Cell internal resistance matching importance (parallel vs serial)

[alcatraz]

Hi everyone

Some of you might be into building battery packs and even fewer might have experimented with using used cells.

For those of you who have, and for those that are interested I'd like to talk about internal resistance matching of cells.

Because our chargers don't see individual voltages they are risky to use as the cells start to deviate from oneanother.

One way is to get a hobby charger and hook things up individually but lets say you don't want to do this.

Lets say your task is to match a battery pack from used cells.

You have internal resistance cells ranging from 30-70mOhm and they have capacites from 2500-3000mAh.

For the sake of arguing lets say you have:

10pcs of 2500mAh and 70mOhm

10pcs of 3000mAh and 30mOhm

The objective is to make a balanced pack of 20s. (20 cells/groups in series to create an 84V pack)

Putting the same kind in parallel would be very unwise because upon charging the 2500mAh cell groups would be go over voltage very fast. They would upon discharge also go under minimum voltage first.

However, what if we take one of each and connect in parallel. Then connect 10 pairs in serial. Theoretically the pack should be balanced because each group has the same resulying capacity and internal resistance.

I'd like to discuss what would be the risks of doing such a thing? Would the weaker cells accelerate degeneration of the stronger cells?

Would the currents going beteeen batteries in each parallel group be very harmful to the cycle life?

Would putting a 70 and 35 mOhm battery in parallel mean that the 35 mOhm one is discharged twice as fast? The max discharge rate for my cells is 10A and lets say that the max current draw of the pack is 15A. It should be within tolerances.

Thank you for reading and commenting

[esaj]

I'll start with the usual "I'm just a hobbyist and you may want to consult a professional", but here's my take:

On 7/16/2018 at 5:18 AM, alcatraz said:

Hi everyone

Some of you might be into building battery packs and even fewer might have experimented with using used cells.

For those of you who have, and for those that are interested I'd like to talk about internal resistance matching of cells.

A guy who builds battery packs professionally once said to me that cell matching is a waste of money, but, and this is the important bit, we were talking about brand new cells and matching those very closely (ie. buy 100 cells and pick out the 4 x 16 = 64 pieces that are most closely matched for building 4 x 16S4P). His opinion was that it rarely is of any use in the pack longevity or such, and is just money down the drain. But like said, this was in relation to buying new, unused cells of same make and model.

 

Quote

Because our chargers don't see individual voltages they are risky to use as the cells start to deviate from oneanother.

Yes, the charger is a "dumb" CV/CC (constant voltage / constant current) power source, the battery management system (BMS) -board in the packs is (or at least should) taking care of balancing the cells. You need to have a proper BMS for the pack, that's the best safety for the pack.

 

Quote

One way is to get a hobby charger and hook things up individually but lets say you don't want to do this.

Lets say your task is to match a battery pack from used cells.

You have internal resistance cells ranging from 30-70mOhm and they have capacites from 2500-3000mAh.

For the sake of arguing lets say you have:

10pcs of 2500mAh and 70mOhm

10pcs of 3000mAh and 30mOhm

The objective is to make a balanced pack of 20s. (20 cells/groups in series to create an 84V pack)

Putting the same kind in parallel would be very unwise because upon charging the 2500mAh cell groups would be go over voltage very fast. They would upon discharge also go under minimum voltage first.

Not 100% sure what you're saying here, as from 10 pieces each, you could only build one 20S pack, so all the cells would be in series. Putting different types of cells in series is not a good idea, as the lower capacity / higher resistance cells would drop their voltages faster and even get damaged / reverse their voltage. Proper BMS should prevent overcharging and over discharging, but likely the latter would cause a cutout (the BMS cuts the power output from the battery).

 

Quote

However, what if we take one of each and connect in parallel. Then connect 10 pairs in serial. Theoretically the pack should be balanced because each group has the same resulying capacity and internal resistance.

You'd still need 20 of each (to create 20 pairs) to get a 20S2P pack. I've never tried it, but people have put packs made from different cells in parallel, and said that it works just fine (basically if one or the other pack is at higher voltage level, it starts charging the other pack). Having cells of different capacity / internal resistance in parallel might cause some weird effects, at least the bulk of the current will run through the lower resistance cell, but as the parallel cells connected through very low resistance, they should "even out" the voltage...

