Combining Batteries Of Different Discharge Rates

[The Mighty Volt]

What happens if we combine batteries of the same chemistry, the same dimensions, the same voltage and capacity......but different discharge rates?

Let's say....50 A123 18650 batteries and 50 "generic" 18650 LiFePo4 cells..........100 cells, mixed up, different cells in different strips of the battery, operating under stress.

What happens?

 

[dogman dan]

I was asking the same question about lipo a few days ago. As I understood the replies, it would depend on the c rate of the discharge. What I planned was 30c mixed with 15c , discharged at 5c. No problem there. But in your case, if the generic 18650's were being discharged at max c rate, then there would be a problem with those cells. Being paralelled with better cells won't improve the weak ones. But paralell them with enough other cells, and the discharge rate of each cell goes down lessening the chances of problems.

 

[vanilla ice]

Like dog says if you series low and highC packs together it does nothing for your maxC rate. You are still limited by the lowC cells, weak link in the series chain..

If you parallel highC with lowC you might be ok. On throttle would presumably be ok as your lowC cells would Vsag out and the stiffer highC cell would pickup the slack. My biggest worry would be- you could get some strange cross currents happening when you let off the throttle. If I were betting on it, I'd bet it would be ok though..

 

[busted_bike]

Funny, I was wondering the same thing recently - basically thinking about optimizing cost/energy/power/price/safety/weight by combining different types of cells (even different chemistries) in parallel.

For example, I'm thinking of putting together a relatively small pack (36V*5Ah) for my bike. I'd like to use LiMn for various reasons, but my controller has a fixed 22A current limit so discharge rate would be a problem. So I was thinking of paralleling 4Ah of LiMn with 1Ah of Lipo and treating them as a ~4C pack, charging at low rate and counting on the internal resistance of the cells to get them to share current appropriately - basically used for acceleration out of stop lights and as a hill assist, so bursts of 1-2 minutes at relatively low duty cycle. But it's not clear to me how long the Lipo would hold up before stressing the LiMn.

Has anyone tried / characterized this?

 

[The Mighty Volt]

I intend trying it...

I think the guys have answered the question: it makes no difference if the actual discharge rate fails to exceed the discharge rate of the lowest rated cell.

I think that a big enough Ah pack coupled with the correct controller would fix the issue.

In case anyone is interested, the only reason I ask is I have some A123 and a lot of generic 18650's....and I want to use them all.

Cheers.

 

[cell_man]

I'd like to try this myself. I would have thought you could exceed the max current of the lower C cells, as the high C cells would just supply most of the current in this situation and the low C cells would just supply the current that they are able to, at that particular voltage level. My concern is not what happens when under load, it's what happens when the load is removed.

If you have a scenario where you have a large low C group of cells with a smaller high C group of cells, the high C cells will get more deeply discharged quite quickly when under a heavy load. If the no load voltage drops down low enough on the small group of high C cells, the larger low C pack will try to dump lots of current into the small high C pack. I'd be worried about how much current was being moved between the cell groups, although I think it could be ok and if the cell groups were roughly equal Ah, that would likely reduce possible issues IMO. Another possible problem I could envisage is that if you are using the mixed pack hard it will have little voltage sag initially but the voltage sag increases when the high C cells become significantly discharged. With a proper control system and isolation between the packs, I think it could be really effective for larger EV applications, but I've never heard of it being done in such a way.

The best thing is to just try it and monitor the current flowing between the cells. I do remember a guy complaining that he bought bad A123 cells as they didn't last well, but he was running them in parallel with lower C cells, just like this

 

[busted_bike]

My concern is not what happens when under load, it's what happens when the load is removed.

If you have a scenario where you have a large low C group of cells with a smaller high C group of cells, the high C cells will get more deeply discharged quite quickly when under a heavy load. If the no load voltage drops down low enough on the small group of high C cells, the larger low C pack will try to dump lots of current into the small high C pack. I'd be worried about how much current was being moved between the cell groups, although I think it could be ok and if the cell groups were roughly equal Ah, that would likely reduce possible issues IMO.

When the load is removed, I would expect the initial response to be that the low C cells would continue supplying current at the same rate at they were under load. Only difference is that the current path would be into the high C cells to equalize the state of charge, rather than being delivered to a load. The current from the low C cells certainly wouldn't increase with reduced load... at least not in any scenario that I can imagine.

 

[The Mighty Volt]

Is there a fire risk or anything?

