Fechter's Capacitor Coupled Cell Balancer

[fechter]

OK, here's the beginning of another half-baked idea. One of the major issues with most BMS/cell balancing schemes is the amount of heat given off by the shunt resistors. Various other approaches have been tried to avoid this, like switched capacitor schemes or separate dc-dc converters for each cell.

The separate dc-dc coverters for each cell approach has been proven to work but is a bit bulky and expensive.
Switch capacitor charge pumping schemes work OK at very low currents, but don't work so well at 1 amp and have a hard time shuttling charge from one end of a pack to the other.

So here's my latest half-baked idea (the half baked ones appear late at night when I go into sleep deprivation mode):

Capacitor Coupled Balancer 1.jpg

(86.64 KiB) Downloaded 2 times

If we put a bridge rectifier on each cell fed by a capacitor coupled bus line. The coupling capacitors will pass an AC signal, but block the DC present from the cell string. If the frequency is very high (over 100KHz?), then the coupling capacitors can be quite small. The coupling capacitors need to be rated for for the full string voltage.

The balance supply would be something like a typical 5v switching mode power supply, but we would take the output coming directly from the transformer, before the rectifier. I think most any switching power supply could be modified to work like this. A single voltage adjustment on the balance supply works for all cells. The balance supply need to supply enough to overcome the forward loss in the bridge diodes.

To the balance supply, the load would 'look' like a bunch of cells in parallel. Whichever cell had the lowest voltage would present the path of least resistance, so would take the most current. This balancing current could run simultaneously with the bulk charge current running through the battery string.

While perhaps not fancy, this topology looks to me like it should work (that usually gets me in trouble). This would need to be combined with a more conventional looking bulk charger and cell level voltage protection.

Now, where's my breadboard?

[johnrobholmes]

In!!! Sounds like the perfect concept for fast charging with integrated balancing!

[GGoodrum]

Yes, I think this is a very clever way way indeed, one with very few "moving" parts. Worth a try, for sure.

-- Gary

[liveforphysics]

Everyone of the de-coupleing caps would need to be rated for the full pack voltage+20% or so, and to pass meaningful current, the freq needs to go way above 100hz unless you plan to use giant low ESR caps.


I can visualize how the cells could all be charged with out removing the series connections from one 4.2v high freq AC supply, but I'm unclear how it would balance during charging.

[MikeB]

One of the major issues with most BMS/cell balancing schemes is the amount of heat given off by the shunt resistors.

I can't voice an opinion about the design, but I really hope you guys can make this work. In my opinion, the heat given off by the shunt resistors isn't nearly as big a problem as the wasted energy. While my new electric scooter might consume just 1kWh for my daily commute, the energy wasted during balancing meant that I had to spend 2kWh just to get it back to a full and balanced battery pack each day.

[fechter]

Everyone of the de-coupleing caps would need to be rated for the full pack voltage+20% or so, and to pass meaningful current, the freq needs to go way above 100hz unless you plan to use giant low ESR caps.


I can visualize how the cells could all be charged with out removing the series connections from one 4.2v high freq AC supply, but I'm unclear how it would balance during charging.

I'm thinking it would run at a pretty high frequency, like over 100kHz, so the caps don't need to be very big, and they could be super low ESR MLCC or something like that. The diodes would need to be Schottkys.

As far as balancing during charge, the lowest voltage cell will tend to draw the most current as it will have the lowest impedance to the balancing source. This will happen from the beginning of charge, so maybe additional time at the end of bulk charging may be unnecessary.

[johnrobholmes]

In essence, the diode and cap array allow you to charge from both parallel and series connection on the pack. Just two connections to deal with.

[fechter]

That's the idea anyway.

The balance charger will see all the cells like a big bunch of parallel cells attached by diodes. The cell with the lowest voltage will naturally suck up most of the current. The voltage applied to the cells by the balance charger can be regulated at the charger, but I suspect the regulation won't be very tight.

At the same time this is going on, there's no reason you can't bulk charge. Most of the balance charger output will still go toward the lowest cell.

[johnrobholmes]

So is this considered analog, digital, or just a mix of parts?

[amberwolf]

Technically analog, since there is no logic control of anything, AFAICS.

Very interesting idea--I hope it works...after you fail to smoke a pack with it I'll see if I have some little rectifiers and stuff around.

[swbluto]

Sublime.

[number1cruncher]

Whichever cell had the lowest voltage would present the path of least resistance, so would take the most current.

Well, I could understand the schematic without any help, so it must be really simple. Plus it follows one of life's major rules...

How would this setup prevent overcharge of the cells? Would it be based on the power supply setting? Or is there some sort of governor in the circuit?

Thanks!

[swbluto]

Whichever cell had the lowest voltage would present the path of least resistance, so would take the most current.

Well, I could understand the schematic without any help, so it must be really simple. Plus it follows one of life's major rules...

How would this setup prevent overcharge of the cells? Would it be based on the power supply setting? Or is there some sort of governor in the circuit?

Thanks!

