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

[Battery_Mooch]

@ducktaperules

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

Yes

@ducktaperules

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.

Yes, these are single cell run times.

I mention that the run times and energy in each cell are not modeled after any existing cells as the actual timing wasn’t important (as mentioned), only the interactions of the cells.

You’re right, the actual numbers would be different than what I used. I wanted to avoid “real world” numbers for my first post though since, in my experience, it leads to a lot of debate about what numbers to actually use versus concentrating on just the theory the sims were demonstrating.

I am happy to do other sims though.

@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.

The actual amount of power doesn’t really matter except to help hide, or expose, differences in the per-cell current sharing. These simulations were set up only to demonstrate that current-sharing differences exist and should be considered.

But for many setups (most?) these sims seem to indicate that mixing decent, but different, cells isn’t a big issue.

I agree, “typical” use could very well result in much smaller cell-to-cell current sharing differences (depending on how badly the cells were mixed).

@ducktaperules

I would be really interested to see the same separate and parallel simulation performed with some kind of more representative usage profile.

I’m completely open to suggestions that the community could agree on.
This would require creating more accurate battery models though and I do not know when I will have the time to do that. It could take a lot of time.

@ducktaperules

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.

Essentially, if I understand what you are saying, you feel that each acceleration and cruising segment of the ride would roughly emulate what we see in the entire discharge for the simulations I did? Or a good portion of it?

You could definitely be right!

Get some input on what folks here feel is a “typical” ride (power draw variations, timing, pack size, voltage cutoffs, etc.) and maybe a ride that is more extreme too (for comparison).

Pick two different cells to model too. Maybe two sets of cells, one realistic and another set of cells that is a worst case setup? I’m happy to run the new sims when I can finish the battery models.

I’m not sure about modeling regen current right now so assume that can’t be part of it. It’s possible to do but I just can’t spend hundreds of hours on this.

Actually, that’s a lie. I’m happy to spend weeks on it but someone is going to have to pay my bills since I would be giving my day job work to someone else for that time.

[ducktaperules]

@Battery_Mooch

I mention that the run times and energy in each cell are not modeled after any existing cells as the actual timing wasn’t important (as mentioned), only the interactions of the cells.

Right so are we assuming these simulated cells are not accurate in capacity? That would make sense as i couldn’t work out how a 20w load was emptying a ~15 wh cell in 13 mins.

@Battery_Mooch

I wanted to avoid “real world” numbers for my first post though since, in my experience, it leads to a lot of debate about what numbers to actually use versus concentrating on just the theory the sims were demonstrating.

This makes sense but if we were to pick known cells that people might be likely to mix like the P42a and 50e, wouldn’t it be easy to fiddle the cell parameters until the load test results are similar to your previous measured load test graphs? i have no idea how this simulation software works so maybe its not.

@Battery_Mooch

The actual amount of power doesn’t really matter except to help hide, or expose, differences in the per-cell current sharing.

i guess the original question was “can you safely mix cells in a parallel group?” and this work clearly shows that you can at lower discharge levels. My question going forward are how important is this when considering an EV style usage profile?

I realize that coming up with an idealized usage model is going to be tough due to the huge variance in board type and usage scenario, but i feel that it would be interesting to at least speculate on how your current cells would perform in a scenario with a less continuous load. eg. how would the graph look if it was repeated as a 120W for a second, then 10W load for 8 seconds. This would result in short term peak currents of 120w but an average load (and presumably runtime) similar to the 20W simulation. Do we still see the same dangerously high currents in the energy cell towards the end of the discharge curve (like the 120W graph) or do they remain closer to the safe currents seen in the 20W graph?

[Battery_Mooch]

@ducktaperules

Right so are we assuming these simulated cells are not accurate in capacity? That would make sense as i couldn’t work out how a 20w load was emptying a ~15 wh cell in 13 mins.

Correct. Mentioned a couple of time in my original posts.
The model only simulates the current-sharing interactions of these two cells, not the actual run times for real cells. This was done on purpose to force the focus on the cell’s interactions instead of their accuracy in simulating a particular setup.

@ducktaperules

This makes sense but if we were to pick known cells that people might be likely to mix like the P42a and 50e, wouldn’t it be easy to fiddle the cell parameters until the load test results are similar to your previous measured load test graphs? i have no idea how this simulation software works so maybe its not.