 

[scoot]

I tried someting like this once, but using SLAs and LiFePO4 packs, but not with entirely naked cells.

My scooter had 2s2p HR9-12 SLAs (24v).

I paralled them with a 24v Headway 10ah (8s1p) pack with its own BMS rated at only 20 amps continuous. It was connected on the opposite side of my CA shunt as the SLAs.

I was running on hills where I was typically drawing about 60 amps or so rather coninuously before this test. So during the test run I was seeing about a +40 amp draw through the CA at full throttle (i'm sure the Headway pack was contributing the other 20 invisible amps to make up the difference).

All was good for the first half of the ride... every time I coasted I could see the CA showing about -2 to -3 amps as it was essentially recharging my SLAs which had been carrying the brunt of the load during the ride. So just like having an onboard charger for my SLAs... sweeet . But later in the ride as my SLAs got further depleted, the CA was was showing about -4 to -5 amps during a coast or stop . I figured that was bad so I ended up returning that Headway pack.

I always thought afterwards that if that Headway BMS been able to deliver about 30 amps continuous (another 10 amps), it may have worked out to be a nice match. Of course, it would have needed to used in such a way where it was not left connected to the SLAs when not in use during a ride. I feared it would have cooked the SLA pack which is at a slightly lower pack volatge than the Headway pack when they are both full and at rest :? . It was a fun and educational experiment for a newb like me. 8)

 

[cell_man]

When the load is removed, I would expect the initial response to be that the low C cells would continue supplying current at the same rate at they were under load. Only difference is that the current path would be into the high C cells to equalize the state of charge, rather than being delivered to a load. The current from the low C cells certainly wouldn't increase with reduced load... at least not in any scenario that I can imagine.

I don't know that they would necessarily try to instantly equalize the SOC, but for sure there would be current flow from the low C cells to the high C cells. I would have thought that the cells would just try to equalize voltage and current would flow to the point with the lowest potential, the high C cells in this case. I can't see how the pack voltage would remain at the voltage it was whilst under load, it would simply bounce back to the resting voltage somewhere between the resting voltage of the high C cells with a lower SOC and hence lower resting voltage and the low C cells with a higher SOC and therefore higher resting voltage. The resulting voltage would be a result of the impedance of the cell groups and the resting voltage of the cells determined by the SOC. The issue is further complicated by the fact that A123 claims a 3.3V nominal as opposed to 3.2V which is the standard for typical LIFePO4, but this could be just due to the low Ri and hence higher Voltage they maintain at the standard 0.5C that these values are based upon, more than any thing else.

Is there a fire risk or anything?

I doubt it, but I wouldn't rule it out :lol: With a few fairly simple checks I think you could work out the likely situations. For example, A123 26650 can accept 10A continuous charging (4.5C). If you apply 10A charge what does the terminal voltages rise to at various SOCs. For the low C pack, the surface charge will quickly go and it will fall into the 3.3-3.4V range very quickly. Check a few worst case scenarios and see what happens. It doesn't take much to pull a low C cell to 3V or less for example, so I can't see an issue. It would be interesting to do some calcs and later try it and see what happens with current meters between the cell groups, at the cell level. If the high C cell group had a much lower capacity than the low C group I think this could possibly be more problematic. However, having said all that I have no experience of actually trying this, so there could be a whole load of other things going on that I never even considered :?

 

[busted_bike]

I don't know that they would necessarily try to instantly equalize the SOC, but for sure there would be current flow from the low C cells to the high C cells. I would have thought that the cells would just try to equalize voltage and current would flow to the point with the lowest potential, the high C cells in this case. I can't see how the pack voltage would remain at the voltage it was whilst under load, it would simply bounce back to the resting voltage somewhere between the resting voltage of the high C cells with a lower SOC and hence lower resting voltage and the low C cells with a higher SOC and therefore higher resting voltage. The resulting voltage would be a result of the impedance of the cell groups and the resting voltage of the cells determined by the SOC. The issue is further complicated by the fact that A123 claims a 3.3V nominal as opposed to 3.2V which is the standard for typical LIFePO4, but this could be just due to the low Ri and hence higher Voltage they maintain at the standard 0.5C that these values are based upon, more than any thing else.

Yup. We're saying the same thing in different ways. My point was that the cross-conduction draw from the low-C cells when the load is removed wouldn't be any worse than the current draw under load. Just need to ensure that the pulse width and duty cycle of the load as seen by the pack is light enough so as to allow the SOC to (mostly) equalize during unloaded or lightly loaded segments to prevent the low-C cells from being overstressed by high currents.

