diy solar

diy solar

Parallel Cell Capacity Balancing (PCCB) Procedure.

Posplayr

New Member
Joined
May 27, 2021
Messages
152

INTRODUCTION


I’m going to describe my setup and the initial electrical performance results as well as a balancing technique that I think is pretty effective at matching parallel cells (e.g an 8 cell 4S battery or a 16 cell 8S battery).

I’m a semi-retired Electrical Engineer, but this is my first foray into doing a DIY LiFePO4 build. This battery is planned for a van build that is in the process for that will include 350W of solar/40A Solar Charger, 40A DC-DC charging, 1100W MSW/600W PSW inverters.

The process is related to this thread which is discussing a charge capacity balancing process of paralleling cells to get better capacity match across battery arrays. This is essentially the same approach I have tried to formalize and test with hard numbers for a 4S battery using two parallel cells per 1S. This works well for example building a 200 Amp-hour 4S battery using 100Amp-hr cells. The goal is to achieve an improved capacity of used B cells to make them comparable to new A cells.

https://diysolarforum.com/threads/diy-cell-matching-process.16784/

FaFrd describes his possess in this post. He seems to be going through an iterative process to rank strongest to weakest cells. For example, he similarly does a top balance t o3.65V, then configures a 16S 48V battery to discharge and determine the lowest capacity cell. He then dropped that cell and continued down using a 15S battery. He continues down apparently draining each cell 2.5V depletion.

https://diysolarforum.com/threads/diy-cell-matching-process.16784/#post-191353

In my procedure, I think the ranking can be achieved a little simpler, by simply balancing the current draw from all cells using a parallel 4S configuration where I have a total of 8 cells. The trick is to get the current draw balanced which is easies with cell pairing as will become obvious when looking at the Step 2 mirror image battery cell configuration. This can be extended but even numbers of cell per 1S are going to be the only thing manageable. If you are going to be paralleling more than 2 cells then it might be best to do multiple discharges of 4S batteries to obtain the 48V load equipment that faFrd is describing.

Any comments and or advice is welcome. That is why I'm posting.

SYSTEM COMPONENTS


So I just recently received an order from AliExpress for:

VariCore 3.2V 102Ah Battery LiFePO4 Lithium phospha Large capacity DIY 12V 24V 48V Electric car RV Solar Energy storage system
Qty 8 100Amp hour LifePO4 cells $343.36 shipped.
Ordered May 28, 2021
Received July 28, 2021
https://www.aliexpress.com/item/4001253443449.html?spm=a2g0s.9042311.0.0.24b84c4dJ9NxR9

View of my bench test setup.

IMG_4027.JPG

Closeup of the 4S battery pack with active balancer installed. The pigtails are prewired from the OverKill BMS (not connected).​

IMG_4028.JPG

The AiLi Battery Monitor (with 100 amp shunt)
IMG_4029.JPG

Description of Components:


Overall the VariCore 100Amp-Hr cells are in good shape except for obvious swelling > 0.15” (both sides) per cell. This will make compressing the cells a bit more challenging. These are supposed to be Grade A cells but are probably refurbished cells that hit their 80% EOL and have been derated 20% , repackaged and resold as Grade A. They are rated for 200 cycles which is over 5.5 year of daily cycle use so for the price I am not going to complain.

Amp hour reading are based on the AiLi battery monitor. The measured capacity is meeting or exceeding 200Amp-Hr but I have yet to do a final test of the balanced cell pair and using active balancing. After doing an initial top balance, I used a Kreiger Modified Sinewave 1100W inverter with a heat gun adjusted for 500W load to do the first discharge capacity test.
https://www.amazon.com/gp/product/B00T564EIY/ref=ppx_yo_dt_b_search_asin_title?ie=UTF8&psc=1


AiLi Battery Monitor Voltmeter Ammeter Voltage Current Meter 8-80V 0-100A Auto Car Motor Boat Caravan RV Motorhome
https://www.amazon.com/gp/product/B07CTKYFTG/ref=ppx_yo_dt_b_search_asin_title?ie=UTF8&psc=1

The inverter shut down for low battery voltage when the cell voltage was still at 12V (3V cell equivalent). The invertor was probably sensing about <11.5V owing to voltage drops of the 500W heat gun load (48A DC load). So my initial discharge capacity estimate is based on the Ali Battery monitor going from 200 Amp-hr down to 20.75 Amp-hr indicated with 3.0V remaining nominally per cell. This means that 179.25 Amp-hrs is 92% of SOC (assuming 3.0V is 8%SOC). Therefore 179.25/.92=194.8 Amp-Hr total capacity. At this point, I’m not concerned with being below 200Amp hours. The Ali battery monitor is supposed to be accurate to +/- 1% or +/- 2 Amp-Hrs but I take this with a grain of salt. I have compared the current reading from the invertor, the power supplies, a clamp-on meter and the Ali Battery monitor. While doing the 500W load test the Ali-was indicating a significantly lower load current (25-35 amps) so I am surprised it came as close as it did based on above measurements.

I will also be doing some initial evaluation of a 5A active balancer also purchased from AliExpress. I will only use this to top balance the final battery configuration following a 12V/12A recharge.