That’s how it would be done but it’s not often an easy thing to get right across a wide range of power levels. Well, not quickly that is. A fairly accurate model for a li-ion cell might have a dozen of those series resistor/capacitor strings being paralleled along with other resistors, capacitors, and inductors sprinkled about…depending on what the model was being used for.

If I accurately simulated a particular pack being discharged using a particular profile then everyone would be wondering what would happen to a slew of other pack configurations and usage profiles instead of just concentrating on the fact that cell current sharing can be unequal, can change during a discharge, and probably isn’t an issue for most reasonable packs.

That’s all these first sims were trying to do, just demonstrate that current sharing isn’t equal but probably isn’t a safety issue. We can certainly build on that though.

@ducktaperules

My question going forward are how important is this when considering an EV style usage profile?

I have my predictions but let’s find out!

@ducktaperules

…how would the graph look if it was repeated as a 120W for a second, then 10W load for 8 seconds.

That would result in about a 5,760W burst/480W cruising profile for a 12S/4P pack. One second burst, eight second cruise, repeat until the pack reaches the low voltage cutoff. How do others feel about using this profile?

I would sim the pack running from 50.4V down to 36V. No regen and throttle adjusted to keep the speed constant (constant-power load).

Next decision…what cell mix for the p-groups (having them all the same)? Assuming P42A’s and 50E’s are available, we have five options for the four-cell groups. Use two P42A’s and two 50E’s for each p-group?

[ducktaperules]

@Battery_Mooch

That would result in about a 5,760W burst/480W cruising profile for a 12S/4P pack. One second burst, eight second cruise, repeat until the pack reaches the low voltage cutoff.

Im not suggesting that this would be representative use, just that it might be an interesting test to study the effect of an uneven load cycle. it was chosen purely because the numbers aligned nicely with your other theoretical tests so may be somewhat comparable if run with your current models.

although im now realizing that the 20W load was for the single cell tests so maybe you would want a profile that makes the average 40w, to make it directly comparable with the other mixed cell load tests.

Maybe 120W for 1 second, 30W for 8 seconds. Thats total of 360W in 9 seconds or average 40W/s.

Same peak load as your 120W graph, but same average load as your 40W graph for the mixed 2p scenario.

[Battery_Mooch]

@ducktaperules

Maybe 120W for 1 second, 30W for 8 seconds. Thats total of 360W in 9 seconds or average 40W/s.

Same peak load as your 120W graph, but same average load as your 40W graph for the mixed 2p scenario.

I like it, makes sense. Let’s see what happens.

[Arzamenable]

Those magical early COVID runs may or may not have been synced.

Serineh rides this board sometimes. We are quiet neighbors now. We’re just like any other evolve rider. We want to get low and visit wobble city.

(But seriously, the 911 v. 44 rematch is coming)

[BenjaminF]

@Battery_Mooch

Let’s see what happens.

Watching intently. This has been an amazing read so far, I love learning this stuff

[glyphiks]

@Battery_Mooch in this circumstance, is it any different to mix these cells together in individual p-groups making one battery, to paralleling two different complete batteries together?

[Battery_Mooch]

In the end I don’t think it would make much of a difference. Mixing p-groups would result in cell level differences in current and paralleling different packs would result in pack level current differences.

Hmm…it might make a big difference to the BMS’ though.

Worth simulating once I have the more accurate models done for the previously mentioned sims.

[glyphiks]

@Battery_Mooch

BMS

That’s only a problem if you are using a discharge bms though right?

[Battery_Mooch]

There’s a good chance that maybe, perhaps, possibly that it potentially, perhaps, might only be a problem if using a discharge BMS.

All levity aside…yea, you’re probably right. I can’t think of anything that a charge-only unit would get zapped with as long as each pack had its own BMS (if paralleling packs). Might spend a lot more time doing its job though.

[glyphiks]

Good thing I don’t use a BMS then

[emaxon]

I’d really like to revisit this topic before I do something stupid (again.)