What's not clear to me is the max pulse duration/amplitude... which I suppose would depend a lot on the specific cells used. Hrmmm, I think I just convinced myself that the only way to answer the question is to try it since batteries don't come with spice models. :wink:

The issue is further complicated by the fact that A123 claims a 3.3V nominal as opposed to 3.2V which is the standard for typical LIFePO4, but this could be just due to the low Ri and hence higher Voltage they maintain at the standard 0.5C that these values are based upon, more than any thing else.

As long as the high and low cutoffs are the same, I can't see how this would be an issue. Anywhere in between, the cells are always going to be trading current to equalize the terminal voltage. The main thing is just to make sure you don't try to parallel two cells that are at very different OCV's, or large currents will flow. But what we're discussing is treating the different types of cells as a pack, never decoupling them.

 

[cell_man]

What's not clear to me is the max pulse duration/amplitude... which I suppose would depend a lot on the specific cells used. Hrmmm, I think I just convinced myself that the only way to answer the question is to try it since batteries don't come with spice models. :wink:

I think we are definitely singing from the same hymn sheet

From what I've seen LifePO4 likes to stay between about 3 to 3.3V. If the the resting voltage drops much below that on a cell with low SOC, it doesn't take much charge current to bring it up above 3V. My gut feeling is that I can't see a situation where the difference in the resting will be big enough to cause an in issue with either the discharging or charging cells. I can't see a situation where the discharge from the low C cells can ever get out of hand as they can only be pulled so low and the same for the high C cells, the voltage supplied from the low C cells just wouldn't be high enough to drive any very high current levels into them IMO. Current needs voltage to flow and with small potential difference and some Ri on both cell groups, I can't see the current levels between them ever getting dangerously.

What happens when the high C pack is almost discharged though and the charging doesn't have time to bring it back up? I think, what was initially say a 10C combination might revert back to something closer to the 2C of the lower C group?

However, it's all just my tuppence

 

[The Mighty Volt]

Thanks everybody for contributing.

I think it's important to ask these kind of obscure questions. People are inevitably going to try and do things "Ghetto", so it is as well to talk about them as not.

 

[dogman dan]

Wouldn't it be better to take some of the cells and make paralell strings of them, then take the lower c rate cells and make paralell strings of them. Then series connect the strings so you don't have mixed batteries in a paralell string.

Tha'ts how I'm going to do it with my lipos. 30 c lipos paired up paralell, series connected to 20c lipos. That way one c rate won't try to charge up the other. then I'll be discharging at about 3c max.

 

[The Mighty Volt]

Wouldn't it be better to take some of the cells and make paralell strings of them, then take the lower c rate cells and make paralell strings of them. Then series connect the strings so you don't have mixed batteries in a paralell string.

Tha'ts how I'm going to do it with my lipos. 30 c lipos paired up paralell, series connected to 20c lipos. That way one c rate won't try to charge up the other. then I'll be discharging at about 3c max.

Yeah that's an idea, that's probably what I will do.

 

[The Mighty Volt]

Been giving this some more thought as part of an attempt to limit my spending and use what I have at my disposal in the quest for 100v of reliable battery.

Given that my "strings" are basically 1s10p for the 26650, that equates to 22Ah or so of capacity.

Now, this means all my other strings, the supplementary strings of lower discharge generic 15c LiFePo4 and 18650 A123, would have to be the same capacity, that is to say, 1s20p or so.

By my reckoning, twenty 15C cells in a string should be more than capable of serving the 65Amp or so discharge rate which I am asking them to meet.

If I am correct, it is the relatively long string in which they are paralleled that makes this possible.

22Ah @ 1c would be 22Amps of continuous current {over one hour}

22Ah @15C would be 330Amps continuous current {over 4 minutes @330Amps or over 20minutes where the discharge amperage is limited to 65Amps of the available 330Amps}

Thus, in conclusion, once the peak Amps of a given string exceeds the maximum Amperage drawn, the strings should all be capable of working in tandem without any major issues.

Clearly, the governing issue in this is a question of the string size, in Ah, being sufficiently high in the first instance so that when combined with the pertinent C rate of that string, that the Amperage required can be comfortably drawn.

Any opinions on this, or any qualified revisions of my statements would be gratefully received.

 

[Gordo]

Here is how it works;

http://enertracrawfeed.blogspot.com/

 

[Holocene]

thinking about... combining different types of cells (even different chemistries) in parallel... Has anyone tried / characterized this?

