Here’s what can happen when you parallel (mix) different cells in a pack

[Battery_Mooch]

Why use different cells in a pack? Some want to extend the range of their packs and they don’t have more of the same cells. Others only have a mixed bunch of cells to build a pack with, perhaps salvaged from other battery packs.

There are firm opinions on whether this mixing of different cells is dangerous or not. To add a little bit of data to the discussion I ran some computer simulations to show what can happen when you parallel two different models of li-ion cells.

TL;DR…Can we build a pack using different cells or add cells to a pack to extend range without causing a lot of safety issues? In my opinion it’s possible. As long as we’re not being ridiculous about it though.

Different cells will not equally share the current being drawn from the pack. These simulations show how different cells can share current differently at different times during a discharge. You will need to take this into account when paralleling different cells. Make sure you do not discharge any of the cells too hard.

THESE SIMULATIONS DO NOT TELL YOU HOW YOUR PARTICULAR PACK WILL BEHAVE WITH MIXED CELLS!

This is true even if you figure out which cells I have very roughly simulated. These simulations IMO just tell us that good cells that aren’t wildly different can probably be mixed in a pack if we aren’t running them at near their current ratings.

But if someone asks me if it’s okay for them to mix different cells in a pack I’ll say no. Why? Because I have no idea what the condition of their cells is. I am forced to err on the side of caution and to recommend not doing it. Each of us can make our own decision though. These posts are only to help give you a little more information. They are NOT some kind of approval to go ahead and mix cells in your pack.

Read post #2 for more info.
Read posts #3 and #4 for the results of the simulations.
Read post #5 for info on the battery models used for the simulations.

[Battery_Mooch]

Overview:
The greater the difference in capacity and internal resistance for the cells you are connecting the greater the difference in how much current each will supply when they are paralleled. This is also true as the power level of the discharge is increased.

A pack using the same model cell, all from the same batch and purchased at the same time from the same vendor, will only have slight differences between the current shared by each cell. You might not even need a BMS to keep them balanced until you have cycled them a lot.

A pack that mixes very different cells though can result in a big difference in how much current each cell has to supply. So while you can often mix cells that aren’t too different in a pack I recommend not doing so unless you are well below the current ratings for all the cells and aren’t discharging the pack to a low cutoff voltage (where the differences between cells can have the biggest effect).

Don’t do something like trying to add 5A-rated ultra-high capacity cells to a pack using 30A-rated low capacity cells to try to increase your range. You cannot assume that adding two 5A cells to a p-group using three 30A cells has a new total rating of 5A+5A+30A+30A+30A = 100A. The current rating for that p-group will probably have to be a lot lower in order to keep the 5A-rated cells from supplying more than 5A each.

Can paralleling different cells be dangerous? Yes, if you insist on not paying attention to current ratings, use old or damaged cells, build a crappy pack, or misuse or mishandle the cells or pack, then it can certainly be dangerous…like any pack would be.

But paralleling good “power” cells like the high current rated Molicel P42A’s with good “energy” cells, like the high capacity Samsung 50G or Molicel M50A (or even the so-so Samsung 50E or Tesla Model 3), shouldn’t result in a pack that is unsafe to use. It can definitely lead to reduced performance though.

Adding energy cells to a power cell pack means you won’t get the higher current rating you would have gotten from adding more of the same power cells. Adding power cells to an energy cell pack means you won’t get the extra run time you would have gotten from using more of the same energy cells.

In both cases though your range would be extended since adding any cells means more run time is available. You just have to be aware of the current ratings for the cells and how they might divide the pack current.

As always this assumes the packs are well constructed and the cells are discharged and charged within their ratings and not when hot or cold, etc. Never misuse or mishandle any cells or packs. We use these cells at our own risk and mixing cells in a pack increases the risks, be nice to them.

But if the cells are in good condition and not nearing the end of their life then there shouldn’t be any big danger in paralleling different cells if you do not exceed their current ratings.

The Cells:
The “power” cell for the simulations is a good performing, low internal resistance cell rated at about 25A. The “energy” cell is also good performing cell, with about 30% higher internal resistance and about 15% more capacity and rated at about 10A.

The power cell has just over 4200 units of energy storage available in these simulations. The energy cell has just over 5000 units of energy storage available.

The cell internal resistances and the differences in run times are roughly based on two popular cells used by the esk8 community. The voltage sag and cell interactions when paralleled should be pretty accurate but I have made no attempt to accurately model the actual run times so capacity numbers mean nothing.

These are not accurate simulations of actual cells for use in estimating what your packs will do. These simulations are only to demonstrate what can happen when different cells are paralleled, to show how they interact versus when operating alone.

The Simulations:
I ran two sets of computer simulations to show what happens when two different cells are used separately and when paralleled at different power levels.