5A - 8A Active Balancer Equalizer 4S Lipo Lifepo4 Li-ion Battery BMS High Large Blance Current Dynamic Conversion Module Board Paid $40.53 shipped.
https://www.aliexpress.com/item/4001013797039.html?spm=a2g0s.9042311.0.0.24b84c4dJ9NxR9

Although I have not done connected a BMS yet it is worth mentioning that, I originally purchased (Qty 2) Daly smart BMS LiFePO4 4S 60A for a drop-in equivalent 100Amp hour batteries. Later, changed my plan to a combined 200A battery with parallel cells and have the Overkill Solar 4S 120A BMS. At this point I have not connected the BMS, and have only used the Balancer to top balance the (Step 4 configuration) Combined 4S battery at 14.6V (3.65V cell equivalent).

https://overkillsolar.com/product/bms-120a-4s-lifepo4/


 

Attachments

  • PCCB_Steps.png
    PCCB_Steps.png
    93.8 KB · Views: 53
  • PCCB_TheorticalCombinations.png
    PCCB_TheorticalCombinations.png
    106.2 KB · Views: 44
  • LiFePO4_4S_voltages_SOC.png
    LiFePO4_4S_voltages_SOC.png
    111.7 KB · Views: 35

(continued)​

Balancing Strategy with Summary Benefits

Top Balancing is a standard practice recommended by Over-Kill Solar(see BMS manual) and Will Prowse

Top balancing involves placing all cells in parallel and charging to some maximum voltage. For the VariCore cells the maximum is specified at 3.65V (4S equivalent of 14.6V). I’m going to report results on a Parallel Cell Capacity Balancing (PCCB) procedure as well as the operation the 5A active balancer. I assume someone else has probably done PCCB before but I have never seen it described so I will describe it here.

PCCB uses the Top Balance procedure as an intermediate step, but goes further by matching cell amp-hr capacities. This is achieved by pairing up cells with MIN and MAX cell voltages (SOC) following a discharge from a top balance. For example, in a 4S battery involving 8 cells, there are theoretically 50 cell combinations that provide significant opportunity to match cell pair capacities. By combining in parallel a low capacity cell with a high capacity cell, all parallel cells tend to the mean of the group of cells. This goes well beyond the top balance because the PCCB is reducing the capacity variation between cells. The cells can then be re-top balanced and should achieve even better balance under discharge. In fact with a balancer, the initial top balance becomes less and less significant as the system is cycled through charge-discharge cycles whereas the capacity balance of PCCB remains unchanged (assuming capacities stay relatively stable).

Top Balancing matches voltages and therefore SOC for all individual cells, but it does nothing about individual cell capacities. In PCCB, by combining in parallel low capacity cells with higher capacity cells, the average capacity of cells tends to the mean and so the overall voltage spread of cell voltages in a series configurator in significantly reduced.

In the data to be presented, we did a discharge test of a top balanced set of cells and found that by the time the inverter shut off at a rested 3.0V nominal cell voltage there was a 1.306V MIN to MAX spread between up paired cells. After an optimal 4S cell pairing, this MIN-to-MAX spread had been reduced to 0.176V. This is a 7.4:1 reduction in the cell voltage spread and essentially represents a significantly better set of matched cell voltage. This corresponds to better capacity pairing of cells without any cell replacements.

In summary, we have to recognize that the cell voltage spreads depend on how deep into SOC discharge or how high the SOC each cell is. However, this example shows that the method is effective in reducing spread albeit dependent on the specific capacity mix of your set of cells. PCCB is very effective for the DIYer that wants to use cheaper used grade B cells as the grade A matched cells are going to be more expensive. In this reference example of a 8 cell 4S battery, using PCCB with active balancing appears to essentially mitigate any capacity reducing side effects of using mismatch grade B cells.

4S Parallel Cell Capacity Balancing (PCCB) Procedure.

In this section, I will summarize the procedure I used for PCCB on a 4S battery. There are essentially 4 battery configuration sets (labeled #1-#4)

View attachment 58370

This is a summary of the theoretical number of combinations of cells array available for paralleling individual 1S cells.

View attachment 58371
This is my Voltage to SOC reference set largely culled from Powerstream.
https://www.powerstream.com/LLLF.htm

Note that they claim:
" When measured with a differential scanning calorimeter (DSC) the exothermic heat of the chemical reaction with electrolyte after overcharge is only 90 Joules/gram for LiFePO4 versus 1600 J/g for LiCoO2 . The greater the exothermic heat, the more vigorous the fire or explosion that can happen when the battery is abused."

"A LiFePO4 battery can be safely overcharged to 4.2 volts per cell, but higher voltages will start to break down the organic electrolytes."

This is equivalent to 16.8V!!! for 4S and far exceeds 3.65V (14.6V 4S) typically quoted. So while not advised to make a habit of it, as centra in amount fo voltage above 3.65 is tolerable. That said there does not seem to be much benefit for setting the upper voltage above perhaps 3.5V of 14.0V 4S provided cell balance is no tan issue.