 

Quote

I'd like to discuss what would be the risks of doing such a thing? Would the weaker cells accelerate degeneration of the stronger cells?

Possibly, if you're using used cells, the cycle life is already lower. Another thing to note is that not only does the capacity of the cell degrade over time, but the internal resistance also might go up. While the cells may work just fine under normal usage, they may drop a lot of voltage at high current draw, and it usually just takes one bad cell to ruin the entire pack, like the BMS cuts the power if one cell drops too low, or even if it doesn't, that one cell reverses or overheats and destroys the entire pack... On the other hand, it might just work fine?   Safest way I can think of would be measuring each cell with an electronic load to see if they "behave" under higher load correctly, and discard any "bad" cells.

 

Quote

Would the currents going beteeen batteries in each parallel group be very harmful to the cycle life?

Possibly, but maybe not. If you have separate packs (ie. batteries, 2 x 20S1P) with their own BMSs, even if the other pack conks out, you still have one pack that you can limp home with (but if you're drawing a high current out from the packs and one cuts out, all the load drops on the second one, which might also buckle under the pressure...). The moment the other pack starts sagging, the second pack will start to charge it (simultaneously with powering the wheel), which may or may not be too much for the pack. But this is all just theoretical hand waving without any real measurements to show   

 

Quote

Would putting a 70 and 35 mOhm battery in parallel mean that the 35 mOhm one is discharged twice as fast?

Somewhat faster, but as said above, when the packs are in parallel, the lower resistance one will also output current to keep the parallel pack in the same voltage (at least at the output-level, what the individual cell voltages are at that point is "invisible" to the other pack). Basically, you wouldn't probably get as much capacity out of the parallel packs as their combined watthours (/amphours) should be.

 

Quote

The max discharge rate for my cells is 10A and lets say that the max current draw of the pack is 15A. It should be within tolerances.

Depends on your wheel, high powered wheels can draw very high current during takeoff and acceleration, but they (usually) don't last very long. If the voltage drops too much (or one or more cells are in bad shape and drop their voltage enough), the BMS might cut out, or if it doesn't have undervoltage protection on the discharge side, then you probably end up with a dead cell sooner or later.

 

Quote

Thank you for reading and commenting

No problem, but I underline that I'm not a professional, so I don't really know...     Plus I might be a bit drunk now...  

As a hobby, building battery packs for your own use might be fun (and somewhat dangerous), but if you're planning on saving money, likely you won't in the long run. If you want proper packs, contact someone like 1RadWerkstatt and get them to make you proper pack(s) from brand new, high quality cells and BMSs.

[esaj]

To dig a little bit deeper, I did a simulation with 30mOhm and 70mOhm cells paralleled. It's a bit "make-shift", but the basic idea is to simulate a pair under 10A draw.

So, I put two voltage sources in parallel, both at 3.6V (about the nominal of li-ion cells, typically 3.6...3.7V), one with 30 milliohm internal resistance, other with 70 milliohms. If you directly connect two voltage sources in LTSpice, that means "0 ohm" resistance between them, which would make it impossible for the voltages to differ, so I used two 100 micro-ohm (0.1 milliohm) resistors to simulate the connection between the cells. I then use an "ideal current source" to cause a draw of 10A through the pair.

What the graphs show (nevermind that the current is shown as negative, LTSpice just "wants" to show it that way, as I need to put the current source the "wrong way around" for it not to cause the voltage to raise before the voltage sources ) is that the voltage of Cell1 (the 70mOhm one) is around 3.3897V, while Cell2 is at 3.390116V, so a slight voltage difference exists. What the current graphs (well, lines, this is just a DC steady state simulation) show is that the lower resistance (Cell2) is passing about 7A of current, while the higher resistance Cell1 only passes about 3A.