Sure, b_b. Read Justin's "Cross-Canada by Ebike" thread:
http://endless-sphere.com/forums/viewtopic.php?f=29&t=7134

He did 7,000 km on $10 worth of electricity... not sure he had to pay for all of it, either! 8)

Cheers!

 

[liveforphysics]

Wouldn't it be better to take some of the cells and make paralell strings of them, then take the lower c rate cells and make paralell strings of them. Then series connect the strings so you don't have mixed batteries in a paralell string.

Tha'ts how I'm going to do it with my lipos. 30 c lipos paired up paralell, series connected to 20c lipos. That way one c rate won't try to charge up the other. then I'll be discharging at about 3c max.

You really want them paralleled at the cell level my friend. Otherwise a continuous load is going to only drain the high-c cells.

Remember, a battery is like a fixed voltage source dependent on the state of the SOC, and then a resistor in series with that (the batteries resistance).
So, the current it pulls is going to be proportionate to the voltage drop across that resistor, divided by the value of that resistor. Which means, when cells are paralleled directly, the one with lower resistance is the one where the most coulombs of charge get taken from. Fortunately, by means of electrochemical action, the cells with a direct physical connection (and at the same temperature) will equalize SOC between each other at a rate determined by the resistance of the connection between them.

Essentially, the goal with multi-chemistry paralleling is to have them directly paralleled, so under heavy loads, the voltage stays at a reasonable point, keeping the pack cool and delivering the acceleration demands needed. Then, during the off throttle times, the higher resistance (lower-C) cells which will inherently always be at a higher SOC will discharge into the high-C cells to equalize the SOC between them during the off-load times. It works perfectly as long as your high current loading is kept intermittent. However... with LiFePO4, the voltage vs SOC curve is quite flat, so it requires either very low resistance connections between the cells (to enable the most current transfer from the <0.05v difference they will have), or very extended periods between high load intervals. A chemistry like LiCo/LiMn with a large voltage change relative to SOC% can much more rapidly equalize SOC between parallel cell groups.

Also worth noting, the additional back/forth Lithium motion in the high-C cell does cause a decrease in useful cycle life, but if the motion is happening while mainly around the nominal voltage of the cells, it's largely mitigated due to substantial available lithium and space for insertion/extraction.

 

[The Mighty Volt]

So, Luke, are you basically suggesting that the strings should consist of mixed-C-rate cells which then exist in differing relationships depending on load?

That is to say, a 1s10p string of 5c and 10c cells act in a sort of chemical and electrical symbiosis, with the 10c cells providing the bulk of the current to meet the load demand, and 5c cells then "repaying" the 10c cells after the load comes off, with everything panning out even, so to speak??

You refer to "Multi-Chemistry Paralleling"- suppose the chemistry is identical but the C rates are different....what is the situation then?

 

[dogman dan]

That's the way I see it too. even with same chemistry, and identical cells, you can have different internal resistance in two different cells. Under load, the one with lower resistance will tend to put more of the amps into the load. After the load is removed, if there is a voltage difference between the two cells, the paralell connection will allow one to charge the other till the voltages are equal.

If the load is continous, then as the one battery discharges more than the other, it's resistance will increase causing it to provide less current. In the end, the cells tend to reach end of charge at the same time. But one cell could get overdischarged if the load is continuos and the difference is too great. Like 2c cells paralelled with 20c stuff.

For sure though, the best bet is to paralell stuff at the individual cell level with wire large enough to carry the load if you expect a lot of current flowing from one cell to the other when the load is removed.

If your amps are moderate, chances are you can get away with just about anything mixing c rates. For example, some 10c lipo mixed with 50c. Then you pull 3 c from the entire pack. Whoop de do at those amps. Pulling 3c is not going to bother either battery one bit. So in such a case the c rates don't matter at all.

BTW, I ended up paralelling my 20c lipo with 30 c lipo, then series connecting the paralell strings. It seems to be working, I'm drawing only 2c out of it. But I think I will want to keep the stuff paired by c rate on the racing bike that uses more amps.

 

[markcycle]

thought I'd show what I'm doing in regard to paralleling cells of different C rates

http://enertracrawfeed.blogspot.com/search/label/Battery pack

I'm doing a interesting mixed chemistry setup that uses the best of LIFEPO4 and Lipo in a diode summed system.
My motivation in doing this stems from the fact that the 40AH Thundersky cells just don't put out enough voltage under load (Sag). For EnerTrac this makes it difficult for my customers to get a true understanding of the power and speed of the 600 series motor, when test riding my bikes. So I need a way to boost voltage under load, without replacing the complete battery pack.

