Effect of switching current on Li-ion battery packs

[VEGETA]

Hello,

I've nearly completed the design of a linear lab PSU with SEPIC switching pre-regulator which can deliver 30v 3 amps from 2S2P battery pack.

The pack should be around 8.4v maximum, say 8v typical... simulation in LTSpice is good.

However, the pack is about 5000 mAH or say 4500 mAH but the switching maximum current according to attached image is around 20-30 amps and zero amps.

So I wonder if this could be done? looking forward to your answers.


Note: green is final output current at 30v... and blue is the ripple switched current from the battery pack (nevermind the negative since it is probed that way).

[fcb]

It will depend on the inductance of your battery pack and switching frequency.

But I would expect significant heating from your pack - can you try reducing the ripple with capacitors on the input of your converter?

[Rigolon]

Batteries are really complex if you go deep in how they work, and i'm not an expert but here is my thoughts on the subject.

In a 2S2P battery pack you have 4 batteries. 2 pairs connected in series, each pair consists in 2 batteries connected in parallel.
So to have 5000mAh 2S2P battery pack, each battery has the capacity of 2500mAh. This is important, so keep that number in mind.

The capacity of the battery represents the value that would fully discharge a battery if draw constantly in 1 hour. (There are more variables to this but I will not get into that). Let's say a 2500mAh means you can supply 2.5A for an hour and then the battery dies, if you supply double (5A) the battery dies in half an hour.

But here is the catch.
The thing is that the capacity of the battery it's not a real value for every case. Usually lower consumption means even more time that can be used and higher means less time. You would expect that in a 2500mAh battery that supplies only 250mA you would get 10 hours, but in reality you can get more, lets say 11~12hours. The same applies for higher consumption. If you go 10 times the capacity current (25A) you'd believe that you can supply current for 1/10 of the time (6 minutes). But in reality will supply less, perhaps 3 to 4 minutes or so. All these values are only illustrative as this characteristics changes with the type of batteries and a lot more specs.

Although when using high frequency the batteries will show behavior of capacitors that would allow higher current output, I wouldn't think it would be able to output 10 times higher the value specified in the capacity for much longer than in constant DC.

Also there is a max current output that over this value will damage the battery, this are all specs to care for.

I'm not saying that you wouldn't be able to do what you want. But as I'm not sure that you will, I'm sharing what I know about the subject so you can check a few things before risking blowing up your battery pack.

EDIT: my knowledge is more related to lead-acid batteries and they are really different than Li-Ion batteries. As I checked a few information it seems they don't lose much capacity even when discharging with 10C rate (10x 2500mA). And since you are using pulsed I guess you won't have much problems.

But on important thing to remember is that when using pulsed discharging the battery allows less cycles (life-time) than rather using constant low DC consumption.

[VEGETA]

It will depend on the inductance of your battery pack and switching frequency.

But I would expect significant heating from your pack - can you try reducing the ripple with capacitors on the input of your converter?

Hmmm for the absolute state which is 30v 3a total output power would be 90 watts. The design is optimized to reduce heating by using the pre-regulator at 1v difference from linear one which is done using ltspice and all loops are compensated for stability.

for the pack it would get 20 amps of pulses but also 0 amps within the same pulse... shouldn't this count? we know that in order to get overheating you need prolonged overload current but this is not the case here.

Notice that this is the maximum case, normal uses are way below this. However, I wanted to make sure this works.

if you supply double (5A) the battery dies in half an hour.

Yes, but that is the user's responsibility right? i mean he knows he will drain his batteries fast. However, my responsibility is to make it safe and working perfectly without errors.

Although when using high frequency the batteries will show behavior of capacitors that would allow higher current output, I wouldn't think it would be able to output 10 times higher the value specified in the capacity for much longer than in constant DC.

I put 2 in series and 2 in parallel to make it 5000 mah to allow more capability. Now each pair should get 10 amps in pulses.

[fcb]

It will depend on the inductance of your battery pack and switching frequency.

But I would expect significant heating from your pack - can you try reducing the ripple with capacitors on the input of your converter?

Hmmm for the absolute state which is 30v 3a total output power would be 90 watts. The design is optimized to reduce heating by using the pre-regulator at 1v difference from linear one which is done using ltspice and all loops are compensated for stability.

for the pack it would get 20 amps of pulses but also 0 amps within the same pulse... shouldn't this count? we know that in order to get overheating you need prolonged overload current but this is not the case here.

