, here are my current thoughts:

a) The battery bank as-is isn't functional. It's unreasonable to constantly have to take it apart and top balance it. As I mentioned, there's also no way the terminals are going to survive that.

b) The battery bank cannot be "fixed" by replacing a single cell. It appears A and C are pretty close to matched and B and D may be pretty close. This means at minimum 2 cells have to be replaced but I have no idea how I'd get cells that are going to match whatever 2 cells are kept. Replacing just 2 cells wouldn't make the pack live up to its rating anyway as none of the individual cells test at or above 280Ah.

c) The bulging issue is a serious problem since these were listed as Grade A Matched cells.



OK, can you guys think of anything I'm missing here or any other way to patch the problems I've got going?
 

time2roll

Photon Sorcerer
Joined
Mar 20, 2021
Messages
2,419
Are you saying you bought 280 cells and the battery produces 272 Ah? Not exactly a disaster.
Maybe an active balancer would help avoid a manual effort to rebalance but the capacity is what it is.
 

williaty

New Member
Joined
May 14, 2021
Messages
23
Are you saying you bought 280 cells and the battery produces 272 Ah? Not exactly a disaster.
Maybe an active balancer would help avoid a manual effort to rebalance but the capacity is what it is.
Correct, listed as Grad A Matched 280Ah cells and I'm struggling to get them to be useful.
 
Top