Charge Pump style BMS?

[mwkeefer]

Hello all,

I've been thinking about this for a while - to be honest the concept comes from an old Issac Asmiov book who's title escapes me but...

It seems to me that the best way we could possibly balance (efficiently) would be to use some form Charge Pump.

I would assume it would need to have 2 seperate busses (cell VCC and cell GND busses) which were isolated with switching transistors from the bus (NO) - this circuit would then have a 2 part charge pump system.

Part #1 - MCU would determine highest voltage cell and set the charge pump system into STORE mode then connect the high voltage cell to the cap via the isolated bus until the cap was charged. Then the high voltage cell would be disconnected / isolated again

Part #2 - The MCU would determine the lowest voltage cell and would connect to the charge pump in mode 2 - charge... this would deliver an amount of current into the lower voltage cell and raise it's SOC.

This process would be repeated until the cells equalized...

The fundamentals are fairly simple - I think I could design this type of system (for up to XS packs easily) without the need of anything more than a minor mod to allow the Charge pump to be disconnected / isolated entirely so that I could use the same isolation transistors on the bus to enumerate through the various cell voltages using a single A2D channel of an MCU... this means no additional MUX is required and also that because I must use transistors to source the amounts of current which will be bursting back and forth the package size will still be about the size of a normal BMS.

The questions I have for the experts here -

1.) Am I correct in assuming this type of balance would produce the least amount of heat and waste the least amount of energy while using excess energy from adjacent cells to increase power in lower cells?
2.) What level of isolation (beyond the 2X busses with cells isolated) would you suggest.
3.) Does anyone see an issue with skipping the multiplexors for A2D input from each channel and rather using just 2 x A2D channels and getting the individual cells voltages in sequence one at a time with the MCU... I figure in this way it enables scaling to X number of cells, limited only by MCU speed and coding efficiency.
4.) If this is not a new idea - does anyone know if its patented, or in production.
5.) With proper isolation, or at a minimum an isolation resistor - couldn't this system actuall run while the pack was in discharge to maintain better cell balances?

Again as I always say - I am not this type of engineer, just have been thinking over this issue for months and figured I'd post what has come to me for peer review and hopefully constructive feedback.

One other item... if my CC max limiter circuit functions, I suppose this could also control the input current for charging with bulk supplies - allowing the SP-320 or the SP-350 to be used in various configurations of parallel and series but without needing to alter the current limiting inherint in the supplies themselves.

Thanks in advance for your time and consideration!

-Mike

 

[BatteryMooch]

Hi Mike,
Great idea...but it's already in production.
You're right about a lot of its advantages though, like high efficiency, very little heat, etc. It's expensive, has a high parts count, and can't easily handle high balancing current levels like a passive resistor-bleed balancer does though. But, that definitely hasn't stopped me from using it for more than one project.

Check out Texas Instrument's bq78PL114 and bq76PL102 chip set. It uses active-charge balancing via inductors and works beautifully, balancing during charge, rest, and discharge. There's a terrific app note on the TI web site about their "PowerPump" balancing that explains how it all works. You can check out the schematic for their evaluation kit to get a parts list for a complete BMS.

There's another thread in this group somewhere about the chip set and the experiences of a couple of people using it too. And there's also a bit info on the Web regarding custom active-charge balancing sytems for EVs being researched at a couple of universities.

 

[mwkeefer]

Thanks John - it sounds similar to my idea, my only immediate concern would be efficiency (given inductive transfer) though that solves isolation - if it works, why re-invent the wheel...

Thanks again for the insight = )_

-Mike

 

[BatteryMooch]

The inductor doesn't provide isolation, it's the charge-storage component.
And the efficiency is as high as any other method out there, e.g., capacitor-based charge transfer systems. Some of the details still make my head spin but luckily I can just concentrate on the BMS design and let the chip handle the ugly 200KHz charge transfer cycles.

 

[mwkeefer]

The inductor doesn't provide isolation, it's the charge-storage component.
And the efficiency is as high as any other method out there, e.g., capacitor-based charge transfer systems. Some of the details still make my head spin but luckily I can just concentrate on the BMS design and let the chip handle the ugly 200KHz charge transfer cycles.

I think I follow... I would have assumed direct connection via isolating transformer at 1:1 between adjacent cells would be the way to go so losses would be incurred though I have no idea how to calculate them... I suppose losses really don't matter much when the numbers should be so low either way.

I'm going to check out the chip (try for samples first, then just go ahead and bite the bullet if need be)...

You reference 200kHZ charge transfer cycles - I assume you understand this so please, correct me if I am mistaken but wouldn't that mean 200 charge transfers (store, pump) per second... seems it would move an awful small amount of current - the other interpretation I can take is that 200khz really performs 1 full pump cycle (1 for drain, 1 for charge)...