In the above picture you'll see the CA meter (lower) and a second smaller meter up top. The top meter monitors the Lipo pack current while the CA meter monitors the total system. Up to 1.5C or 60 amps, the LIFEPO4 batteries run the bike, then as current exceeds 60 amps or beyond 2C the LIPO kick in. At 100 amps the LIFEPO4 pack maintain 1.5C and the Lipo pack does the remaining 40 amps. My 40AH LIFEPO4 cells now never see much over 60 amps while my 20C rated Lipo see as much as 200 amps peak.
The Lipo are bulk charged with a separate Meanwell type power supply. Separate balance units are used as needed, they are connected in parallel with the display units.

I accomplished this by exploiting the fact that Thundersky cell have much higher IR than the Lipo cells and setting the Thundersky pack to be at a higher no load voltage. The diodes are 150 volt schocky diodes.
Testing was done on my Dyno. The box under the bike is the Lipo pack 25S 2P

I have only done 4 miles of road testing before the snow hit. I had the Lipo up to 100 amps while the LIFEPO4 stayed under 80 amps during this on road test.
My Dyno Currently can only do 100 amps continuous, measured at controller input. My plan is to do a video record of the display system so I can record in real time the speed, current of both batteries and then put the data in a spreadsheet.
My data so far shows temperature not to be a big problem I've had the batteries on test from 45F to 70F
My big concern is
The 4 to 1 AH ratio - is it correct, only on road testing can determine that.
Battery pack

I think the battery box weights more than the batteries.
The big challenge was getting all the hardware into the bike and still be able to get the skins on.

The bike with the skins back on

Conclusion:
More data is needed but my testing to date, shows it's possible to mix two chemistry types and use the best characteristics of each respective battery type.

Mark

 

[The Mighty Volt]

That's the way I see it too. even with same chemistry, and identical cells, you can have different internal resistance in two different cells. Under load, the one with lower resistance will tend to put more of the amps into the load. After the load is removed, if there is a voltage difference between the two cells, the paralell connection will allow one to charge the other till the voltages are equal.

If the load is continous, then as the one battery discharges more than the other, it's resistance will increase causing it to provide less current. In the end, the cells tend to reach end of charge at the same time. But one cell could get overdischarged if the load is continuos and the difference is too great. Like 2c cells paralelled with 20c stuff.

For sure though, the best bet is to paralell stuff at the individual cell level with wire large enough to carry the load if you expect a lot of current flowing from one cell to the other when the load is removed.

If your amps are moderate, chances are you can get away with just about anything mixing c rates. For example, some 10c lipo mixed with 50c. Then you pull 3 c from the entire pack. Whoop de do at those amps. Pulling 3c is not going to bother either battery one bit. So in such a case the c rates don't matter at all.

BTW, I ended up paralelling my 20c lipo with 30 c lipo, then series connecting the paralell strings. It seems to be working, I'm drawing only 2c out of it. But I think I will want to keep the stuff paired by c rate on the racing bike that uses more amps.

Hi dogman thanks for the contribution- you raise the point, and I referred to it also- the discharge rate of the cells and relation between that discharge rate and the current being drawn by the controller.

I think that if one has a 3C string paralleled with a 30C string, then the current being drawn by the controller comes into focus once we go beyond a specific value.

However, how do we determine that value?

Suppose I have a 22Ah string, at 3.3v, of 15C cells.......is there any reason to believe that a controller, programmed to draw 65 Amps max load, should draw from that string more current than that pack is capable of comfortably discharging?

22Ah at 1c = 22 amps continuous current over 1 hour. Clearly, this would not support the current rate required and there would be a serious issue within the pack, both at cell and at string level.

22Ah at 15c = 330 Amps, continuous, for 4 minutes, or 65Amps continuous for 20 minutes, all other things being equal.

I think, and I do not wish to be contradictory or dogmatic {no pun intended} about it, but should the overall capacity of the "weaker" string not be taken into account once that capacity is high enough, when combined with the C rate, to support/meet the constant current demands.

bear in mind here that my hypothesis is based on the following

1. That the same chemistry is being used, namely LiFePo4
2. That the maximum current drawn is 65 Amps
3. That the string size is 1s20p for the 1100mAg cells and 1s10p for the 2200mAh cells
4. That the lowest C-Rate in question is 15c.