As a baseline for comparison I did three simulations for each cell when used alone at different power levels, 20W, 40W, and 60W, from 4.2V down to 3.0V. These wattages simulate use of a battery pack by someone who keeps bumping the throttle up as the pack voltage drops in order to keep the speed constant.

This is what is known as a “constant-power” discharge and results in more and more current being drawn from the pack as the voltage drops in order to make the same amount of power at the motors. It can result in a lot of heat being created in the pack at lower voltages since internal resistance goes up as the pack nears empty and the current level can increase significantly at lower voltages too.

I then paralleled the two cells and did three simulations at what would be the same power levels each if the cells divided the power equally, 40W, 80W, and 120W.

In the next two posts I describe what happens for each discharge above each graph. The power cell is always the red plot line and the energy cell is always the green plot line. The blue plot line in graphs #4-6 shows the voltage of the paralleled cells.

[Battery_Mooch]

In these simulations each cell was discharged separately to show how they perform individually at 20W, 40W, and 60W.

Graph 1: Each cell discharged separately at 20W
Power cell supplies about 5.10A at the start to 7.04A at the end.
Energy cell supplies about 5.15A at the start to 6.76A at the end.

The voltage plots show the power cell (in red) running at a higher voltage at the start. This is due to its lower internal resistance. At almost halfway through the discharge the energy cell (in green) starts to run at a higher voltage than the power cell. This is because of the higher capacity of the energy cell. Eventually the power cell can no longer hold its voltage up due to its lower capacity.

Looking at the start of the current plots we can see that the power cell doesn’t need to supply as much current as the energy cell. The power cell’s higher voltage means that less current is needed to make the 20W of power. As the two cells discharge though, and the power cell’s voltage drops below the energy cell’s voltage, we can see that the power cell ends up needing to supply more current than the energy cell.

This is a classic example of how a higher capacity energy cell can be the better choice at low power levels, typically under 75% of its current rating. At these low power levels the high internal resistance of an energy cell doesn’t cause a lot of voltage sag and the higher capacity of the energy cell results in longer run time. This can be seen in the lower right hand corner. In this simulation the energy cell ran for about 775 seconds and the power cell for about 720 seconds.

Note that this run time difference is only about half of what you might expect when just looking at the energy storage for the two cells. The energy cell stored about 15% more energy (“capacity”) but only ran for about 7% longer.

Two things caused this. One, any discharge above the less-than-one-amp discharge used when determining the cell’s capacity will result in less efficient operation, i.e., the delivered capacity will be lower. Two, the cutoff is set to 3V per cell and you typically won’t see most of the differences in run times between two cells of different capacity cells until you get down to lower voltages.

Graph 2: Each cell discharged separately at 40W
Power cell supplies about 10.3A at the start to 13.55A at the end.
Energy cell supplies about 10.6A at the start to 13.95A at the end.

Looking at the voltage plots we can see the higher voltage the power cell runs at for the entire discharge compared to the energy cell. At this power levels the high internal resistance of the energy cell results in a lot of voltage sag and the power cell becomes the clear winner.

Note that this is a discharge above the approximately 10A rating of the energy cell. Using any ultra-high capacity cell at much above about 75% of its rating means a lot of voltage sag and a huge loss in run time.

In this simulation the power cell ran for longer, about 260 seconds.
The energy cell ran for about 240 seconds.

Graph 3: Each cell discharged separately at 60W
Power cell supplies about 16.0A at the start to 20.3A at the end.
Energy cell supplies about 17.0A at the start to 22.7A at the end.

The difference in voltage while being discharged, and the run times, for the two cells is now huge. The power cell ran for about 116 seconds and the energy cell ran for about 69 seconds.

The energy cell could be used at a power level this high but that would result in truly terrible performance and significantly speed up the aging of the cell. Just because a cell doesn’t explode when being used hard doesn’t mean it’s a good idea.

[Battery_Mooch]

In these simulations the two cells are connected in parallel to show how they perform together at 40W, 80W, and 120W. This would be 20W, 40W, and 60W per cell if they divided the power equally. This shows how two different cells interact in different ways that you might think.

Graph 4: Different cells in parallel at 40W
Power cell supplies about 5.65A at the start to 6.45A at the end.
Energy cell supplies about 4.25A at the start to 7.35A at the end.

The power cell supplies more current at the start than the energy cell and less current at the end. This is the opposite of what happens when they are discharged separately.

In this graph the two cells would be running at the same 20W power level used for Graph 1 if they divided the current equally and ran at the same voltage. However, these two cells have different internal resistances and different capacities and that leads to some interesting behavior when they are run in parallel.

Since these cells are paralleled they both run at the same voltage. That is shown by the blue plot line.

