Sudden connection of Li-Ion cells in parallel

[opampsmoker]

Hi,

Hi, supposing I suddenly connect in parallel two 18650 size Lithium ion cells….one with voltage of 4.2V and the other of voltage 4V.

What will happen? How much current will flow between the cells and for how long?
What is the lowest voltage 18650 Li ion cell which can be placed in parallel with a 4.2V Li ion cell and not cause damage to either cell?

How much does it depend on internal resistance, and how much on chemical effects?

The connection will be  done very quickly with very large, 1 milliohm RdsON FETs (back-to-back)

Lithium Ion battery

[Siwastaja]

Never ever do that.

18650 DC ESR is approximately 50 mOhm (depending on cell type, age, temperature, etc., somewhere between 20 and 200 mOhm). Let's assume "worst case" (one producing highest current) 20mOhm.

Assuming 1mOhm Rds(on) switch and some 5 mOhm in wiring, the current will be

I = U/R = 200mV / (20mOhm + 20mOhm + 1mOhm + 5mOhm) = 4.3A.

18650 cells are typically rated to some 2A - 3.5A maximum charge rate (or about half of that for long cycle life), so you are clearly exceeding it.

It's likely there is no catastrophic incident or even serious damage if the rating is exceeded by some 30% done rarely, but do yourself a favor and recalculate with different conditions, like one cell at 4.2V and one at 3.5V and see what happens!

You need protection logic which only allows connection when the voltages are "close enough". How close? Do the math as shown above. 100mV would be usually acceptable; 50mV would have some safety margin as well. Then you need to add actively current-limited power conversion to balance the cell voltages before connecting. Getting quite expensive. Finally, add classical fuses to protect the cells in case your protection logic goes haywire.

But usually, the answer to this classic X-Y problem is: you don't have to do shit like that. Just connect the cells together permanently, creating a pack that is big enough for the job.

[MosherIV]

What will happen? How much current will flow between the cells and for how long?
What is the lowest voltage 18650 Li ion cell which can be placed in parallel with a 4.2V Li ion cell and not cause damage to either cell?

The cell with the higher voltage will discharge into the cell with the lower voltage until the 2 cells equalise in charge (ie their voltages match).
As you have already hinted at, the balancing current is limited by the cell internal resistance.
This may well exceed the cell rated current or C rating, therefore may damage the cell.

Cell should alway be matched in voltage, to mV, before they are joined in parallel or the balancing current may damage 1 cell.

[opampsmoker]

Hi,
Thankyou , please could you also discuss for the following cell

https://www.18650batterystore.com/products/samsung-25r-18650

[tunk]

There's a test of a cell with the same model number here:
https://lygte-info.dk/review/batteries2012/Samsung%20INR18650-25R%202500mAh%20%28Green%29%20UK.html

[Siwastaja]

Samsung 25R is a so-called "high power" cell for power tools, from about a decade ago. 29Q is a newer power tool cell with higher capacity but likely somewhat higher cost $/kWh.

I have personally measured the DC ESR of this particular 25R cell to be approx. 25 mOhm at around 30degC, if I recall correctly.

The cell type doesn't make that much of difference; a cell with lower resistance of course puts out higher current under such condition, but OTOH, such cells can typically also take higher current safely!

So all in all, I reiterate my suggestion to match the cell voltages within +/- 50mV before connecting in parallel. Maybe +/- 20mV if you want to be extra careful.

There is no need to match to 1mV accuracy. Cells don't even come out of the factory to this accuracy, and they are supposed to be connected in parallel by pack manufacturers as-they-are (of course, verification step of voltage is highly recommended).

[NiHaoMike]

Some way to soft connect the cells would ensure it will be OK regardless of the charge levels. That could be by operating the MOSFETs as linear loads or by using PTCs.

[ejeffrey]

To reiterate: this is not something you should be doing regularly.

Either connect the cells together permanently at manufacture time and leave them that way forever or treat the cells as independent storage elements with their own battery management and the ability to route power from either to your device.

It is also ideal if you only connect identical new cells together in parallel.  If you connect used cells in parallel or used with new, even if you match voltages the cells may have different internal resistance.  This will cause poor current sharing.  If you limit your charge and discharge currents to what a single cell can tolerate you should be fine but you don't want to do this in a scenario where you are going to be trying to fast charge the pack.