Notice that this is the maximum case, normal uses are way below this. However, I wanted to make sure this works.

if you supply double (5A) the battery dies in half an hour.

Yes, but that is the user's responsibility right? i mean he knows he will drain his batteries fast. However, my responsibility is to make it safe and working perfectly without errors.

Although when using high frequency the batteries will show behavior of capacitors that would allow higher current output, I wouldn't think it would be able to output 10 times higher the value specified in the capacity for much longer than in constant DC.

I put 2 in series and 2 in parallel to make it 5000 mah to allow more capability. Now each pair should get 10 amps in pulses.

OK. I think you are missing the point of what I'm trying to say.

If you look at your graph, the blue trace is solid - this represents a dynamic signal (what frequency?) - basically you are drawing an AC current from your battery.  So what matters is how your battery is able to cope with an AC current!  And this will depend on cell construction, chemistry, state of charge etc..

If your battery has any inductance at all, then there will be an effect and you can't treat your load totally within the DC domain.

With sufficient capacitance (and the right type of capacitors) between the battery and the converter you will 'smooth out' the AC ripple - reducing the AC component you are presenting to the battery and then you are treating the battery more as DC device.

[VEGETA]

If you look at your graph, the blue trace is solid - this represents a dynamic signal (what frequency?) - basically you are drawing an AC current from your battery.  So what matters is how your battery is able to cope with an AC current!  And this will depend on cell construction, chemistry, state of charge etc..

No, the current is DC only. I told you the graph is just how I could probe in LTSpice. No AC current. The design is a bit unorthodox which uses a floating regulation with reference to the output... etc.

So kindly assume all DC current and rethink about it. Maybe my problem of probing misguided you.

[fcb]

If you look at your graph, the blue trace is solid - this represents a dynamic signal (what frequency?) - basically you are drawing an AC current from your battery.  So what matters is how your battery is able to cope with an AC current!  And this will depend on cell construction, chemistry, state of charge etc..

No, the current is DC only. I told you the graph is just how I could probe in LTSpice. No AC current. The design is a bit unorthodox which uses a floating regulation with reference to the output... etc.

So kindly assume all DC current and rethink about it. Maybe my problem of probing misguided you.

You've edited your original text. But still it doesn't seem clear to me that you are not presenting the battery with a large AC component - but if you say it's DC; then it's DC and will proceed as DC.

The answer is 'it depends' - it depends on the type of cells in your battery pack as to whether or not they will stand the high current discharge.  I've done tests on many types of Lithium cell in recent times and in general they are all capable of 5-10C discharge with some heating.  Cells with built in protection tend to crap out at between 1-2C discharge rate.

You'll just have to build the system and try the battery to see how the cell's react - you'd have no problem using some generic battery for a 1C discharge, but at the rates you want - you should defintly test/control what battery your clients use.

[VEGETA]

but if you say it's DC; then it's DC and will proceed as DC.

yes it is. It is a floating regulator design with the output positive is "GND" of the battery. However the load gets DC current as well as from battery. However, the direction depends on where and how you probe... thus magnitude is the only important thing.

I've done tests on many types of Lithium cell in recent times and in general they are all capable of 5-10C discharge with some heating.  Cells with built in protection tend to crap out at between 1-2C discharge rate.

I will use normal quality batteries assuming they are not protected, since there will be a charging and protection circuit available.

However, shouldn't the C rating be about the continuous current? we are talking about instantaneous current which is averaged through inductors (DC-DC converter) to give the continuous maximum current of 3 amps which is totally ok. any battery will be able to give such a current but the point is the instantaneous current pulses as shown.

In your tests, did you make 10C continuous current? what is the configuration (single or dual cells? parallel?)?

 

You'll just have to build the system and try the battery to see how the cell's react - you'd have no problem using some generic battery for a 1C discharge, but at the rates you want - you should defintly test/control what battery your clients use.

I should do tests of course. The ultimate solution is to put in another 2 cells in parallel which keeps total voltage at 8.4 and gets more current which should be enough... still using the exact same circuit. This adds cost to those who wanna get or make the circuit but the performance is gonna be a monster. Or better yet, they could choose to put only 2 or 4 batteries according to their uses... with maximum of 6 batteries.