Like said, this is a bit "make shift" -simulation, as in reality the current isn't steady (the motor's being "pulsed" on and off, causing both high and low current draw alternatingly, as well as there are capacitors on the mainboard to steady the voltage and supply the very high current spikes, plus the motor inductances and mosfets, switching frequencies, parasitic capacitances and inductances etc. play a role...), and plain voltage sources with internal resistances cannot simulate the behavior of a lithium cell correctly.

[alcatraz]

Thank you very much esaj.

It seems to confirm my suspicions. The problem with dummy charging is you don't know in 100 cycles if the values are off again. Dummy charging relies on packs being balanced by so much more than capacity. And the slightest imbalance you will deviate outside the maximum 4.2v which starts to degrade the cells. I'm told you can't trust the bms to balance anything except tiny mismatches.

Lets say I'm leaning towards using a hobby charger to charge my packs. (Route leads out through some kind of custom ATX connector so I can quickly connect everything (and monitor voltages) I should be well off.

I read an article that measured cycle life with NCR18650B (previous no.1 cell): http://blog.evandmore.com/lets-talk-about-the-panasonic-ncr18650b/

0-100% 500 cycles

6-94% 7500 cycles

8-92% 15000 cycles

10-90% 28000 cycles

20-80% 35000 cycles

30-70% 40000 cycles

So I'm inclined to get a hobby charger to reliably get to the 4.10-4.15v the 90% would imply. (Author writes 90% = 4.15v which seems a bit off, possible?)

If the BMS doesn't need to do charge balancing and doesn't need to overvolt protect (and doesn't have undervolt protection = EUC bms) then why should I keep it at all?

What do you do with your EUC esaj? What's your charging setup?

Cheers! /a

[esaj]

13 hours ago, alcatraz said:

Thank you very much esaj.

It seems to confirm my suspicions. The problem with dummy charging is you don't know in 100 cycles if the values are off again. Dummy charging relies on packs being balanced by so much more than capacity. And the slightest imbalance you will deviate outside the maximum 4.2v which starts to degrade the cells. I'm told you can't trust the bms to balance anything except tiny mismatches.

Most BMSs just "shunt" (bypass) the cell once it reaches the full 4.2V, but there are more "intelligent" BMSs that have active balancing that can balance the voltages all throughout the charging. Don't know if any wheel uses them, but I know they exist    Probably much more expensive than "normal" BMSs though.

 

13 hours ago, alcatraz said:

Lets say I'm leaning towards using a hobby charger to charge my packs. (Route leads out through some kind of custom ATX connector so I can quickly connect everything (and monitor voltages) I should be well off.

I read an article that measured cycle life with NCR18650B (previous no.1 cell): http://blog.evandmore.com/lets-talk-about-the-panasonic-ncr18650b/

0-100% 500 cycles

6-94% 7500 cycles

8-92% 15000 cycles

10-90% 28000 cycles

20-80% 35000 cycles

30-70% 40000 cycles

So I'm inclined to get a hobby charger to reliably get to the 4.10-4.15v the 90% would imply. (Author writes 90% = 4.15v which seems a bit off, possible?)

Could be, some chemistries can go up to 4.25-4.30V per cell (NMC? Maybe, don't remember...)

 

13 hours ago, alcatraz said:

If the BMS doesn't need to do charge balancing and doesn't need to overvolt protect (and doesn't have undervolt protection = EUC bms) then why should I keep it at all?

Of course you could go without BMS, although care must be taken since the pack won't be protected at all, for example in case of an accidental short circuit. AFAIK, the EUC BMSs also have undervoltage protection, but it's really low, something like 2.5V per cell. EUC Extreme uses unprotected LiPo-pouches and custom charger setup for high output and really fast charging (Something like 2 x 1.4kW, don't remember the model, it's somewhere in the forums).

 

13 hours ago, alcatraz said:

What do you do with your EUC esaj? What's your charging setup?

Cheers! /a

I use the Charge Doctor V2 with two standard "brick chargers" in parallel, although I don't limit the voltage, but I don't always charge it all the way to full:

 

 

[alcatraz]

We all know charging IR (internal resistance) mismatched cells in series is bad.

What about discharging mismatched chains? Imagine every cell(group) has the SAME capacity but has a mix of different IR cells?