[opampsmoker]

Thanks,
Is There a close model of an 18650 Li-ion cell for LTspice so I can simulate suddenly paralleling two Li-ion cells? Or simulate say suddenly  putting a Li-ion cell in parallel with a layer of Li-ion cells which are at a higher voltage than it.
I think the cell model that’s “shipped” with LTspice is just an ideal voltage source, and so isn’t like a real cell. For a real cell, I think the internal resistance doesn’t exactly apply when a cell is charged. This is because the electric current is kind of formulating the cell chemistry, so in fact the internal resistance is much less when charging, would you agree?....i don’t think the LTspice cell model exactly and correctly simulates the electrochemical situation of a cell?

[Siwastaja]

Voltage source with 25mOhm series resistor is good enough model for this purpose.

To simulate full cell, set the voltage source at 4.2V.

To simulate empty cell, set the voltage source at 3.2V.

For intermediate states, interpolate between these voltages.

Although, with such simple model, why do you need Spice? It's elementary school level math.

[opampsmoker]

Thanks..but i suspect internal resistance is complex...

The Samsung “INR18650 25R” Li-ion cell datasheet says that the 1kHz internal resistance is 13 milliohms, but the DC internal resistance is 22 milliohms….one then extrapolates that at 1MHz the internal resistance would be well under 13 milliohms…..would you agree?

INR18650 25R   cell datasheet
https://www.powerstream.com/p/INR18650-25R-datasheet.pdf

….Now suppose a sudden pulse of current suddenly flows into the  INR18650 25R   cell. Now, the rising edge of a sudden  “step” pulse current has extremely high frequencies associated with it…potentially well above 1MHz….so one would conclude that the internal resistance to a very fast rising edge pulse would be extremely low….well under 13 milliohms…..would you agree?

[trobbins]

There is a difference between a step pulse and a small signal sinewave impedance measurement (whether that small signal measurement is at a low frequency, high frequency or the sum of many frequencies).

A step pulse starts to change the voltage source level over time, and the ESR measured under large-signal conditions is likely to be fairly constant across a wide range of operating SoC, and current level.

[opampsmoker]

Thnaks,
For the INR 18650 25R Li ion cell, i have attached what i believe to be accurate equivalent circuit parameters for the Li-ion cell INR18650 25R

INR 18650 25R Li ion cell datasheet
https://www.powerstream.com/p/INR18650-25R-datasheet.pdf

I parallel them suddenly and well over 4A flows for 20-30 seconds....this shows surely, that sudden paralleling of Li-ion cells is a nono.
Would you agree?
LTspice and pdf schem attached.

[trobbins]

opampsmoker, perhaps it's a good idea to add technical and summary data about your test to make the post informative to read, without every reader having to scurry away and look at links and and assess circuits etc just to work out what you have done and what it really means.

You don't say what the resting voltages of the two cells were, or the effective shorting resistance (or a photo), or a plot of the current flow and how you measured current, or a plot of each cells terminal voltage, or if the measured current was as expected from cell and circuit resistances, or summarise how the measured current related to the datasheet max current rating.

You may also want to revisit your first post and add in your latest 'answers' to each of your initial queries.

[ejeffrey]

Thanks..but i suspect internal resistance is complex...

The Samsung “INR18650 25R” Li-ion cell datasheet says that the 1kHz internal resistance is 13 milliohms, but the DC internal resistance is 22 milliohms….one then extrapolates that at 1MHz the internal resistance would be well under 13 milliohms…..would you agree?

That doesn't matter for this purpose.  Just the DC resistance is fine.  Maybe there will be a surge past there for a millisecond or so?  That isn't going to matter.

Shorting two random lithium ion cells with unknown histories and state of charge together is not a "precision operation."  You are overthinking this.  Everyone here had provided the reasons this is inadvisable and also how to go about it if you still need to.  Either do it or not, but making spice models is only going to make meaningless plots.

[Siwastaja]

Say, if your load is a motor controller for example, you'll be running it at least at 10kHz or maybe 20kHz, the inductance of the motor limits ripple current amplitude (motor basically runs in CCM with a large DC component) and the DC link capacitance is taking care of most of that ripple current; battery sees basically DC. If you have sudden accelerations, maybe you can say battery sees 1Hz. But the ESR at 1Hz is already quite close to the ESR at DC (or say, 0.05Hz, because true DC doesn't exist and can't be measured).

If you are building a mobile phone with wifi and need to save space and cost of capacitors, then your battery might be seeing considerable 1ms peaks and you could benefit (i.e., get more accurate result) from modeling the battery 1kHz impedance.

But generally, the 1kHz impedance of the battery is lower than the DC ESR, and most applications have significantly higher currents at DC than at 1kHz, so using (only) 1kHz impedance instead of DC ESR generally results in overestimation of battery capability to deliver current, underestimation of voltage drop, and a sad failure unless you have ample margin to compensate against using the wrong parameter. I have seen this so many times yet certain people have a fixation to model the battery ESR over the whole frequency range by using just the datasheet or measured value of 1kHz AC Z real part.