View attachment 58372


Excel Spreadsheet in Dropbox

 
(Continued 2)
Step #1 Top Voltage Balance

This is essentially the standard Top Balance recommendation. The point is to get all cells to the same voltage at the top of the voltage range. For PCCB , by pushing the voltages to say 3.65V we have all SOC the same at the point. In the middle of the voltage range even some voltage difference can correspond to a large SOC difference. Note: just because a cell is at the same SOC level does not mean it has the came capacity. It is like having two full glasses of water. They are both filled to the brim (e.g. same SOC), but since one is larger than the other corresponding to a higher capacity (e.g. high Amp-Hr ).

Once all cells are voltage balanced we reconfigure to a pair of 4S batteries for a balanced discharge. Note that if the battery connections are not balanced, there can be large difference in current flow to/from a cell.

Step #2 Discharge Current Balance

In this step we want to expose the differences in cell capacities so we can match the final pairs by pairing a high capacity with a low capacity in a particular cell pair. We also need a 4S configuration so that we can use out invertor and resistive heat gun. So we create two Parallel 4S batteries. While the top and bottom of the pair will be the same, the variation between the cells in each battery can vary depending upon their respective capacities. High capacities will have higher voltages while the lower capacity cells will discharge lower under the same total load. Please note it is very important to match the paths as shown in the figure using bus bars. The cells are arranged in a mirror-image of each other so this can be accomplished. Not doing this will essentially create imbalances invalidate all results. In my testing I used the inverter and 500W heat gun load to discharge the 200Amp hour pack down till the invertor shut off.

Step #3 Cell Voltage Pairing

This is the important set to pairing to achieve more consistent capacities in cell pairs. In my example, I am taking a set of 8 100 Amp-hr Grade B mismatched cells and creating 4 Grade A matched 200 amp-hr cells pairs. As show the process is very easy for pairing. Take the highest voltage cell and pair it will the lowest voltage cell. Continue this pairing as shown. If you have 12 cells instead of 8 you could achieve even meter matching but the optimal algorithm is more complex and beyond the scope of this discussion.

In the table below extracted from the spreadsheet, we can see that following the Step 2 discharge, the lowest cell voltage of 2.747V corresponds to a 10.988V 4S voltage while the highest of 3.074V corresponds to 12.294V. This is a 1.306V difference. After pairing in Step #3, the minimum cell voltage is 2.978V and the maximum is 3.022V corresponding 4S voltages of 11.912v and 12.088V respectively. As mentioned above we are now down to only a 0.176V MIN-MAX voltage dispersion.

Step2_Spreadsheet.png

In the final step #4 we will use the selected pairs to construct the final 4S battery and ready it for recharging. We should be able to observe how well the un-rebalanced cells charge to the maximum at 14.6V (4S terminal voltage at 3.65V cell voltage) . We expect less dispersion at the top than previously because the now the capacities and SOC of each pair are better matched.

Step #4 Final 4S Installation.

The results of the recharge are shown in the table below. The balanced pairs are now nominally 3.0V for a 4S total of very close to 12V and 0.176 MIBN-MAX spread. After an overnight charge of 11Amps we see a 13.44V terminal voltage with 150 amp hours indicated (net 130 amp-hour charged). All cells are at 3.36 indicated indicating there are perfectly balanced (improvement even over the previous 0.176V spread. This is a somewhat surprising result because this is without an active balancer.

Step3_Spreadsheet.png

The next row indicates the completion of the non-active balance phase when one of the cells exceeds the 3.65V limit by achieving 3.73V (corresponding to 14.92V 4S). At this point the MIN-MAX spread is equivalent to 1.32V because one of the cells is at only 3.4V corresponding to a 13.6V 4S cell pack.

At this point, the observant reader might suggest that this is now worse than the Step #2 1.306V voltage spread. But we have to realize that this prepared configuration of cells has never really been top balanced and so this type of comparison is apples to oranges. My assertion is that while the pairs are (the best combination) capacity balanced, they are not top balanced in this configuration. In a future test I will top balance the pairs and do the full discharge test to see fi the dispersion at the bottom is reduced prior to pairing. But at this point I will show what happened after connecting the active balancer.

The last row shows the results from an active balance at maximum voltage of 14.56V (approx the 3.65 cell maximum). The charging voltage for the Topward 6306D power supply was adjusted to keep the overall 4S voltage at 14.6 while letting the active balancer adjust the voltages of the cells. Over a period of about 20 minutes #D , the lowest cell at 3.4V, was charged up to 3.65V. The remaining cells are 3.66V, 3.62V, and 3.63V corresponding to #A, #B, and #C respectively.

The conclusion is that the active balancer is capable of balancing the individual cells in a relatively short period given it has a 5A capacity and using 200Amp-hr cell pairs. In effect, the active balancing diminishes any longer-term effect due to top balancing. While the PCCB procedure requires a top balance to measure relative cell capacity, the active balance negates the long-term operational effects while complementing the PCCB cell pair capacity matching.