Do you have any understanding with this regard... also do you have some idea of the total amount of transfer abililty in, lets say a 60 second period?

Again, I need to go read the app notes and such... just trying to get as much insight as possible.

Thanks in advance

-Mike

 

[Toorbough ULL-Zeveigh]

isn't charge pump a synonym for switched cap?
anyways ignore this if what u have in mind is something different & have already seen this bms from ebikes.ca

Switched Capacitor Balancing BMS Circuits are In

Well it has been quite a long wait but we have finally received the proper cell balancing battery management circuits for the Iron Phosphate battery packs. For any customers who purchased a 24V, 36V, or 48V LiFePO4 battery from us, please send us an email and we will ship you a new BMS circuit that will bring all the cells back into balance, restoring the full capacity from the pack. These BMS circuits are also less prone to tripping from the capacitor inrush current when plugged into the larger motor controllers.

 

[bigmoose]

This will answer a lot of the questions as mentioned previously:
Application Report
SLUA524A–July 2009–Revised October 2009
PowerPumpâ„¢ Balancing
http://focus.ti.com/lit/an/slua524a/slua524a.pdf

 

[fechter]

the PowerPump thing is interesting, but in reading the app note I didn't see what kind of average current the thing actually runs at. Just guessing from the graphs without doing the calculus, it appears to be somewhere under 500ma average. They indicated a less than 50% duty cycle, discontinuous mode with a peak of around 1.5A. I designed a similar (but much cruder) circuit a long time ago, that uses the same inductor switching topology.

The inductive charge pumping scheme will work much better than a switched capacitor at high currents.
Capacitive charge pumps are great at low currents, but the ESR of the caps + switching FETs kills the performance at higher currents.

The weakness with both of these schemes is the if the high cell is many cells away from the low cell, it has a hard time pumping across a lot of cells. The real advantage is the minimal heat dissipation. The abiltiy to shuttle current during discharge is an advantage too, but at high (typical EV) discharge rates it would be a tiny percentage of the total capacity.

 

[GGoodrum]

Yep, there's a couple different threads on this idea. You could probably make this work for smaller balance currents, like the one Justin was selling before, but I think one of the issues his unit had is excessive drain on the pack. It never stops transferring between cells.

I still think the best idea to eliminate the waste, due to heat, is the idea Richard had which is to use flyback transformers to feed the bleed/shunt current back into the whole pack. Once version 4 of our current BMS is done, and we finally have some time to do something other the endless stream of tests we've been doing, maybe Richard will have time to get back to looking at this promising idea.

-- Gary

 

[mwkeefer]

Gary,

You make good points as do all...

I think the "chip based" charge pump (or whatever they call it) is limited severely in current and duty cycle...

You are correct there are some inefficiencies in dealing with FET losses but I think that for quick bursts of ON/OFF style charge - discharge this would be the fastest most efficient manor currently on the boards - don't get me wrong, I don't know the losses of the flyback transformer but if you can make v4 recycle the energy which is now basically bled off... I'm in for a unit or two.

The thing I like about the "switched capacitor" idea is it's relative simplicity, a single MCU and in essence a SPDT center off transistor to connect each side of the bus...

I still can't find a way around how to make this work while charging at the same time... Unless, it were to function:

Sub Charge()

// Should the process begin with a cell level balance if using a pump style to low end balance the pack prior to bulk charge?

TurnOnChargingRelay();

While (ENDOFCHARGE = FALSE)

 If HighestCellVoltage = > 4.0v Then

   If (OutOfBalance > AllowableOutOfBalance) Then

      TurnOffChargingRelay(); // Actually disconnect (relay or other) charger input to pack
      TurnOnChargePump();  // Only enables the circuit then enters a wait loop

      While (OutOfBalance > AllowableOutOfBalance)
          Pause(5000);  // Wait 5 seconds at a time
      End While

      TurnOffChargePump();
      TurnOnChargingRelay();

   Else

      // Here we just wait for the end of charge, this needs alot more to ensure proper voltage at end of charge but it illustrates the concept
      If HighestCellVoltage = CellHighVoltageCutout Then 
         ENDOFCHARGE = TRUE
         TurnOffChargingRelay();
         Exit While
      End If

 End If

End While

End Sub

That's just an idea.. the capacitor style charge pump is fairly straightforward to control from any MCU and the code should be evident.

-Mike

 

[fechter]

we did some capacitor charge pump testing, but the results were not great. To make it work well, you need really low Rds FETs and low ESR capacitors. The rest of it is fairly simple other than driving the gates.

My attraction to inductive or flyback based systems is not so much the efficiency, but just overall heat generation during balancing. I suppose it's sort of the same thing. If I could shunt 1 amp off a high cell without generating 4W of heat, it would be great. Even if it still dissipated 1W that would be a huge improvement. Making a flyback circuit that's 75% efficient is quite possible. Again, the switching transitors and how to drive them are the biggest challenge. Logic level FETs are still kind of borderline at 3.6v. Low saturation bipolar transistors work out about the same.