OTOH, if you have a very specific application which takes considerable current at say 100Hz to 5kHz, if you still use just DC ESR (higher value than 1kHz Z) to model it, the error plays in your favor adding extra margin.

By all means create a complex model, but if you understand your application, you may be able to use very simple model instead, with less room for human error.

Also remember that temperature and SoC affect the DC ESR more than they affect the 1kHz Z real part. Those who want to play with meters like having more "stability" in readings; yet the DC ESR defines actual in-application voltage drop and power loss, which are what a system designer should be interested in. So remember the battery can't deliver as much current when near empty, or at cold temperatures.

[Siwastaja]

The Samsung “INR18650 25R” Li-ion cell datasheet says that the 1kHz internal resistance is 13 milliohms, but the DC internal resistance is 22 milliohms….one then extrapolates that at 1MHz the internal resistance would be well under 13 milliohms…..would you agree?

Of course not. At 1MHz, inductance increases the Z significantly, and skin effect increases the real part of Z (the ESR).

But no one would pull or push currents at 1MHz to/from the cell; if nothing else, it's already an EMI issue with such large loop. Use bypass capacitors to supply currents above a few kHz.

The sweet spot for lowest ESR is approximately around 1 to 10 kHz. Which is also one of the reasons manufacturers like to publish that number in datasheets.

[opampsmoker]

OK Thanks, I take your point about the 1MHz situation etc.
The attached LTspice simulation shows the sudden paralleling of two Li-ion cells. (their equivalent circuits).  Prior to paralleling,  one cell is at 3.2V, and the other is at 4.2V…..in spite of this, the “equalisation current” only flows for 109 seconds. Surely this must be wrong? (ie the Li-ion cell model as in the attached sim and pdf, must be wrong?)  In real cells the “equalisation current” would surely flow for far longer?
This  “equivalent Li-ion cell” model was taken from a respected professional text from Pennsylvania State university.

[trobbins]

opampsmoker, can I suggest that you have the cells, so can make actual measurements, and report back to this thread on how much the measurements deviate from the model you are using.

[KE5FX]

Then he'll have to change his handle to batterysmoker... 

[ejeffrey]

OK Thanks, I take your point about the 1MHz situation etc.
The attached LTspice simulation shows the sudden paralleling of two Li-ion cells. (their equivalent circuits).  Prior to paralleling,  one cell is at 3.2V, and the other is at 4.2V…..in spite of this, the “equalisation current” only flows for 109 seconds. Surely this must be wrong? (ie the Li-ion cell model as in the attached sim and pdf, must be wrong?)  In real cells the “equalisation current” would surely flow for far longer?
This  “equivalent Li-ion cell” model was taken from a respected professional text from Pennsylvania State university.

Please put down the simulator and step away from the computer.  You can't use a simulator if you don't know what you are doing.

Just because a model comes from a "respected professional text" doesn't mean that it is applicable the way you are trying to use it.  A battery is not a capacitor, and any model based on capacitors is _obviously_ wrong.  Indeed any linear model is wrong.  That doesn't mean it is bad: it might still be useful when used as intended but if you pull it out of a book and just do a simulation you are likely to get nonsense.  In particular, when you are looking at operation over the full SoC state from 3.2 to 4.2 V you can't use any sort of linear circuit model and expect to get a correct quantitative result.

You can work out the important qualitative characteristics on a piece of paper and looking at the discharge curves in the datasheet.

[Siwastaja]

To solve most product-level li-ion design issues, including this particular problem, you don't need complex modelling. All you need is a set of good parameters, understanding them, and using standard ohms law. Napkin is enough, but I prefer Excel spreadsheets very much because they allow you to change values retaining formulae, and compare different options.

You definitely do not need Spice for this.

The formulas you need are: U = RI, P = UI, and all the combinations thereof.

Add the understanding that the Samsung 25R is:
* at about 4.17V with zero current when full
* at about 3.4V with zero current when empty
* ESR at room temp is about 25mOhm when between 40%-100%
* the ESR rises to some 50mOhm when approaching empty
* ESR rises considerably from the aforementioned values when the cell is cold. Rough guesstimate doubling every 15 degC? Check the actual value out somewhere if important.
* the ESR rises when the cell ages until it's considered End-Of-Life when the ESR has doubled after maybe 5-10 years depending on the use

Now you have parameters to plug in the Excel.

And 99.99% of battery models available or discussed in academy do not show these crucial effects of temperature and aging but focus on high-frequency behavior of the cell which is of no interest for most application designers, because their products take DC current in.