I will follow up with test results of the PCCB matched 4S battery active top balanced and discharged using the same 500W load till the inverter shuts down. Test in process
 
Last edited:
EDIT: I was able to restart the Inverter after it cooled off. In the original posts, you will see that I reported the inverter (overheated) stopping at about 100 Amp-hr 13.04V (3.26V at the cell level). The final part of the discharge test ran a bit lower than before (approx 11.979 volts) to an average of 11.58V or equivalent to 2.895V at the cell level. This voltage might rise more with rest but for now, I'll call it close enough. In fact, it should have more spread as the voltages are lower. The result is that the MIN-MAX spread at 4S is only 0.4V compared to the 0.160 predicted at the end of Step #4 but much better than the 1.32V MIN-MAX spread in Step #2 2X 4S parallel configuration of single cells. The improvement for this capacity match is 1.32/0.4=3:1

This last data set validates the improvement obtainable in the Parallel Cell Capacity Balancing (PCCP) Procedure. The capacity matching process is effective at improving voltage/SOC matching across a larger population of cells. (e.g. 8 cells in a 4S battery pack).

(The original post before the incomplete termination continues here).

I ran a load test after Top Balancing but after 3 hours and 15 min the invertor seems to have failed after drawing 100 amp-hrs of 200 amp -hrs.
The cells were as matched as they were at the beginning (0.040v 4S equivalent MIN-MAX).

I will let the invertor rest and see if I can draw the batteries down anymore. I'm not claiming this level of balance will hold down to 3.0V but I'm hopeful for an improvement over the Step #2 results. Those results indicated a 7.5:1 improvement.

Unfortunately, that test also blew/reset my AiLi battery monitor.

Step5B_Spreadsheet.png
(
 

Attachments

  • Step5_Spreadsheet.png
    Step5_Spreadsheet.png
    30.1 KB · Views: 13
Last edited:
Wow, a lot of work there. Thanks for sharing.

I guess your testing will eventually show that no matter how you configure your cells, the total battery capacity is limited to that of the weakest cell, unless you use active balancing.
 
Wow, a lot of work there. Thanks for sharing.

I guess your testing will eventually show that no matter how you configure your cells, the total battery capacity is limited to that of the weakest cell, unless you use active balancing.
Well, I like to dive in and understand something well.

Actually, the calibration is to match weakest with strongest cells so the weakest pair of cells tend to be somewhat random but not as weak as the weakest individual.

As far as the balancer, yes it can compensate for a lot of error at low loads, however, it does not act instantaneously as it has a limited current carrying capacity. So it is always playing catchup and the better the cells are matched the less equalization required and probably the better the efficiency to stay in sync.
 
Last edited:
Actually, the calibration is to match weakest with strongest cells so the weakest pair of cells tend to be somewhat random but much less than the weakest individual.

In a pair of cells, both cells will have the same capacity of the weakest cell. They don’t average out.

ie: you can parallel a 50ah and a 200ah, and the capacity will be 50ah, not 100ah.
 
In a pair of cells, both cells will have the same capacity of the weakest cell. They don’t average out.

ie: you can parallel a 50ah and a 200ah, and the capacity will be 50ah, not 100ah.
It is a sum, not an average. A 110 amp hour cell and a 92 amp-hour cell have a 202 Amp-hr capacity based on the principle of conservation of energy( energy can be neither created non destroyed). In your example, 50 and 200 amp hours single cells in parallel is 250 amp-hrs.

If you could combine 50 plus 200 and get anything but 250 you are violating CE.

I realized you must be thinking of something akin to series solar panels which have a current source. If one of the panels has a reduced current then it reduces the current in all panels.

In your example, you could add a million 1 million amp-hour cells to your little 50 amp-hour cell and get only 50 amp-hours.

What if you added zero amp-hours to a million 1 million amp-hour cells would you get zero amp hours? Clearly no.

You get X=X+0 whatever X is.
 
Last edited:
You are correct, sorry my response was wrong. What i meant to say was the capacity of the battery will be limited by the weakest cell.

Got my tongue twisted over parallel cell pairs, and cells in series for a battery.

As you say, matching the strongest and weakest cells raises the capacity of that parallel pair.
 
What i meant to say was the capacity of the battery will be limited by the weakest cell.
Well, this is true if you cant do capacity matching (e.g PCCB) and have no active balancer. However, PCCB mitigates this effect to a significant amount. In the case of my 8 cells, it reduced the MIN-MAX variation by 3:1 in terms of MIN-MAX voltage.

A simple example for 8 cells. All cells are 100 Amp-hr except there are two oddballs; one is 96 Amp-hr and the other is 105 amp-hrs.

8 individual cells: 100,100,100,100,100,100,96,105 (all amp hours)

If you took the 8 cells and built 2 4S batteries in parallel then the 96 amp-Hr battery would trigger the BMS (low voltage cell) first from a discharge test after top balancing.

However, if we pair cells with the HI and the LOW together then we end up with 201 Amp-hr and 3 remaining perfectly match 200 amp-hour cells.

4 paired cells: 100+100=200, 100+100=200, 100+100=200, 96+105=201

Now you build a 4S battery with these pairs and there is only a 1 amp-hour (of 200 Amp-hr) or 0.5% capacity error in that single cell. Before you had 4 amp-hour deviations out of 100 amp-hours average which is 4%. This example is an 8:1 matching improvement on a percentage basis.

