Paralleling dissimilar cells (hybrid), interesting results

Overclocker

10 kW
Joined
Mar 17, 2014
Messages
794
Location
Philippines
so i was planning on building a "hybrid" pack composed of 17s6p Samsung 26F 2600mah, paralleled to 17s6p Amperex INR18650 2200mah. the amperex cells are older and have a bit higher internal resistance

so i set them up with their own volt/ammeters to show independent readings, then wired them up in parallel to a ~20W resistor load. the amperex on the left (blue), pink samsung on the right:

8JzSrFu.jpg


as expected, the cell with the lower internal resistance puts out more current. but the interesting thing is that when the load if removed the cell with the higher voltage, depending on where it is on its discharge curve, starts charging the other cell:

o3RLV3L.jpg


PS these ammeters cannot show negative currents so it doesn't show the amperex cell actually charging
 
I7RlhGv.jpg


OK it makes perfect sense now. further down the discharge the curves have intersected and now the amperex is charging the sanyo when the load is removed

at around the intersection point there would be minimal inter-cell charging

it seems that in order to parallel heterogeneous cells successfully (i.e. minimal inter-charging) their discharge curves have to be at least similar. bonus points for curves that intersect :mrgreen:

i'll try to graph to better visualize
 
it seems that in order to parallel heterogeneous cells successfully (i.e. minimal inter-charging) their discharge curves have to be at least similar. bonus points for curves that intersect :mrgreen:
Why do you believe that to be successful , you need to minimise "inter-charging" ??
..isnt that just a form of self balancing ?
 
Why is it bad?

The scenario that interests me most is a large energy dense pack paralleled with a very small very power dense pack, for example 15p Samsung 30q in line with 4.5ah "100c" lipo. Obviously the small pack with low IR will be drained more quickly than the 18650's, but given the lipo can accept very high charging rates I don't think this is actually a problem unless the discharge lasts for a relatively long time.

For peak absorption it seems to stack up..
 
E-S member aerowhatt did something similar with 3 hawker 42 AH SLA and two Ping 36V 20 AH batteries;
http://endless-sphere.com/forums/viewtopic.php?f=14&t=8815&start=75#p139016

http://endless-sphere.com/forums/viewtopic.php?f=14&t=6285#p125438
aerowhatt said:
Another nice outcome is that even though the AGM's do the heavy lifting for acceleration and hill climbing, they generally are still above 50% DOD when 3/4th's of the max range is used up! They should last a long time with that usage and already have more miles on them than previous, all lead packs, have lasted on the same platform.

http://endless-sphere.com/forums/viewtopic.php?f=14&t=8815&start=60#p138760
aerowhatt said:
Basically it doesn't take much additional voltage drop on the LiFePO4 to get the SLA's to step up a bit more and give a more ideal balance of current delivery during peaks loads. So if you have a dual battery system that is naturally pretty close to each performing where you want them too. Changing the harness resistance of either, or both batteries to fine tune it results in a very low loss.
 
Hillhater said:
it seems that in order to parallel heterogeneous cells successfully (i.e. minimal inter-charging) their discharge curves have to be at least similar. bonus points for curves that intersect :mrgreen:
Why do you believe that to be successful , you need to minimise "inter-charging" ??
..isnt that just a form of self balancing ?

mzoCy2Q.png


if their curves are very dissimilar (like above) the inter-charging might become too severe at the points where they diverge. since dis/charging is an electrochemical process that's not 100% efficient then it would be best to minimize it. but i guess w/ li-ion it's actually quite efficient so it might not be that bad, except when inter-charging becomes too high (probably >0.5C). or like in the graph where one cell falls off the cliff at 3.5v while the NCA cell keeps on trucking until 3.1v