[fcb]

However, shouldn't the C rating be about the continuous current? we are talking about instantaneous current which is averaged through inductors (DC-DC converter) to give the continuous maximum current of 3 amps which is totally ok. any battery will be able to give such a current but the point is the instantaneous current pulses as shown.

someone else pls

[VEGETA]

However, shouldn't the C rating be about the continuous current? we are talking about instantaneous current which is averaged through inductors (DC-DC converter) to give the continuous maximum current of 3 amps which is totally ok. any battery will be able to give such a current but the point is the instantaneous current pulses as shown.

someone else pls

Why are you saying this? I meant any li-ion battery (18650) can give 3 amps of current... is this weird??

about the 20 amps inst. current... yes it goes through inductors to be smoothed to maximum of 3 amps.... is this wrong?

please tell me if i made a big mistake.

[fcb]

OK. Lets try again.

Is this statement correct: 3A is your average (RMS) current draw from the battery and this is being drawn from the battery in short pulses (PEAK) of perhaps 30A.

[fcb]

C is a ratio, and is calculated thus: (dis)charge divided by rated capacity of the battery.

So 0.4A discharge current of a battery rated at 4Ah would be called a "0.1C discharge" and should exhaust the battery after 1/0.1C=10hours.
Likewise charging a the same battery at 2C would mean charging the same battery at 8A and result in a 30 minute charge time.

My tests where on various 18650 cells rated at 2 to 2.8Ah.

[ogden]

With sufficient capacitance (and the right type of capacitors) between the battery and the converter you will 'smooth out' the AC ripple - reducing the AC component you are presenting to the battery and then you are treating the battery more as DC device.

Exactly. Next thing to do - redo simulation. Introduce real capacitor(s) between battery and converter, use real components with real impedances everywhere including battery and wires.

[VEGETA]

OK. Lets try again.

Is this statement correct: 3A is your average (RMS) current draw from the battery and this is being drawn from the battery in short pulses (PEAK) of perhaps 30A.

When I set the final voltage at 30v and final current limit to a little bit more than 3 amps (to allow 3 amps) + at 10 ohms of load... I get 30v and 3 amps of output through that 10 ohms load. this is supposed to be the absolute maximum rating ever of this design. Normal uses are way below this.

At this statuation... switching current through the pre-regulator taken from the battery is about 20-30 amps of pulses as seen in the picture and attached complete ltspice simulation file.

____

My tests where on various 18650 cells rated at 2 to 2.8Ah.

so these 2000 mah batteries gave you +10 amps of continuous current? no severe voltage drop or something?

Exactly. Next thing to do - redo simulation. Introduce real capacitor(s) between battery and converter, use real components with real impedances everywhere including battery and wires.

I have attached the simulation so you can check yourself. it is very well done and compensated through help from a certain member in this forum long time ago... I spent sleepless nights to learn and modify it until we made it stable and working.

kindly check it, it is slightly modified to use consolidated parts like 40k resistors (in real-life would be 4 of 10k in series) instead of 43.2k which gave 300khz switching frequency while 40k gives slightly more than 300khz. this has no major effect on the design.

Also, the design uses not-orthodox type of grounding and referencing since it is a floating regulator... so when you probe, keep holding mouse left button and drag to get the drop voltage needed. You'll know when you see.

thanks!

[ogden]

I have attached the simulation so you can check yourself.

Yes indeed you have zero impedance battery, ideal (and small) capacitor after battery, no battery wires "modelled", L3 & L4 also are ideal. You would want to update circuit and run simulation again.

That "not-orthodox type of grounding" is  . What's the problem to make OUT- as simulation ground?

[VEGETA]

I have attached the simulation so you can check yourself.

Yes indeed you have zero impedance battery, ideal (and small) capacitor after battery, no battery wires "modelled", L3 & L4 also are ideal. You would want to update circuit and run simulation again.

So what is your suggestion? i posted the spice simulation so you can modify and test your ideas to see if it fits. I am open to these modifications.

That "not-orthodox type of grounding" is  . What's the problem to make OUT- as simulation ground?

The circuit is indeed not so straightforward, if you are interested and have time to waste then check this thread were this circuit was developed from scratch.