In reality, there is more variation than this and for 8 cells and 4S battery configuration, there are 50 different coimbinations to minimize the spread.

As you say, matching the strongest and weakest cells raises the capacity of that parallel pair.
More specifically, parallel combinations of cells have a total capacity of the sum of individual capacities. This assumes of course you have them right next to each other with bus bars so there the voltages at the terminal are virtually identical.
 
I hope i didn’t kill your post by posting a brainfart for a first response.

I think what you have posted is very thoroughly done, and i’m surprised more people haven’t commented.

I know balancing is like a religion on this forum - surely anyone going for the optimum initial setup would like to see what you have done.
 
I went back over my data and corrected any transcription errors in the spreadsheet and did a little reformatting to be able t better assess the charge/discharge cycle results after the initial to balance for the capacity matched cells.

Here is a Summary of 1 1/2 round trip cycle. (Discharge/Charge/Discharge) of the top balanced final step #4 configuration. I may continue this but I feel I need to package the cells with compression before continuing too much more. Unless noted there is no BMS nor active balancers. All charging-discharging and data are recording are manually performed using the Fluke VOM.

Charge/Discharge Cycles

1.) After the top line top balance the cells are nearly Voltage/SOC matched at 3.65v nominal After a discharge to 11.58v (all cells under 3.0v) the #2 cell is lowest (2.843v) and the #4 is the highest(2.943v). This indicates the relative capacity of each respectively as well. #4 has the highest capacity because it dropped the least and #2 has the lowest capacity because it dropped the most.

2.) I did a little Active bottom balance to see how effective the 5A balancer is. You can see that although all the voltages were rising during a rest period (no active charging), the group of cells was tending to a mean value. regardless the net results left the same cell #2 at the lowest SOC and the #4 at the highest. The total balance time was 1:15 min using the 5A balancer on 200 Amp-hr cells. I suspect it would take 4-5 hours to completely balance the cells.

3.) On the first charge cycle, the Amp-hr monitor was set to 200 amp hour, and the batteries were charged to a nominal 13.7V. Consistent with the initial assessment, Cell #2 comes to the highest SOC at 3.47 while the rest are all at 3.4V. This indicates that #4 may not have as high of a capacity of we had assumed and perhaps #3 is higher because it ended with the same 3.4V although it started with the lowest of the cells excluding cell #2. In any event, we have very good voltage tracking during the bulk of the charge cycle. The charge cycle was terminated when The #2 cell popped up to 3.47V above the other three which were at 3.40V.

4.) The final discharge began after an 18 hour rest during which time the #2 cell had dropped relative to the rest of the battery cells but still claimed the top spot along with cel #4 at 3.35V. After 180 amp-hrs of 40A discharge, the inverter shut down due to low voltage limits at approx 11.37v. The #2 cell is now the lowest cell at 2.765V and the #4 cell is highest at 2.909V.

Summary:
Comparing paragraphs 1 and 2 results with this paragraph, and while the values vary somewhat, we have a consistent pattern that cell #2 is the lowest capacity while cell #4 is the highest capacity.

In the voltage range between 3.20v-3.40v (12.8v -13.6v which is nominally 40 of capacity ), there is nearly ideal open-loop tracking of the 4 capacity matched cell pairs. The voltage spread tends to increase at voltages above 13.6 and below 12.92v. While there is divergence at these endpoints it seems the MIN-MAX 4S Equiv metrics are generally below 0.5V (.125 volts on one cell). The initial Step one voltages deviations were approx 1.3V ( corresponding to 0.327 at the cell level).

We have generally validated a reduction of the cell to cell capacity variation that can not be compensated for using top balancing but can be minimized using parallel cell matching. The number of combinations for 4S, 8S, and 16S cell configuratuions are:

Combinations for cell voltage balancing:
•4S Paired cell configuration there are 28 combinations
•8S Paired cell configuration there are 120 combinations
•16S Paired cell configuration there are 496 combinations

So the expectation is that this type of capacity balancing will be even more effective for larger cell arrays

.

ChargeDischargeCycle.png
 
I hope i didn’t kill your post by posting a brainfart for a first response.

I think what you have posted is very thoroughly done, and i’m surprised more people haven’t commented.

I know balancing is like a religion on this forum - surely anyone going for the optimum initial setup would like to see what you have done.
Thanks for the support. Any traffic is better than no traffic. I posted a wrap-up at least for now.
 
In the first post:
"While doing the 500W load test the Ali-was indicating a significantly lower load current"
In my first solar, DC, and AC system I had a lead acid battery and a modified square wave inverter. The inverter current is a bunch of pulses and my amp hour meter was way off and inconsistent. The microcomputer in the amp hour meter samples the current at a fixed rate that is not related to the frequency of the inverter's 120 pulses per second. In theory that could work but in practice the result was of no value. I resorted to using a slow ammeter and recording the current every 15 minutes.