but that's just speculation. my experiment involves 2 cells wit similar curves
 
As always, the answer is "it depends" As in the graph above, you'd have to stop the discharge when the first battery (red line) is about to go over the cliff. Otherwise risk over discharge of the first pack. At 3.6v, the red line battery needs to stop. You could continue, and they would continue to have equal voltage after the intercharging finished, but they would balance out at a voltage that is still over discharged for the red line cell. You'd have to leave about .75 ah of the blue cells capacity on the table. Not ideal to pair those two. But it could still be ok,, if the ratio of the blue cells was ok. like 5 red cells to one blue. Still leaving some on the table, but % wise, less than if it was 1 to 1. You were talking about a 1 to 1 ratio hybrid pack, so that would be leaving a lot of blue cell capacity on the table. 6x.75 ah. 4.5 ah left unused. That's 250 wh or more!! Bleah.

As for the divergence of the two lines earlier in the discharge,, hell that will be the day when I worry my ass off over .1v. The amperage of the intercharging between packs that differ by .1v is not going to make a wire heat up, or get a cell all hot and bothered. But that discharge stop point for the red cell is a deal breaker for a 1-1 ratio pack.

If you tried to stiffen a big pack of laptop cells with lipo, then you'd lose some of the capacity of the laptops, because you'd be stopping at 3.5v for the lipo when the laptop cells would not be fully discharged. Now the ratio works the wrong way, unlike the red blue example above. Most of your capacity in the laptop cells, but now you have to stop early because of the big ah battery, losing a large % of your capacity to save the small high c rate pack.

It would be possible though, to unplug the lipo, and then slowly run home with just the laptop pack.

In other cases, the 100% discharged voltage is more similar, or maybe you never plan to reach 100% DOD. At one point,, I found some dodgy NiCad really worked well when combined with lead. The lead got less peukerts with the NiCad stiffening it, and the NiCad never got hot as it had been doing when run by itself. the end result was all win. To recharge, I simply plugged in the NiCad charger, and let it charge the lead too. Both packs were happier, but that may have been primarily because I could return home less than 100% dod with the bigger combined capacity.
 
Well, you could do that, or you could have a battery pack with homogenous characteristics that's easy to charge, maintain, monitor and provides good performance. In a world of readily purchasable new batteries this is more like a "if you have to" kind of thing.
 
Behaves basically just like I thought it would in theory. And confirms that there is no good reason to leave cells of various types paralleled. Now a turbo button or something on an ebike that adds in a smaller high current pack temporarily would be another situation entirely. :evil:

Are you able to do another test with a more exaggerated cell difference? Such as a laptop grade cell (like the Samsung 26F) with ~50 Mili ohms pared with a ~15 mili ohm (20-30A) rated power tool grade cell (25R, HE4, HE2, VTC5, NSX etc)? Because those are the main scenarios that people suggest trying.

Also worth mentioning is that depending on the load you expose the cells to, you might find situations where the cells are above 50C and one cell is charging the other. Li-ion's don't like being charged when hot, which is why basically all reputable power tool pack makers put temperature sensors and the packs generally don't charge until they cool down when they have been used under heavy loads.
 
i just did discharge tests (to 2.7v) on the iCharger:

26F (solo at 3.0A): 2456mah
ATL(solo at 2.5A): 2208mah
26F + ATL (in parallel at 5.5A): 4540mah*

i call that a success 8) minimal inter-charging well below 0.5C. should i require more range i could go to this comparator: http://lygte-info.dk/review/batteries2012/Common18650comparator.php and select something with a similar curve


*4540 is less than (2456+2208). the icharger probably terminated a bit early coz it doesn't use the balance port. it senses end voltage over the main leads and 5.5A induces a bigger voltage drop
 
eTrike said:
Thanks for the examples gogo and dogman. They clearly highlight some benefits like improved power and longevity.

@flattire & redilast, if that is your takeaway then you must not be comprehending what has been shared. Please attempt to set aside your confirmation bias.

I've looked at your data and sources (most of them talk about hybriding with super caps with special circuitry) and besides super caps, I still firmly believe its a bad idea and should be avoided. I trust my own instinct, and the knowledge and recommendations from the battery industry.