It took me a year, and >1k mi to realize that the pack I built for my mountainboard, out of Model 3 “Energy” cells, couldn’t come close to properly feeding the 6374 motors I used. The reason I’ve never seen more than a 60A draw from my pack, which once suggested to me that your pack doesn’t need to be able to feed max motor amps x2, is because that’s all the cells are capable of delivering!

Since I built it using Batteryblocs, I’m free to reconfigure at will.

Here’s the burning question: If I replace one or two 10A cells in each 6P group (It’s a 10S pack now, will be 12S after the rebuild) with a P42A (45A max draw) to increase ampacity from 60A to (in theory) 50A + 45A=95A or 40A+90A=130A, is this going to help keep temps of the Energy cells down, or am I just shooting myself in the foot?

Per the above, I’m not worried about what happens under constant load, but rather what happens when I let off after skating hard for a while, or if I maintain the heavy load long enough to deplete the one power cell. If the current draw from each cell is proportional to the internal resistance or ampacity, I’d expect the power cells to be drawn down more easily, then for them to draw current from the energy cells when I’m coasting, similar to the way my range extenders charge the internal pack when I’m coasting. What I’m worried about is the voltage differential growing the longer I’m on the throttle, to the point where when I finally let-off, the energy cells will dump all 40-50A into the power cells to equalize them, vastly exceeding the 4.2A recommended max charging rate per cell.

Aside from the simulations above, has anybody (here) actually tried mixing power + energy cells within the same pack, have something to add?

Since the experiment is so easy, I may try it anyway, just like to hear some BTDT first.

[Soflo]

I think you are over rating the p42a.
In a 6p configuration if you replace 1 cell, you’ll need to add the amps of all 6 cells together then divide by 6. By adding 1 cell to a 6p you do not magically get an extra 45a from that p group,

[Battery_Mooch]

@emaxon

Since the experiment is so easy, I may try it anyway, just like to hear some BTDT first.

I’ve done some actual testing but you’ll have to run your own experiment to know for sure. Trying to guess what your results might be from tests I’ve run with different cells and conditions can’t be done.

Paralleling cells like you outlined can work. But as you mentioned it depends on how hard you use them and the duty cycle of your high current draw.

If you rate the Model 3 cells at 10A then rate the P42A’s at 30A to provide the same level of stress to each. Replacing one Model 3 cell with a P42A won’t increase your current handling by that theoretical 20A IMO as the current sharing will affect things. Too much to think through, brain hurts.

Run the experiment with two cells and measure the current from each cell during low and high discharge current levels. It’s the only way to know for sure.

[emaxon]

Soflo, I think we’re both mostly wrong… If I followed your math, I’d only be able to draw 15.83A out of the whole pack. ((5x10A + 45A)/6 = 15.8333A) I realize I won’t be drawing 10A from each Mod3, nor 45A from the lone P42A, but I expect to draw “somewhat” under those 2 figures, and must be able to draw more than 60A from the whole group of 6 cells. Sounds like the 30A approximation below is useful. Bottom line is that cell mixing is an advanced form self-flagellation. I’ve decided to practice battery eugenics, stick with power cells for my main packs, play games with disposable range-extenders only.

[sleepless]

@Battery_Mooch

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?

@rusins

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.

@Battery_Mooch

In the end I don’t think it would make much of a difference. Mixing p-groups would result in cell level differences in current and paralleling different packs would result in pack level current differences.

Hmm…it might make a big difference to the BMS’ though.

Worth simulating once I have the more accurate models done for the previously mentioned sims.

Mooch, would similar lessons from this experiment apply if paralleling two packs of the same cells, but different size, to extend range? i.e. a 12s2p in parallel with a 12s3p (of all brand new P42A if we want to get real specific)

I’m out of my element here, but I’m thinking this is somewhat analogous to a large capacity cell in parallel with a medium capacity cell, of similar discharge curves.

[I ask totally selfishly bc I’m planning on doing this. I was convinced it was safe and effective at one pt. But I’m having second thoughts]

details

NESE modular jumper - 🐉 Dragon MTB build

The plan is to use 60 fresh P42A cells in NESE modules, and wire the two 12s packs in parallel to the ESC. No BMS onboard - Daly 12S BMS will be mounted to the 6A YZPower charger and the two packs will be charged individually. No charge port - everything will use XT90s connectors - must disconnect pack from ESC to charge, but that’s easy with the Nanuk case.