I have a question. In step 2 where the cells are discharged to determine their capacity, what makes the left branch cells discharge at the same rate, amps, as the right branch cells? The discharge time, hours, is matched by a common wire to disconnect the discharger. If all the cells are instead series connected then every amp hour passing through one cell passes through all cells. Since the point is to identify the mismatch of cell capacity the cells must be presumed to have differing capacities which would cause differing rates of change of state of charge, even if the internal resistances and external resistances are well matched. The 4S2P configuration doesn't force the same amps in both of the 2P paths the way an 8s configuration would.

I don't have a solution to the problem of how to load a 1P8S 24 volt battery when you have a 12 volt inverter. I have a 120 volt resistive space heater and in my case with 32 cells, 1P32S, they will be 117 volts at 3.65 volts per cell dropping to 96 volts at 3 volts per cell.
 
I have a question. In step 2 where the cells are discharged to determine their capacity, what makes the left branch cells discharge at the same rate, amps, as the right branch cells? The discharge time, hours, is matched by a common wire to disconnect the discharger. If all the cells are instead series connected then every amp hour passing through one cell passes through all cells. Since the point is to identify the mismatch of cell capacity the cells must be presumed to have differing capacities which would cause differing rates of change of state of charge, even if the internal resistances and external resistances are well matched. The 4S2P configuration doesn't force the same amps in both of the 2P paths the way an 8s configuration would.

I don't have a solution to the problem of how to load a 1P8S 24 volt battery when you have a 12 volt inverter. I have a 120 volt resistive space heater and in my case with 32 cells, 1P32S, they will be 117 volts at 3.65 volts per cell dropping to 96 volts at 3 volts per cell.
This is a good question and centers on a fundamental element of the approach I am advocating. While yes you are quite correct, putting all cells in series and discharging them ensures that all cells have the same charge going through them would in a theoretical sense be the most accurate means of ordering the capacity of all the cells. For reference see the link to FaFrd's post on the procedure he used.

FaFrd describes his possess in this post. He seems to be going through an iterative process to rank strongest to weakest cells. For example, he similarly does a top balance t o3.65V, then configures a 16S 48V battery to discharge and determine the lowest capacity cell. He then dropped that cell and continued down using a 15S battery. He continues down apparently draining each cell 2.5V depletion.

https://diysolarforum.com/threads/diy-cell-matching-process.16784/#post-191353
The objective of what might appear as a flawed configuration (i.e. 2 parallel 4S batteries) is an attempt to rank capacities using the planned equipment rather than some other calibration equipment you might have on hand. Voltage is the primary measurable of our BMS, so this indirect method of SOC matching is more direct in relation to how the BMS will operate.

So what is going on in what I describe? As a general procedure, say we are trying to build an Ns battery (N=4,8,12,16..etc) and we plan to parallel two cells for each s cell. We can rely on the statistics of the situation and the assumption that a higher voltage is indicative of a higher SOC especially at the upper and lower limits of cell voltage. So our proxy for SOC is voltage and we are not going to concern ourselves with the fact that we did not measure the exact current through all cells. As I mentioned previously you can parallel up even more cells, but the matching becomes more complicated, whereas higher numbers of stacked cells is a simple MIN-MAX selection process that gives you an adequately rich number of combinations with simple cell pairing.

What is the relevance of this voltage/SOC assumption? Well, when we place two Ns batteries in parallel, although we don't know exactly what the current split is, we do assume that the two parallel nS batteries have the same average cell SOC. That is because all series cells in each Ns battery sum to the same voltage as its dual. Of course, you can make your life miserable by purposeful daisy chaining the charging batteries in parallel and causing a huge mismatch that requires hours to settle. But if you keep the configuration well balanced and let it rest after the charging process, the two parallel Ns batteries should converge/rest to the same voltage. You will know this when there is no relative flow between the two parallel batteries. I should note that I did allow for a rest (15-20min) after the Step #2 discharge but did not confirm that there was no flow between the two parallel Ns batteries before regrouping.

When the two Ns batteries are balanced, not only are each Ns battery at the same voltage but also each battery has the same average cell voltage and by assumption at the same average cell SOC. At this point, whether you compare voltages of individual cells with any other in the grouping or compare cells voltages relative to a fixed constant average voltage, it is all the same. So you can freely group the mixed set of cells based on the cell voltages in the two groups. They are all referenced to the same average voltage/SOC.

To reiterate, this works because we are measuring the voltages where voltage spread will be significant and directly indicative of SOC differences. This does not work in the mid-range. The primary goal is to be able to charge and discharge the Ns batteries with the same operational equipment configuration after we parallel up the individual sells Ns2p.

As an additional sanity check, I went back to reevaluated my pairing of cells based on ordering cells with respect to their individual battery averages. This is to see if my not allowing the Step #2 configuration enough time to settle might have caused a sub-optimal pairing. The net result is that the biggest outliers in the individual top balanced cells are #4 and #8 stayed the same. There is a different mix in pairing after that, but the computed MIN-MAX is always larger than the actual data I measure after doing the pairing. I don't know at this point if it is worth going back and redoing the whole process because I did not confirm that setp#2 was fully settled. The tracking as is more than adequate for a 10%-90% SOC cycle. I'm measuring 90% of the amp-hour in a discharge from 3.5V cell voltage.