The funny thing is what overclocker is doing is almost the opposite of a hybrid pack. He is selecting cells that are very similar in performance and paralleling them. He isn't taking cells with opposite cell characteristics such as high capacity low current (NCR18650B for example) and High Current, low capacity (.e.g Sony VTC4) and mixing them, which I thought was kind of the whole point of a hybrid type pack. Mix high current output caps or batteries with lower current batteries.
 
Those two were so similar,, I was a bit surprised how dissimilar they were on 100% dod voltage.

FWIW,, the only reason I have paralleled dis similar chemistries, was to flog a nearly out of service battery a few more cycles. Like when my ping was dying, at 50% capacity and puffing on one end. I eeked a few more trips out of it by paralelling with 5 ah lipo. It just sagged too much to use anymore by itself.

A new battery was in the works,, but the cash for it had to be raised first,, so for a month or so I did the awkward routine. Charge both,, then run the ping one block to bring it down to the voltage of the lipo. Connect, and proceed. Stop when the voltage of the lipo dropped to 50v. When healthy, the ping would go to 48v.

Truly a pain in the ass. But I got through till the battery order arrived. At the time, the only good battery I had left was that 5 ah of lipo, but I needed 10 ah of range to do anything.
 
test cells: ATL INR18650 (blue) and Sony VTC4 (green)

initially the Sony grabs a bigger share

H4PQTA0.jpg


but since the Sony depletes faster its voltage goes down as well which prevents it from pushing out as much current:

36WnAsl.jpg


wkDU7Mp.jpg


at this point the sharing is almost 50/50:

rJu6rOB.jpg


when the load is removed we see a much greater inter-charging current compared to the ATL+26F combo due to the bigger voltage imbalance

aaJbbqp.jpg
 
Again, exactly what the theory predicts.
What is your conclusion from that ?
Inter-charging...otherwise known as self balancing of parallel cells. A function of every battery containing sets of parallel cells, because no two cells are identical. ,!
 
Overclocker thanks for posting those results with the VTC4. Like I imagined the higher current cell because it held higher voltage provided more current. But what I am most curious about is what happens towards the end of the discharge cycle. Because the VTC4 and other ATL cell have similar capacities, and the VTC4 has much lower internal resistance and the VTC4 was putting out much higher amps at least initially, would there come a point where the ATL cell is actually charging the VTC4, due to the VTC4 using up most of its energy sooner? If so that would be really bad. As the point in hybriding would be to get more current output, but towards the end you would be getting out less current output than just the set of ATL cells due to them needing to charge the VTC4's. Can you try a longer discharge test perhaps checking at 3.5v, 3v and then 2.5v?
 
redilast said:
Overclocker thanks for posting those results with the VTC4. Like I imagined the higher current cell because it held higher voltage provided more current. ......
:?: no ! , you are miss understanding what is happening. I dont see a significant higher voltage ?
The VTC4 is giving more current because of its lower IR !
Of course this boosted discharge cannot go on indefinitely, just as with any pack it will have a limited capacity.
These "hybrid" packs are not suitable for every situation, but can be very benificial in some applications, such as where high energy density is needed together with the capability of brief high power discharge (or regen) boost periods that high energy density cells cannot provide alone. That intermittent discharge cycle allows the pack to re balance between "boost" or regen operation.
 
Hillhater said:
:?: no ! , you are miss understanding what is happening. I dont see a significant higher voltage ?
The VTC4 is giving more current because of its lower IR !

I'm not misunderstanding it at all. To clarify my statement. The VTC4 is giving more current because its lower IR leads to higher voltage under the same load as the ATL cell. The reason the voltage doesn't appear much different from the other ATL cell is because the other cell has higher IR but is put under a lower current lower load to balance the voltages between the two cells.

But again, you have to look at the whole situation. Initial performance, half capacity performance and nearly depleted performance and see what is actually happening between the current sharing and output between the cells.
 