The circuit makes the GND as the reference and as Vout+ itself. There is a DC-DC isolated module which acts as an isolated power source to power op-amps, MCU, LCD, and control circuit (DAC,ADC,etc..) which has its negative reference at the Vout+ itself... this is called the floating regulator. I learned this idea from Kleinstein here in this forum as he witnessed it in commercial supplies. That is all the trick in the circuit, the rest is normal.

The topology of this circuit is not really the most important here but rather the pulsing current, whether 2s2p pack of 18650 can handle it or not.

[ogden]

So what is your suggestion? i posted the spice simulation so you can modify and test your ideas to see if it fits. I am open to these modifications.

It is up to you to find out & introduce impedance of your battery, wires, said capacitors & inductors. Don't ask me.

The circuit makes the GND as the reference and as Vout+ itself.
There is a DC-DC isolated module which acts as an isolated power source to power op-amps, MCU, LCD, and control circuit (DAC,ADC,etc..) which has its negative reference at the Vout+ itself... this is called the floating regulator. I learned this idea from Kleinstein here in this forum as he witnessed it in commercial supplies. That is all the trick in the circuit, the rest is normal.

I am afraid that you confuse real world ground with simulator ground. Those can be two different things. If your simulation ground (triangle symbol) placement actually breaks simulation - you are doing it wrong. Just introduce something like "GND1" for MCU, LCD / whatever (name that floating ground "GND1") but connect simulator ground to OUT- where it shall be.

[edit] Whether 2s2p pack of 18650 can handle pulsing current you will ask *after* you correctly simulate that pulsing current.

[VEGETA]

It is up to you to find out & introduce impedance of your battery, wires, said capacitors & inductors. Don't ask me.

Ok, I will but I wouldn't mind a hint.

I am afraid that you confuse real world ground with simulator ground. Those can be two different things.

Real world ground (triangle symbol) is exactly where I want it to be. The ground of the design is the output positive itself.

If your simulation ground (triangle symbol) placement actually breaks simulation - you are doing it wrong.

Simulation works and it is not broken. As I told you, the design is exactly as we wanted it to be.

Just introduce something like "GND1" for MCU, LCD / whatever (name that floating ground "GND1") but connect simulator ground to OUT- where it shall be.

Hmm what should be the benefit of changing the terminology if the final result in real world is the same and simulation works?

[edit] Whether 2s2p pack of 18650 can handle pulsing current you will ask *after* you correctly simulate that pulsing current.

I fail to understand why is reversing the direction of current or changing the probing method will affect this? the current pulsing is in the design itself as the LT3757A explains and we picked the 0.003R shunt resistor to allow such high current peaks in order to be able to output enough final output current.

[ogden]

Real world ground (triangle symbol) is exactly where I want it to be.

 

Triangle symbol is not real world ground in LTspice. It is simulation ground. There is no real world ground as such in LTspice. Decision of which wire is real world ground is up-to you.

I fail to understand why is reversing the direction of current or changing the probing method will affect this? the current pulsing is in the design itself as the LT3757A explains and we picked the 0.003R shunt resistor to allow such high current peaks in order to be able to output enough final output current.

There are two problems in your simulation: #1 use of ideal, non real components #2 wrong inconvenient for simulation connection of LTspice ground. Problem #2 has nothing to do  with current pulsing.

Hmm what should be the benefit of changing the terminology if the final result in real world is the same and simulation works?

Changing simulator ground changes ground reference of simulation voltages. Illogical simulation ground connection discouraged me to even press "run" button because I knew "measure" tool have wrong reference ground.

[fcb]

Running your simulation (as supplied), if you look at the current draw on the power source (battery) at say 4.85ms, you get this:

You'll see that the current is being drawn by your converter in pulses. Any inductance in the supply to this (e.g. battery chemistry, construction, leads, etc..) will have an effect on this ability to draw current and likely affect your stability/function of your converter.

If I put in a small ESR for the battery (say 0.02ohms) and just look at the ripple current that C13 (which says 4.7uF x2 - but is actually set to 4.7uF not 9.6uF?):

Peak currents of around 3.6A in C13 - that needs to be a pretty special 4.7uF capacitor to take that ripple. I would increase it and model is with at least it's ESR included.