At the start of a discharge it takes a li-ion cell up to a few seconds for the ion flow to settle in the different places and densities it stay at for the rest of the discharge. During this time the cell internal resistance and cell voltage changes and you can see this affecting the current sharing for these cells at the start of the discharge.

The low internal resistance of the power cell means it will supply more current at the start than the energy cell. But as the discharge proceeds the lower capacity power cell will try to drop to a lower voltage than the energy cell. The energy cell will then have to compensate for this by increasing the amount of current it supplies. This results in the two cell voltages always remaining equal.

But at this low power level there’s never more than a 2A difference between what these two cells each supply. This difference in the current each supplies will change for different cells, different age cells, and for different power levels.

After a bit more than 200 seconds you can see that the energy cell begins to supply more current than than the power cell and continues to do so for the rest of the discharge.

The change in current levels from start to finish is not large when discharging at power levels this low and even cells (in good condition) with bigger differences between their internal resistance and capacity could be paralleled.

The shapes of the current plot lines will change for different cells and for discharges at different power levels but this switchover between the energy cell and power cell will still happen each time. At very high power levels it might not happen before the cells drop to the cutoff voltage though.

The run times for separate 20W discharges were about 775sec and 720sec, averaging about 747sec. The run time for these paralleled cells at 40W (20W each cell if divided equally) is about 740sec, essentially equal to the average of the two cells when run separately.

For low power discharges ONLY this means you can just add up the capacities/run times for the cells you want to combine and get a good idea of the new total capacity/run time for the pack you’re adding cells to. This might not be possible for higher power discharges where imbalances between the cells are magnified.

Graph 5: Different cells in parallel at 80W
Power cell supplies about 11.7A at the start to 13.6A at the end.
Energy cell supplies about 9A at the start to 13.95A at the end.

When discharged separately the power cell supplied less current at the start, the opposite of what happens here.

At this higher power level you can clearly see the larger spread between how much each current each cell supplies compared to the 40W parallel discharge in Graph 4.

You can also see that the lower internal resistance power cell again supplies more current until its lower capacity forces the energy cell to supply more and more current as the discharge continues. Eventually the energy cells ends up supplying more current than the power cell.

Graph 6: Different cells in parallel at 120W
Power cell supplies about 18.7A at the start to 21.5A at the end.
Energy cell supplies about 14.1A at the start to 19.2A at the end.

When discharged separately the power cell supplied less current at both the start and the end, the opposite of what happens here.

As the power level of the discharge goes up the current sharing between the two cells gets worse and worse. The differences in internal resistance and capacity gets magnified as the current level gets higher and higher. This is why we need to try to keep a pack using different cells from being used too hard. It becomes easier to damage the pack, or worse, as the discharge current increases.

For this entire discharge the power cell supplies more current than the less efficient energy cell. If we let the discharge continue down to a lower cell voltage the energy cell would eventually start to supply more current though. You can see the two current plot lines starting to converge as the discharge proceeds.

[Battery_Mooch]

The Battery Models:
I used LTSpice IV (https://www.analog.com/en/design-center/design-tools-and-calculators/ltspice-simulator.html) for these simulations. It has a few quirks (okay, a LOT of quirks) but is a great way to simulate all sorts of circuits.

The models for the two cells are shown below. These are simplified as we don’t need to model all of the smaller electrochemical actions of discharging cells. One cell is the power cell and the other is the energy cell. The components for each use “power” or “energy” in their names to represent which cell they are part of.

For graphs #1-3 the cells are discharged as shown below.
For graphs #4-6 one of the loads is deleted and the two cell circuits are connected in parallel.

Each cell consists of paralleled sets of a series-connected resistor/capacitor to represent the energy stored in the cell. Those paralleled sets are in series with another resistor that represents the materials in the cell. Each cell also has a 1mOhm resistor at the positive and negative ends to represent the wiring/strip connected to each cell.

Rbulk/Cbulk = the main energy storage capacitor along with its equivalent internal resistance. This equivalent resistance causes some of the voltage sag seen when discharging a cell.

Rsurface/Csurface = the much smaller energy storage capacitor, along with its equivalent internal resistance, that represents the “surface” charge inside a cell.

Rohmic = the resistance of the materials inside the cell. This is the metal foils, metal tabs to the top and bottom contacts, resistance of the electrolyte liquid, etc.

Rpos/Rneg = the resistance of the wiring or strip connected to the cell.

V_PowerCell and V_EnergyCell = the points where the voltage of each cell as would be measured by us at the positive and negative contacts of each cell.

POWER_LOAD and ENERGY_LOAD = the constant power loads that simulate a motor with the throttle being increased as the cell voltage drops in order to keep the motor power steady.

The text at the bottom of the image sets up the simulation and tells it how long to run.