I hope this is a clear enough explanation. Thanks for the question. It forced me to formalize my gut feelings on this capacity calibration method. I will add this explicit resting balance step to Step #2 prior to disassembling and regrouping. It also helps explains why I saw such a change in the battery voltages while resting. While all of the voltages were rising, they were also converging in the mean. So there are two elements to the resting: balancing and recovery.
 
Last edited:
Steve_s makes the following comments in the post below:


Paralleling Cells.
EV's use Cells in Parallel BUT those are all Matched, Batched & Binned to perform identically with the same Internal Resistance & Impedances throughout their operating range. ONLY Matched cells should be set in parallel to ensure that all the cells are working within their spec and that not a single one is too high or low which result in failures (sometimes really bad fails too).

Most Vendors on Ali* sell Bulk Commodity cells, Grade A or otherwise. Please read the link in my signature regarding ESS Cells. These are NOT suitable for paralleling together. They WILL deviate and unbalance resulting in issues. That's a Virtual Guarantee.

Few Vendors are selling properly Matched, Batched and Binned LFP cells. These all perform very closely together through their voltage & discharge ranges. They do cost a bit more (IE bulk 280's = $92, Matched = $122 USD) BUT you do get what you pay for. Luyuan Tech is ONE of the few who carry Matched & Batched cells.

Hope it Helps, Good Luck
Steve
In the bolded RED comment above he is suggesting that mismatched cells as per my MIN-MAX approach not be done. Not sure why if the next cell voltage is matched to the rest of the series cells.

If we look at an individual MIN-MAX cell pair, under three conditions:

Charging: Presumably the larger capacity cell will rise in voltage slower than the lower capacity cell. This means that the lower capacity cell voltage will at some point exceed the high capacity cell voltage. The charge current to LOWER will have to be reduced to keep the voltages of the pair the same. So while both cells continue to charge, the paralleling should in theory self regulate the currents in proportion to the relative capacities of the pair regardless of internal impedance. I do not think there is any mico cycling unless someone else can enlighten me?

DisCharging: Presumably the larger capacity cell will go down in voltage slower than the lower capacity cell. This means that the lower capacity cell voltage will at some point be below the high capacity cell voltage. The charge current to LOWER will have to be increased to keep the voltages of the pair the same. So while both cells continue to charge, the paralleling should in theory self regulate the currents in proportion to the relative capacities of the pair regardless of internal impedance. I do not think there is any mico cycling unless someone else can enlighten me?

At rest: Presumably, if there is some SOC differential, then there will be a voltage differential and so there will be a net current flow one way or the other. At the high end of the SOC state, the low-capacity cell will generally be at a higher SOC. Likewise, at the low end of the SOC, the low capacity will be at a lower SOC. It has been demonstrated before that there can be wide differences in SOC between parallel cells in the mid-range 3.1-3.3V power cell. Even identical cavity cells can be at 30% SOC differential when in parallel and presumably at the same or similar voltage in this midrange.

Summary:
Given the foregoing analysis, we have to ask what is the difference between paralleling at the battery level vs at the cell level?

If you parallel at the battery level you are stuck with whatever capacity differential you have and are basically relying on the vendor proving matched cells. Basically, they match individual cells that can all be put in series as a single matched cell battery. Somebody else gets that other groups of cells that are themselves matched. If you try and add an additional parallel battery you are not likely to get cells matched to your original set so your reliance on a matched cell approach goes out the window.

If you parallel at the cell level you are getting an automatic capacity weighted current split between the two parallel cells. This split does not involve micro cycling or current reversals, it is a weighted split that only depends on whether charging or discharging.

If I can put a 50 Amp-hr battery in parallel with a 200 amp-hr battery, then what is the problem with putting at the cell level 100+Delta1 amp-hr and 100-Delta2 amp-hour that tends to reduce variance to a 200 Amp-hr capacity in the parallel cells? I don't see any!

Given you have mismatched cells at some level, MIN-MAX pairing cells so they complement each other's capacity seems to be an obvious way to develop internal self-regulating capacity matching within a stacked set of mismatched cells. In fact, Steve_S seems to even contradict his own statements in the same post where widely mismatched cells can be used but not too (100 Amp-hr limit) mismatched. Hopefully, he sees this thread and comes to clarify.


Unlike LEAD, they can be of different capacities. IE 200AH & 280AH sitting side by side, without issues *.
Also, they are NOT age dependent like LEAD. 1, 2 or 5 years apart does not matter, provided the existing packs have NOT been abused & degraded as a result of such abuse.
* Capacities can be different BUT not too far apart, less than 100AH difference is suggested/ The closer the Capacities the better as it is more "balanced".
 
Last edited:
I put together a parallel cell model for some theoretical analysis of the current proportioning between the cells. This can be extended to a series configuration but a single pair is sufficient to see how current sharing might occur as a function of the several variables identified.

The diagram shows that there is a series charging current with a superimposed circular balancing current. The left battery cell is indicated to have a lower capacity and therefore there is a positive current flow to the larger capacity cell. The circular current I 12 tends to equalize the terminal voltages VT1 and VT2. It may not be perfect given the variety of variables involved but it will always tend to keep the terminal voltages closer together than what they would be otherwise without this buss-bared parallel configuration.