That was my point ..
The voltage is a CONSEQUENCE of the current distribution caused by the IR differences.
And theoretically the voltage on each cell is equal.
The real comparison would be to compare this hybrid pair, to a similar test using a pair of the ATL cells....voltage amps etc.
 
redilast said:
But again, you have to look at the whole situation. Initial performance, half capacity performance and nearly depleted performance and see what is actually happening between the current sharing and output between the cells.


from what we've seen so far those scenarios are easy to predict

initial performance already shown

half capacity: just continue the trend and the distribution would be almost 50-50

towards the end the ATL would have a larger share

but that's when doing CONTINUOUS discharge which isn't representative of typical ebike usage. suppose at half-capacity the pack is allowed to rest, the ATL would charge the sony and their voltages would equalize. this gives the advantage back to the sony and would take the lion's share again
 
Overclocker said:
but that's when doing CONTINUOUS discharge which isn't representative of typical ebike usage. suppose at half-capacity the pack is allowed to rest, the ATL would charge the sony and their voltages would equalize. this gives the advantage back to the sony and would take the lion's share again

So basically its only practical if you babysit the pack and never use more than maybe 50-75% of the capacity. Because like I mentioned earlier I believe the Sony is going to give up more of its capacity first due to its lower IR, which means towards the end of the total combined usable capacity of the pack, the ATL is going to have to recharge the Sony and if this is done under the load from the motor (not allowing the ebike pack to rest) that is a very bad situation.
 
Yes I theorized this way back when (wow been that long!) when I was a noob.
https://endless-sphere.com/forums/viewtopic.php?f=2&t=35600&p=986072&hilit=parallel+chemistry#p986072

And yes I was right :-D

Summary:
Thus, a high power cell (low impedance, R1) in parallel with a low power cell (high impedance, R2), assuming similar chemistry and short term power bursts will result in the following behavior:
-Under load, the high power cell will output most of the current, A, draining it more quickly. The Amount of current coming from each cell is actually a function of their impedance relative to both cell's equivalent parallel impedance. The equations simplify to:
(A = A1+A2)
A1/A2 = R2/R1
-Upon removal (or reduction of) load, The high power cell would have a lower disconnected resting voltage (because it was drained more)
-The low power cell would charge the high power cell until their disconnected resting voltages became equal (or the discrepancy in charge was transferred)

For longer-timed loads, it gets a bit more complicated. The amp distribution will start off at A1/A2 = R2/R1, but will slowly shift to the lower power cell supplying more amps as the high power cell becomes more drained.
The problem here is if you have a 2.5V cut off, it's quite possible you may end up in a situation where theoretically once cell becomes over-discharged based on the cut off voltage and the per-cell amps.
Thus you'd want a safety factor here, say use a 2.8 or 3.0V cut off.
 
redilast said:
Overclocker said:
but that's when doing CONTINUOUS discharge which isn't representative of typical ebike usage. suppose at half-capacity the pack is allowed to rest, the ATL would charge the sony and their voltages would equalize. this gives the advantage back to the sony and would take the lion's share again

So basically its only practical if you babysit the pack and never use more than maybe 50-75% of the capacity. Because like I mentioned earlier I believe the Sony is going to give up more of its capacity first due to its lower IR, which means towards the end of the total combined usable capacity of the pack, the ATL is going to have to recharge the Sony and if this is done under the load from the motor (not allowing the ebike pack to rest) that is a very bad situation.


remember that inter-charging happens when one cell has a higher voltage than the other. under load both cells become depressed relative to their respective IR, so if the load is large enough there won't be intercharging even at zero SOC

at very light loads in the same ballpark as intercharging current then the ATL would send some current to the sony and some to the motor. both very small currents so it shouldn't be a problem

the bottom line is in these small-scale experiments, hetero-paralleling works quite well even in "extreme" cases (ATL+VTC4). i'll be building up my 102-cell ATL + 102-cell 26F and will see how they behave full-scale
 
Back
Top