Would the low IR groups get discharged to lower voltages before the high IR groups? Or are we only talking the self discharge perhaps?

It seems to me that the increase of IR and reduction of capacity goes hand in hand. I suspect that the same battery model has a certain rate that is the same (assuming they werent ever over/undervolted).  Wouldn't that mean that if we build a battery pack from cells with varying capacity and IR (but identical cells), sorted so that each group has same capacity, that we will end up with similar IR for each group of cells and that they will discharge at approx the same rate?

I like the idea of using LG MH1 cells with a bit different IR. Balance charged of course. Now I'm worried about discharge.

If my suspicion is true or fairly true it means I only need to measure the capacity of all the cells and group them by capacity to get a nice somewhat balanced pack (for discharging). Later when balance charging it, any self discharge imbalance will be taken care of. 

Everytime I hook the battery up for charging I can see which groups are weaker and can one day choose to service them. Take cells out, remeasure them, replace one or two of the weakest cells.

I like that balance chargers have a storage mode. For cells waiting to be used at a later time I can store them at a certain voltage ready for the next servicing.

/a

[esaj]

On 7/21/2018 at 3:47 AM, alcatraz said:

We all know charging IR (internal resistance) mismatched cells in series is bad.

What about discharging mismatched chains? Imagine every cell(group) has the SAME capacity but has a mix of different IR cells?

Would the low IR groups get discharged to lower voltages before the high IR groups? Or are we only talking the self discharge perhaps?

It seems to me that the increase of IR and reduction of capacity goes hand in hand. I suspect that the same battery model has a certain rate that is the same (assuming they werent ever over/undervolted).  Wouldn't that mean that if we build a battery pack from cells with varying capacity and IR (but identical cells), sorted so that each group has same capacity, that we will end up with similar IR for each group of cells and that they will discharge at approx the same rate?

I like the idea of using LG MH1 cells with a bit different IR. Balance charged of course. Now I'm worried about discharge.

If my suspicion is true or fairly true it means I only need to measure the capacity of all the cells and group them by capacity to get a nice somewhat balanced pack (for discharging). Later when balance charging it, any self discharge imbalance will be taken care of. 

AFAIK, usually the internal resistance goes up as the cells age, and the total capacity the cell is able to store goes down, but whether these are "linked", I don't know. Higher IR cells would sag more (drop more voltage) over themselves when high current is passing through (be it charging or discharging), burning off more of the energy as heat inside the cell, so they get more stressed. Could be that they also drop their "resting" voltage faster then, getting out of balance faster than the rest with lower IR.

 

Quote

Everytime I hook the battery up for charging I can see which groups are weaker and can one day choose to service them. Take cells out, remeasure them, replace one or two of the weakest cells.

This reminded me, how are you connecting the cells? Do you have a battery spot-welder? With the currents used in the wheels, no plastic holder can be used (it will likely overheat and melt), and the cells have to be connected "strongly", usually they're spot-welded through nickel strips:

Soldering might be an option, but the connection has to be good and low resistance, not allow the cells to move much (so that they don't tear the connection over time with vibration & bumps), and care has to be taken during soldering not to overheat or short-circuit the cells. The cardboard (or other material) insulators sitting on top of the cells under the nickel strip is there for a purpose, the outer edge of the plus-side of the cell is actually the casing, acting as negative-pole:

It's actually "negative cathode", but you get the idea... if you accidentally allow solder or the strip to bridge the gap there, the cell is short circuited and starts to overheat.

 

 

[alcatraz]

Yeah I just took delivery of a small spotwelder that can do 0.15mm nickel strips. Would need two layers everywhere.

Heat glue on the batteries or just held together by the welds?

I'm thinking that if I go the route of trying to make use of used cells and testing them and grouping them, I would then need a stategy for handling low voltage groups. These are the groups that require me to balance the pack much more frequently than the average groups would. I'd need to break the spotweld connections on those groups, measure them all and replace the worst cell and replace it with a much newer one. This way that voltage curve will be given a boost.

This could be a yearly or every second year winter activity. After the average capacity is down to 2300mAh, I get completely new cells.