The only way that I12 can reverse direction is if the lower capacity cell has a lower SOC than the larger cell. But this is similar to what an active balancer will do as well for series cells. The difference is the buss bar currents I12 can be much higher than 1-5 amps of a typical active balancer. This type of high current can only occur when there is a large imbalance and it is expected that the currents will normally always be low once a parallel bus bar connection is made between cells.

In effect, the parallel cell connections provide a direct implicit active balancer effect (between the pair). It does nothing about balancing voltages between series pairs, and that is the main distinction that an active balancer can provide.ParallelCellModel.png
 
So the Off-grid garage channel on Youtube has several interesting videos. One of the recent ones (dated Jun 10, 2021) is an analysis of parallel strings vs parallel cells.


He discusses a document written and commented about by Orion BMS.


So I post this because it is comforting to find that the paralleling cell is a standard configuration and not likely to create any problem in and of itself. So given the common practice, it makes the capacity balance technique more compelling.




OrionBMS_StandardCellConfiguration_Page4.png
Here is a summary of the primary benefits of paralleling cells. The argument against this is apparently one of being able to better detect anomalies is separate strings (more on this below). In the meantime, this is Orion BMS synopsis on the benefits of paralleling cells.

Therefore, monitoring the voltage of either cell will show the same results (less the very negligible difference in voltage caused by voltage drop on the busbar). In the event that one of the cells develops a reduced capacity or high resistance (as is typical for aged or failed cells), the stronger cell will take more of the load and essentially prop up the weaker cell. In that event, the BMS is able to see a decrease in the overall capacity or an overall increase in resistance. With two cells paralleled together, a single weak cell can affect the resistance up to 50% and the capacity up to 50%. If three cells are paralleled, a single bad cell can affect the resistance and capacity of the total paralleled block up to 33% (with four cells paralleled, up to 25%, and so forth). As more cells are paralleled, a single failure becomes more difficult to detect, but redundancy is also increased since a single cell failure will have less of an impact on the overall performance of the battery. Cells directly paralleled with each other will automatically balance each other since they are permanently connected. Note: While most lithium batteries can be directly paralleled together, check with the cell manufacturer to ensure that the cells can be safely paralleled and to see if there are any specific requirements for the specific cells used. In some cases (such as with some 18650 style cells), cell manufacturers may require individual fuses or fusible link wire to prevent over current through a single cell in the event of a cell failure or an internal short within a cell. Consult with the cell manufacturer to determine if such a design is necessary.

Here is the comment that OGG used to rationalize doing parallel strings. This pretty much confirms he is not an engineer (as I suspected).

OffGridGarageComentJasonEmanual.png

Fault detection is a whole other issue that seems to be the argument against paralleling cells. If you parallel 280 Amp-Hr cells you have 560 amp hr pairs which is a pretty hefty capacity. In this case, a total failure of one cell of a pair will cause that cell to stick out like a sore thumb while not taking down your whole battery cell array.

A 16S2P battery can use a single BMS/balancer combination, as opposed to one BMS/balancer per parallel string. Since the parallel cell configuration provides automatic balancing by the buss barred parallel cells. the cells themselves provide a large part of the cell management that is automatic and you will to only need to provide cell management at each S level. This provides a higher capacity for the whole battery system.
 
Last edited:
Very interesting to read your testing. I'm still unsure which way to go but you might have convinced me to just build one pack. How do the rack type batteries get away with the problems brought up in the Orion BMS article?
 
Very interesting to read your testing. I'm still unsure which way to go but you might have convinced me to just build one pack. How do the rack type batteries get away with the problems brought up in the Orion BMS article?
It should be easy enough to duplicate my results. The more cells you involve in the matching, statistcially the more likely to achive overall better matches.

I don't think rack-type or modular battery manufacturers guarantee any integrated bank performance (e.g. paralleled series cell configurations). I don't know if I was the ORION BMS article or not but I have read about pretty substantial imbalance in charging which will do a lot to reduce the overall integrated performance. I think there are several threads on this forum describing surprising imbalance which can be due to corrosion, wire length, or other things that can cause unequal current sharing and which is exacerbated by the flat voltage to SOC characteristics of LiFePO4.

It should be quite clear that if you have two parallel paths there is always going to be a greater risk of imbalance sharing. If you parallel up the cells in the two parallel strings then you have limited the imbalance because there is now a single string. Even if you have the same potential for oxidized busbars, the parallel bus bards will always reduce the imbalance of the series configurations.

If you couple the procedure described in this thread, you now have the opportunity to also capacity balance each cell in a string and you are far better off under any scenario than the parallel strings. The only possible issue is the need to use the parallel cells against internal shorts. It might be that an internal short within a single cell might not be a hazard, whereas a parallel combination will involve higher currents and the possibility exists that more current could flow from parallel cells. So it might be prudent to put a fuse between each pair of cells that are in parallel.

two in parallel is 1 fuse
three in parallel is 3 fuses
four in parallel is 6 fuses

There do not have to carry the full MBS current but rather what eve the nominal imbalance currents are. Not having measured them, 10% would be an initial guess but very much dependent on your matching.
 
Last edited:
Back
Top