(Obviously, packs will only be plugged in together when both are charged to 100% on the same charger. The range pack would not be added on the fly)

Pack config like this, with ESC permanently mounted under the range pack, bolted to an aluminum enclosure floor plate, which is then mounted to the deck. Each pack has its own balance harness:

[Battery_Mooch]

@sleepless

Mooch, would similar lessons from this experiment apply if paralleling two packs of the same cells, but different size, to extend range? i.e. a 12s2p in parallel with a 12s3p (of all brand new P42A if we want to get real specific)

The different sizes wouldn’t affect anything. The condition of the cells and interconnects would though.

Directly paralleling different p-groups just makes a bigger p-group. How the cells interact, the levels of inter-cell current flow as they are discharged and charged, depends on how close all the cells in that group are matched. If they are wildly different then you will have a lot of current flow within the p-group as they all constantly try to balance each other out. This creates more heat and ages the cells faster.

This is how any p-group works though, whether all in one pack or two groups of cells paralleled together.

I think the safety wouldn’t be directly affected any more than any other setup since a bad cell, one that was about to fail, would be an issue no matter how it was used in a pack.

@sleepless

I’m out of my element here, but I’m thinking this is somewhat analogous to a large capacity cell in parallel with a medium capacity cell, of similar discharge curves.

Yes, but subsequently acting as one large cell. The same as any p-group.

@sleepless

[I ask totally selfishly bc I’m planning on doing this. I was convinced it was safe and effective at one pt. But I’m having second thoughts]

Safe? Absolutely not. There is no such thing as “safe” use of any li-ion cell. There is always some risk. However, we can considerably reduce that risk by using good pack building practices and having a good BMS in place.

All this assumes that the packs are paralleled at the p-group level and not just at each pack’s main pos/neg leads. Joining the main leads together parallels the packs but only as independent packs. Any differences in the packs will result in current flow between the packs as they try to balance each other out. The p-groups from each pack will not be paralleled and you will need two BMS’.

Each pack’s p-group needs to be connected directly to the other pack’s matching voltage p-group to actually parallel the packs completely and to be able to use a single BMS.

[sleepless]

@Battery_Mooch

Joining the main leads together parallels the packs but only as independent packs. Any differences in the packs will result in current flow between the packs as they try to balance each other out. The p-groups from each pack will not be paralleled and you will need two BMS’.

Sorry, I should have been more clear - I’m specifically talking about paralleling two independent packs by joining the main leads together. This electrical connection would only happen when both packs are 100% charged. No BMS on the discharge. Packs are individually charged with their own BMS.

Mechanically / functionally, this is much easier to achieve (less wiring) than actually connecting each p-group’s pos & neg to the second pack to form one single cohesive battery.
Similar to paralleling slightly different capacity 12s lipo pouches.

[Battery_Mooch]

@sleepless

Sorry, I should have been more clear - I’m specifically talking about paralleling two independent packs by joining the main leads together. This electrical connection would only happen when both packs are 100% charged. No BMS on the discharge. Packs are individually charged with their own BMS.

Ahhh….okay.

@sleepless

Mechanically / functionally, this is much easier to achieve (less wiring) than actually connecting each p-group’s pos & neg to the second pack to form one single cohesive battery.
Similar to paralleling slightly different capacity 12s lipo pouches.

Hugely easier.
But it results in two packs that will “go their own way” and each will need to be monitored and separately balanced as necessary.

The huge mismatch in pack capacities means that there will be a lot more current flow from one pack than the other as the discharge proceeds and the higher voltage pack tries to keep the lower voltage pack charged to the same level (same voltage).

They will indeed act the same way, conceptually, as the individual cells I used in my simulations. As long as neither pack ends up trying to supply too much current, the cells are in decent condition, and they are kept balanced when charged, it seems you should be okay paralleling them like that.

But I cannot guarantee anything as I have no idea of the actual condition of your cells and packs. You’ll have to carefully test the setup at a low current level first, checking voltages/temperatures/current levels, to make sure the current being supplied by each pack is within expected levels.

Naturally, I am going to recommend only paralleling same-capacity packs that have cells of the same age and to use a BMS for each pack. Good luck though!