These models can definitely be made more accurate but for the simulations I wanted to run they are good enough. Feel free to copy these models and tweak them for whatever simulations you might want to do.

[Pedrodemio]

Killer analysis @Battery_Mooch

I have to say the results weren’t what I intuitively expected, maybe because I’ve always imagined that done with energy cells from the past, that had way higher internal resistance than what modern high energy cells have. Something like a internal resistance 2 to 3 times what a power cell had

[rusins]

Fascinating read! I suppose this also applies to mixing new and old cells in a pack, or using a new pack in parallel to an old existing pack.

I guess if someone really wanted an average pack by combining power cells and energy cells, they’d need a separate circuit that split the load correctly / charged one pack with the other pack, or something like that.

Thanks for doing the simulations! Shame the combined packs always under-performed

[Battery_Mooch]

Some energy cells do have triple the IR of good power cells, even higher. For these simulations I used the approximate IR’s for two cells I’ve seen others mention they were thinking about paralleling.

The greater the difference in IR, the greater the difference in current sharing. It might be worth doing another couple of sims using cells with vastly different IR’s. But I think most of us already would know that something like that could be a much bigger issue than using two common cells with IR’s that aren’t hugely different.

[ShutterShock]

This is really cool. Good read and it adds a lot more understanding to the whole discussion of why people wouldn’t want to mix different types of cells. It’s really complex to understand but at least it shines light on the idea that we can keep using the same type of cells and still get great performance.

[jonisingt35r]

If calls are below 40ish ohms then they are generally high current output, above that then current output goes down considerably.

[b264]

milliohms

[jonisingt35r]

Ohms

[Battery_Mooch]

Most cells with IR’s in the 40 ohm range are made from fruits or vegetables.

[jonisingt35r]

Lots of info here if anybody wants to learn about batterys and a really good calculator to make battery packs, and a data base of cells

Second Life Storage

[ducktaperules]

@Battery_Mooch thanks for this detailed write up, its incredibly interesting.

The graphs you have produced are very useful however i wonder how these constant discharge simulations relate to esk8.

@Battery_Mooch

In your low power graphs your seeing run times of <800s (about 13 mins). Im thinking that people with larger more powerful boards are usually running more P cells in their battery packs. Overall, regardless of type or size of board, I have not seen many boards that have less than an hour of ride time. This is over 4 x the run time time of your lowest power draw estimation, I guess this means that the average power draw for most boards would be less than 5W per cell right?

@Battery_Mooch

likewise for your 120W graph, lets apply this to average eskate. 120W per cell in a 12s pack means your using about 1.4Kw from a single p group. lets assume that most here are running at least a 4P pack on an AT board (for example), This means that you would be pulling 5.7KW for the full minuet and a half of the test. I have never seen any logs using this much power for more than a few seconds. For reference 5.7kW is enough power to accelerate a 100kg (board and rider) to 30mph in around 3.5 seconds and up to 60mph in around 12 seconds. 1.5 mins at this power level would be pure insanity.

These high discharge graphs seem to be showing that at the beginning of higher discharge tests the power cell is delivering much more current than The energy cell and this gets worse as the power cell begins to loose remaining capacity with respect to the energy cell. I would argue that this effect will be far less prominent in estate usage as these short pulses of high discharge are separated by large time periods of low discharge allowing the cells to re balance.

I would be really interested to see the same separate and parallel simulation performed with some kind of more representative usage profile. Im not sure how you simulate a “representative ride” but i think you could get closer by mixing different loads for set times, making sure the average power usage is still quite low. Maybe this would include some short pulse of high load (+50W for 3s, throttle punch) some longer pulses of medium load ( 20W for 120s, long hill) lots of low load cruising (maybe less than 5w) and some no load coasting (values per cell).

My hypothesis would be that under the high load the power cell will give more amps for a few seconds but then as the load drops again the power cells current will drop below the energy cells current until the cells come back to a point of balance again.

@Battery_Mooch is it possible to run a simulation with a more complicated usage profile like this?

[b264]

@ducktaperules

This means that you would be pulling 5.7KW for the full minuet and a half of the test. I have never seen any logs using this much power for more than a few seconds.

That’s probably because if you pull 5.7kW for more than a few seconds you’ll reach orbit in no time at all

LoL your edit adds basically this same notion

[ducktaperules]

5.7kW for 5 seconds puts a 100kg board and rider at 40mph

And because im lazy this is also assuming rider has the rolling resistance and air Resistance of a car

[b264]

40mph ≈ 64km/h

[rusins]

@ducktaperules

I have not seen many boards that have less than an hour of ride time

sweats profusely

[glyphiks]

@ducktaperules

I have never seen any logs using this much power for more than a few seconds.

@Arzamenable you know what you need to do brother