To balance the pack I don't want to charge it fully. I'm thinking of using (monthly) a portable passive balancer 500-1000mA balancing current and do a set of 8s one night, and another set of 8s another night. (or just buy three of them and hook them up at the same time, they just discharge the high voltage groups to the lowest one, i hope to be able to find one where I can set the discharge voltage myself so I can set all three balancers to the same voltage)

This way I don't need a balance charger and risk blowing stuff up. Or fear of leaving it all unattended. I already ordered a 3 amp 80-84v charger that I can set the voltage on.

When the cells are matched I can have the voltage higher at 82-83v maybe but as they get worse and worse I reduce the charger voltage to keep the shitty cells from going over 4.15v every time, and when I'm down to like 80v and balancing every 2 weeks then it's time to service the pack.

That's what I've found to be my compromise. Still need to get all kinds of tools like passive balancers, desktop 18650 charger to assess resistance and capacity, spotwelder. But after that I can pretty much mod any wheel battery pack the way I like and I can monitor voltages and balance safely (passively) whenever I need.

Lets see if this theory works in practice. ?

[mrelwood]

On 7/22/2018 at 4:53 PM, alcatraz said:

Lets see if this theory works in practice. ?

Which wheel and at what capacity do you plan to use your packs with? The battery packs are the heart of a wheel, and using mismatched cells might degrade the available power and capacity quite a bit. I'd also be worried if I hadn't closely tested the individual cell voltages under stress. A cell dipping under 3.0V regularly might be a dead cell pretty soon. It will then also kill the adjacent parallel cells. Mine did.

The cells are glued together then wrapped tight in proper packs. Hot glue might get quite soft if the battery temperature rises, so I would use special hi-temp glue sticks. The ones I have ("Rapid Pro") also stick a lot better than regular.

[alcatraz]

I will be using the inmotion v8 with 20s3p instead of the original 20s2p to begin with. Later I might experiment with making a 21s3p pack. But that's another story.

The used cells I'm getting are very low internal resistance and guaranteed over 3000mAh. 

They are also the same cells as the V8 comes with originally.

The wheel isn't very usable under 40% as it becomes sluggish. 

With the increased range of going 50% more cells and 33% less cell current I won't really have to ride under 40% charge very much. Staying over 30% reduces the chance of a single cell group going under my target minimum of 10% charge.

Chances are my current cells are in worse shape than the cells I'm getting. The plan is to put a balance wire harness (4x5s) so that I can monitor my voltages from time to time.

I'm aware that the cell groups that first get to 4.15v are going to be the same groups that are going to see low voltages first upon discharging. The idea is that by swapping weak cells around they will be protected by their stronger group siblings. Another method is to simply replace the worst cell in a problem group with a better one.

These cells are not ready for the bin right now. If I don't service the pack however, then the next 100 cycles is going to do as much harm as the first 500. By applying my methods and staying within 10-90% any further degradation will slow down considerably. 

When I say 10-90% I mean on a cell level. First monitor the voltages then I can set the end charging voltage to stay within 10-90%. I'll check the voltages after a depleting ride and after a full charge to see that all groups are within 10-90%. Maybe sometime in the future I'll build a 20s charger using transformers and single cell chargers. Just haven't found anything where I can set the target voltage to 4.10-4.15 instead of 4.2.

/a

[alcatraz]

I just thought I'd write another idea I had.

Because I want to keep the downtime to a minimum for my battery back rebuild I thought to share an idea I had.

Because I will continue to use the current cells I can't measure them before I spotweld the new pack together. Testing would take forever for 40 cells. 

My idea is to open the pack after a depleting ride and check cell voltages. Because I will increase cell groups from 2 to 3 cells. Whatever group has the lowest voltage is in need of the strongest cell to balance them.

So I will order the groups after their voltages. 

I will do the same with the new cells I plan to get.

Then the strongest cell in the weakest group, second strongest in the second weakeast and so on.

Not a bad idea I think and I don't need to measure 40 cells, and I can go ahead with building the pack right away after taking the old one apart. Thumbs up.