 

[BatteryMooch]

Yea, overall, the cap-switching methods are best (and typically used for) very low current balancing. On the order of 150mA with the ones I've seen. The inductor method can go to higher balancing current levels easier, but they're both expensive compared to passive resistor bleeding balancing.

The Texas Instruments charge-transfer chip PowerPump circuit in the reference schematic has an average balancing current of about 100mA.

 

[liveforphysics]

Keep in mind though, 100mA on a transfer-type BMS puts a pack in balance at roughly the same rate of at least a 200mA bleed-type BMS.

One just robs from the highest cell. One robs from the highest cell, and feeds the lowest cell with what it stole.

 

[Tiberius]

Keep in mind though, 100mA on a transfer-type BMS puts a pack in balance at roughly the same rate of at least a 200mA bleed-type BMS.

One just robs from the highest cell. One robs from the highest cell, and feeds the lowest cell with what it stole.

Hmm, doesn't that only work if the charger is also putting 100 mA into the whole pack? In that case, its the same either way: the full cell is getting a net charge of zero and the lowest cell is getting 200 mA. The difference is that the charger has to supply 100 mA instead of 200.

But a charge transfer balancer would still score in speed, though, simply because you could start it earlier. It doesn't have to wait until the first cell is full.

Nick

 

[Jeremy Harris]

I've found that just using individual switched mode DC-DC converters for each cell seems to be pretty much the best way of balancing a pack quickly. It's dead easy to do, needs nothing fancy in terms of circuitry and just works.

Get a few of these at $2.99 each: http://cgi.ebay.co.uk/ACON-DH50S24033-DC-DC-Power-Converter-24V-3-3V-16A_W0QQitemZ130366323504QQcmdZViewItemQQptZLH_DefaultDomain_0?hash=item1e5a703b30, set the trim up to 3.63V (pretty much the maximum and good enough for a full charge on LiFePO4), hook them up to a suitable 24V supply (a few Meanwell 350W supplies will do) and you have an up to 16A per cell balanced charger.

I've just ordered some more of these converters to build another charger, as the one I have for the 80Ah boat battery works so well. They need fan cooling, but I just have a mains powered fan fitted to the end of the case that blows over them all the time. There's no need to worry about current limiting, as in practice the resistance of the individual cell charging leads is enough to make sure that the converters don't hit their current limit (to drop the ~0.4V between the cell resting voltage and the fully charged voltage only needs about 26 mOhms of lead resistance at the max rated current of the converters).

By default this system always keeps the pack perfectly in balance, as each cell is being charged to the same terminal voltage.

Jeremy

 

[peterperkins]

I quite like those Jeremy but how do you know if they are all working correctly without some sort of monitor on the individual cells.
I presume you use these with a BMS to monitor high and low voltages?

 

[Jeremy Harris]

The simple answer is that I don't really know they are working, but I do have a Cellog 8 that I can plug in to look at the individual cell voltages. I suppose I could leave this plugged in and wire up it's alarm, but I've not really felt the need to. These are high reliability parts, intended for running things like server power supplies, so should continue to work OK if kept cool. They have built in over voltage and over current protection, plus over temperature shut down, so are almost certainly more reliable than anything home made that I might build to monitor it could hope to be.

For discharge monitoring I have a conventional cell-level low voltage warning system, plus the Cellog 8 that I can plug in if I want a detailed look at things.

It would be quite easy to build a BMS around the Cellog 8, or perhaps a bank of them for high voltage packs. They are cheap enough to buy (see here: http://www.hobbycity.com/hobbyking/store/uh_viewItem.asp?idProduct=9282&Product_Name=Cell-Log_Cell_Voltage_Monitor_2-8S_Lipo) that you could have a few spares. The settable LVC/HVC alarm can drive external controls easily enough.

Jeremy

 

[liveforphysics]

I've found that just using individual switched mode DC-DC converters for each cell seems to be pretty much the best way of balancing a pack quickly. It's dead easy to do, needs nothing fancy in terms of circuitry and just works.

Get a few of these at $2.99 each: http://cgi.ebay.co.uk/ACON-DH50S24033-DC-DC-Power-Converter-24V-3-3V-16A_W0QQitemZ130366323504QQcmdZViewItemQQptZLH_DefaultDomain_0?hash=item1e5a703b30, set the trim up to 3.63V (pretty much the maximum and good enough for a full charge on LiFePO4), hook them up to a suitable 24V supply (a few Meanwell 350W supplies will do) and you have an up to 16A per cell balanced charger.
Jeremy

Wow! I like those much better than the bulky inefficient vicors!

Outstanding charging setup for LiFePO4! I would expect nothing less from you of course.