Presumably, 5v - 2 diode drops = 3.7 volts. I don't understand how it's supposed to limit the voltage at lower currents, however, since the diode drop is smaller at lower currents and so it seems like it has the potential to creep up to the supply voltage.

So I'd just read your last line. You state that this would be used with a regular bulk charger and cell voltage protection.

So, would it go something like this?

Attach regular charger to battery normally to bulk charge.
Detach when a cell's voltage get's too high.
Plug in the 5V AC source into this "balancer" to balance the cells quickly.

So, you'd need a bulk power supply AND a 5v AC source? What kind of balancing times would you suspect in normal use?

Also, how does this compare to the balancers offered at hobby-city? (Are balancers at hobby city just this topology with an integrated ~5V AC inverter?)

Now that I think about it, I suppose the BMS could be an integrated 5V source with capacitors, and then the BMS would be powered off the bulk charger and the BMS would 'choose' between balancing and bulk charging depending on the cell voltage detector. A 5V inverter from a variable DC input from a DC bulk charger isn't trivial, though, is it?

[number1cruncher]

Presumably, 5v - 2 diode drops = 3.7 volts. I don't understand how it's supposed to limit the voltage at lower currents, however, since the diode drop is smaller at lower currents and so it seems like it has the potential to creep up to the supply voltage.

Is the forward drop proprotional on a diode? I thought this was a characteristic of a resistor, but not a diode. However, I did have this argument in my head before posting earlier...

[swbluto]

Presumably, 5v - 2 diode drops = 3.7 volts. I don't understand how it's supposed to limit the voltage at lower currents, however, since the diode drop is smaller at lower currents and so it seems like it has the potential to creep up to the supply voltage.

Is the forward drop proprotional on a diode? I thought this was a characteristic of a resistor, but not a diode. However, I did have this argument in my head before posting earlier...

I just ran this through a simulation and surely enough, it didn't run upto the supply voltage when a cell was connected. I think the capacitor ensures there's some current draw when the voltage oscillates, so the current draw is "never" really low such that the diode's voltage drop is low. Wait... no, it's just one of those "It won't charge unless there's current going through the diode, but if there's current going through the diode, then the charge voltage won't be enough to charge the battery at a high enough battery voltage" oddities.

The drop of a diode depends distinctly on the current going through it. But it's different than a resistor because it's not linear with current, instead, at really low currents, it gets close to zero and then with meaningful currents it rises up the "voltage rating" and then sticks pretty close to it.

Here's a typical I-V curve for a diode.

[fechter]

The voltage drop in the diodes needs to be compensated for and would depend on load. I'm thinking there would be another bridge at the balance charger that would provide the feedback to the voltage regulator, so it would be close.

A separate HVC on each cell would still be needed to ensure no cell goes over voltage.

The balance charger would start at the beginning of the charge cycle and could run at the same time as the bulk charger.

There are still a lot of things that need to be figured out...

[Ypedal]

Alot of all that goes right over my head.. wooosh...

But, kinda sounds like what i'm doing with the voltphreak chargers an my soneil 5 amp units.

I hook up the soneil direct to the pack, and also hook up the voltphreak single cell chargers, in essence charging at 7 amps. as soon as the 1st cell hits it's full voltage the voltphreak charger goes green while the rest of the pack keeps on charging, within 30 seconds the entire pack is fully charged and ballanced as one by one the VP chargers lower the charge rate of each cell as they are full.. ( the soneil quits charging just shy of the full mark, this all works quite well )..

a similar setup permanently attached to the pack would be cool if it was reasonably sized. ie: slowing the charge rate of single cells in the pack instead of bleeding power away in resistors.

[scoot]

Glad you mentioned that Ypedal... from what little I could comprehend from this dialogue , that's kind of what I was thinking this concept was about as well. But like you, most of this circuitry discussion gets over my head. So if this proposed set up does indeed mimic your single cell charger array, I should think it is a truly elegant approach to balancing. So Fetcher, is that the idea more or less?

I am going to have to keep my on this thread... and now I'm subscribed to it

[number1cruncher]

Well, I'm definately an electric dummy and make it up as I go along sometimes... I would hope that the corrections from the more enlightened members balance my wing-it-ness.

I enjoy deciphering patterns, especially if I understand the basic stucture.

With Fechter's curcuit I see a capacitor off each leg of a bridge rectifier(4 diodes assembled, I think in a series loop? ). Capacitors to my understanding "smooth" the voltage. Maybe something to do with Ypedal's description of the tapering off?

However, I am wishful that this type of design could facilitate transfer of energy from the cells sitting at the finish gate, to the cells not yet there. Although that may be a "pump" style balancer that needs a processing chip with logic to operate.

We'll have to see what the more experienced have to say...

[fechter]

In this case, the capacitors allow AC to pass, but block the DC.

I did some initial testing using a junk switching power supply. I attached a pair of wires to the output of the switcher transformer to get an AC signal to work with.

This particular switcher runs at about 50khz. Feeding the AC through a capacitor, I measured how much power came through and tested the cap for heating.