Here's what happens if you change:
C13 to C1210C106K3RAC 10uF from KEMET (Mouser stocked part)
L7 (wire), 100 mm of 4mm CSA wire (https://www.eeweb.com/tools/wire-inductance)

You'll notice that the ripple current experienced by the battery is much lower. Perhaps adding a bit more inductance (a filter basically) will reduce this further (lower EMI, lower heating due to AC component in battery). I'd probably add a delibrate PI filter, or perhaps add a capacitor across the battery directly to create this filter with any wiring.

You'll have to do some calculations with C13 to see if it will take the ripple (clue: Dissipation Factor!)

[GeorgeOfTheJungle]

The topology of this circuit is not really the most important here but rather the pulsing current, whether 2s2p pack of 18650 can handle it or not.

Many 18650s can handle 15..20 A max constant current nowadays (e.g. LG INR18650HG2), but you've got to look that carefully in the cells' specs: Intended Use (power tools) and Maximum Continuous Current.

In general the rule (for 18650s) has always been more current (max A) => less capacity (Ah/Wh), and in this sense 4S is better than 2S2P, and 2S2P better than 4P, because 4S lets you get the same watts with less amps.

At those currents you won't get anywhere close to the nominal capacity (Ah/Wh) because the nominal capacity is usually calculated at less than 1C. For example, that INR18650HG2 has 3Ah @0.6A, but I2R losses at 20A are at least a THOUSAND times more than at 0.6A... ((20/0.6)²)

At those currents life expectancy is a disaster.

[VEGETA]

Kindly see newer modified version in attachment, now total battery peak currents are about 1 amps of ripple (14.5A - 13.5A) instead of pulses of 20 amps and 0. Nice improvement right?

what I did is adding around 50nH of inductance with 0.005R series resistance which is the PCB trace since I will surface mount the regulator very closely to the surface mounted batteries... so it would be even better than this.

Combine that with now 3 of 10uF capacitor that you choose to get this result with about 3 amps of these capacitors ripple current current (assuming I could get cheaper ones from LCSC in a big reel with same specs or close). I guess this is what you meant previously right? but your capacitor ripple is lower.

What I have mistaken hugely is the switching current at Isense and Rsense resistor of the LT3757A which is required... I assumed that the current drawn from the battery should be similar... but you guided me that it shouldn't.

At those currents life expectancy is a disaster.

Now maximum battery pack current is about 15 amps continuous which means each battery gives 7.5 amps, this is much better. There is a very short (~3 ms) of more current at first but this is not continuous.

and in this sense 4S is better than 2S2P, and 2S2P better than 4P, because 4S lets you get the same watts with less amps.

I thought about this solution and maybe it is better overall? luckily the design allows this flexibility, any configuration works since it is a wide input SEPIC design... However I will need to adjust the charging and protection circuit for the complete circuit according to the final decision... which has nothing to affect this main circuit.

Do you think I could pull maximum of 30 volts 3 amps out of say 4S1P cells of ~ 2000-2500 mah? the design allows this assuming the source is capable.

[ogden]

As you see - capacitors indeed reduce power supply ripple Yes, high current Li-Ion cells can sustain currents and discharge rate above 1C. There are even 10C cells/packs. Problem with such - they are more expensive than 1C rate cells, they degrade faster. For example 18650 cell at 4C discharge (2500mAh @ 10A) at the end will be around 50..60oC degree hot. Frequently overheated cells do not live long.  (picture from internet search)

[VEGETA]

As you see - capacitors indeed reduce power supply ripple Yes, high current Li-Ion cells can sustain currents and discharge rate above 1C. There are even 10C cells/packs. Problem with such - they are more expensive than 1C rate cells, they degrade faster. For example 18650 cell at 4C discharge (2500mAh @ 10A) at the end will be around 50..60oC degree hot. Frequently overheated cells do not live long.  (picture from internet search)

Well, at 15A max we got 7.5A per cell which is doable even for Chinese 18650 batteries.. which is my aim. To be able to use this thing with any battery.

I made another simulation with 16v input (4 batteries in series) and the total current from battery was around 7 amps with peaks between 6 amps and 7 amps total.

So the question is: should I use 2S2P or 4P? since they are using same current per cell (~7A) -> at maximum ever state.

[BravoV]

Well, at 15A max we got 7.5A per cell which is doable even for Chinese 18650 batteries.. which is my aim. To be able to use this thing with any battery.

 

Seriously ? Especially using those "firey" brand 18650 cells like SureFire, MustFire and etc.