With a 4.7uf capacitor, it looked nearly identical to a direct connection. This is encouraging. I also tried a 0,1uf, but there was a significant loss in output, though it was nearly 1/2 of the direct connection.

Another thing I figured out is the output at the cell end will be the peak-to-peak voltage of the AC waveform minus the diode drop. As the switcher regulates, the duty cycle of the on pulse increases, but the peak-peak voltage remains nearly the same. This means the AC source needs to be a bit more sophisticated than the output of a switcher transformer. My guess is you'd want to take a 5v supply and run it through a FET bridge to change it to a high frequency AC. This way, the voltage of the 5v supply could be adjusted to get the desired output at the cell. You could also crank the frequency up higher to reduce the drop in the capacitors or allow smaller capacitors and have a nice 50% duty cycle square wave.

This circuit won't suck power from the high cells and feed it to the low ones. It will only be able to charge the lower ones at a higher rate than the high ones. I think in the end, this will have about the same effect during a charging cycle. If the total imbalance is corrected during the same time that the bulk charge takes, then there won't be any need for additional balancing time.

[markcycle]

In this case, the capacitors allow AC to pass, but block the DC.

I did some initial testing using a junk switching power supply. I attached a pair of wires to the output of the switcher transformer to get an AC signal to work with.

This particular switcher runs at about 50khz. Feeding the AC through a capacitor, I measured how much power came through and tested the cap for heating.

With a 4.7uf capacitor, it looked nearly identical to a direct connection. This is encouraging. I also tried a 0,1uf, but there was a significant loss in output, though it was nearly 1/2 of the direct connection.

Another thing I figured out is the output at the cell end will be the peak-to-peak voltage of the AC waveform minus the diode drop. As the switcher regulates, the duty cycle of the on pulse increases, but the peak-peak voltage remains nearly the same. This means the AC source needs to be a bit more sophisticated than the output of a switcher transformer. My guess is you'd want to take a 5v supply and run it through a FET bridge to change it to a high frequency AC. This way, the voltage of the 5v supply could be adjusted to get the desired output at the cell. You could also crank the frequency up higher to reduce the drop in the capacitors or allow smaller capacitors and have a nice 50% duty cycle square wave.

This circuit won't suck power from the high cells and feed it to the low ones. It will only be able to charge the lower ones at a higher rate than the high ones. I think in the end, this will have about the same effect during a charging cycle. If the total imbalance is corrected during the same time that the bulk charge takes, then there won't be any need for additional balancing time.

This is the best solution to cell balancing I've seen to date.

[Toorbough ULL-Zeveigh]

it's a nice (and simple) idea, effectively substitute the shunt resistors in the conventional setup with a dissipationless Xc.
what i don't see how the voltages stack to increase from a single balance supply.
the problem i'm having is with the caps that hook to bridge neg, which i take it is supposed to incrementally raise the voltage floor to the next more positive stage.

where does the top of the uppermost cap ultimately hook to?

cuz it doesn't connect to anything right now, or is that one cap left off entirely?
sorry if that's a previously established given assumption i'm not aware of & would just like a little clarification.
with say a 4 cell balancer as pictured, what would the completed circuit look like, that might help me out.

also what kind of load did u run the test at?
can u really pull one amp thru a cap, if that's ur target?

[fechter]

where does the top of the uppermost cap ultimately hook to?

cuz it doesn't connect to anything right now, or is that one cap left off entirely?
sorry if that's a previously established given assumption i'm not aware of & would just like a little clarification.
with say a 4 cell balancer as pictured, what would the completed circuit look like, that might help me out.

also what kind of load did u run the test at?
can u really pull one amp thru a cap, if that's ur target?

I'm not sure I understand the question. The uppermost cap in the drawing goes to the next cell (not shown). At the top cell in the string, it just goes to the bridge.

Also not shown in the drawing is the bulk charger, which charges the whole string.

Yes, I can pull one amp through a tiny little capacitor without it getting hot or blowing up. The limitation will be heating in the diode bridge. I found some tiny little 2A shottky bridges that would be in the ballpark.

[fechter]

OK, I redrew it so that it is a bit easier to understand (hopefully):

I left out the capacitors that go across each cell. Those may be unnecessary, but I think would be good practice.

The AC output from the balance charger would be a high frequency square wave with a P-P voltage of one cell plus the drop of two diodes.

Capacitor coupled cell balancer 3.jpg (29.67 KiB) Viewed 475 times

[rhitee05]

I like the concept. It's elegant.

A suggestion - you could make a multi-secondary transformer and using that to achieve your isolation instead of capacitors. Your new transformer would replace the switcher's transformer, with one primary and N secondaries (or N+1 if you wanted a loop-back for regulation). Each secondary would drive the rectifier on a cell directly, no capacitors needed. Besides saving some parts, this would make the switcher frequency less important. This sort of topology is how a lot of isolated multi-output power supplies work.

Another suggestion - save yourself some diodes and use a half-bridge or even just a single diode rectifier on each cell. You'll lose some efficiency, but for large packs the part savings could be significant. I gather that this would be powered from the wall, so a little loss of efficiency isn't that important.