The care and feeding of a123-based packs...

[Doctorbass]

I would like to thanks you GGoodrum for the great job you did about the LVC pcb that you have sent to me . Great packing, very clean pcb, that's Cool!

Here are some pics i toke after assembling those:

I will use 2 board for my existing 20s 4p setup and 2 next boards for my future konion pack 20s 16p of 2000Wh

Doc

[lawsonuw]

ok, the standby current of the given circuit is about 0.5mA plus any leakages in the silicon parts. The circuit should still work just fine if the values of R2 and R3 are increased 10x to 100x. This mod would get the circuit into the ~10uA leakage range, but part by part calibration would likely be needed. (someone should still test the final circuit for the leakage current though)


@GGoodrum: Yes this circuit impliments the CV part of the charging profile. The main issue is that any current the charger is supplying that the batterys aren't using is dissapated as heat in the large power darlington transistor, TIP142 in the schematic. So if you're pumping 10A into the pack and all the circuits have a lit LED, each circuit has to get rid of up to 36 watts of heat. So... the charger really should look at the feedback line from this protection circuit and cut the current back at the end of a charge. Having the circuits dump 1-2 watts max at the end of charge shouldn't be too hard to manage. The way Jeff's cutoff modules use the power transistor's tab as the posative terminal is a good way to deal with this waste heat. With a good themal connection, the cell paralleling bus bar should make a good heat sink for the protection circuit.

VarrrooooOOOOMMMMM!
Marty

[GGoodrum]

Okay, I'm starting to understand better now. The problem is that if these are buried in packs, what good does an LED do if you can't see it? For that matter, having an optocoupled output for all these doesn't do any good either, because there is no off-the-shelf charger that I'm aware of that can make use of this output.

Instead, what if we just used one of these clamping circuits to simply limit the voltage that each cell can reach to 3.7V. Then we can use any one of many existing SLA chargers, that also have the CC/CV charging profile, but at a pack voltage level. The way I think it would work is that as soon as a cell hits 3.7V, it stays there. Once all the cells hit that point, the charger will see that the pack cutoff voltage has been hit, and will start reducing the current. For me, this would be the best of all worlds, because there are lots of existing SLA chargers that would work great for this. I really don't need to know which cell block hit 3.7V first.

So, if this would work, which parts of the circuit above could be eliminated? Would a max of 36W per channel still need to be dealt with. and if so, could the PCB be used for this?

-- Gary

[Malcolm]

what if we just used one of these clamping circuits to simply limit the voltage that each cell can reach to 3.7V. Then we can use any one of many existing SLA chargers, that also have the CC/CV charging profile, but at a pack voltage level.

Couldn't have put it better. There are so many decent SLA chargers available that can do most of the job (and do it cheaply) that it seems overkill to start messing about with lab power sources and replicating stuff that's readily available. But how do you shut off the charging circuit if one cell takes too long to reach the cutoff voltage?

[GGoodrum]

Couldn't have put it better. There are so many decent SLA chargers available that can do most of the job (and do it cheaply) that it seems overkill to start messing about with lab power sources and replicating stuff that's readily available. But how do you shut off the charging circuit if one cell takes too long to reach the cutoff voltage?

Well, in that case the SLA charger won't get to the CV mode and most CC/CV chargers at least have an LED or light that shows what mode it is in. Many have timeout features as well. In any case, if you've got a cell that is not hitting the cutoff, chances are it has been highly stressed, and has lost capacity. It will likely trip the LVC much earlier than it should, which will also give you an indication that something is amiss.

I'm still on the fence about whether or not it makes sense to do this internally (i.e. -- buried in the pack...), or do it as an external board, and still bring out balancer plugs out of the pack. In the latter case, however, it is also possible to simply do individual cell charging, like I'm doing now. Doing it internally does have the advantage of making packs made with a123, and/or other LiFe cells, act just like SLAs, at least from a charging point-of-view.

-- Gary

[brandonh]

Some RCers already use CC/CV supplies to charge their A123s, and there's even a YouTube video showing how it's done with a Mastech 5020E supply.

I think there's an issue with your overvoltage-threshold-in-series idea. I'll assume that a CC/CV charge is being used, with voltage set to (#batteries*3.6v). In an out-of-balance pack, one cell will need longer to reach full voltage. That lagging cell might be at 3.4v, while the other cells are at the 3.6v target. Since the charger starts in CC mode (not at Vlimit yet), the other cells rise in voltage until the total of their voltages plus the lagging cell voltage equals the charger limit (#batteries*3.6v). At Vlimit, the charger switches to CC mode, with current through the string dropping quickly - many cells won't accept as much current. The cycle stops when all cells have reached at least the 3.6v threshold. But while the lagging cell voltage is rising, the other cells are all being overcharged! The cycle does nothing to balance the pack.

This is the whole reason for the shunt regulator "voltage clamp" design; it enables each cell to rise in voltage to exactly 3.6v, then accept no more current. The pack comes out of every charge cycle balanced.

If your pack stays in reasonable balance, then the above setup would work, but it seems to get you nothing beyond a basic CC/CV charger, other than knowledge of whether each cell has reached its threshold. A123's are especially tolerant of a little overvoltage, so I think it could work just fine, as long as you have some way to shut off the CC/CV charger after a timer or when the current dips low enough through the pack.

My desired (if I can find the $$$) charge solution for a 10s4p A123 pack is to use:
- 1 Mastech 5020E 50V/20A charger set at 36v and 10A
- 10 combination LVC and shunt regulator circuit boards
- 1 charge cutoff FET or relay between the Mastech and the pack, triggered when all the overvoltage signals are high, or after a specified time delay has passed (safety backup)

The problem, like GGoodrum mentioned, is heat dissipation through the shunt regulator at higher charge currents. If you use Jeff's shunt reg board, there's very little space to dissipate the heat, and you'd probably at minimum add a heatsink. Maybe the "balance" phase, where some regs are dissipating while others aren't, is short enough that the heat generated can be easily dissipated. Maybe you find some clever way to reduce the current through the pack when the first cell reaches the 3.6v cutoff, but this would likely increase charge times. Does anyone have an intuitive understanding of how much heat we could push through the TIP?

One solution is to remove the shunt regulator boards from the pack, and keep them external; but then, for every battery in the pack, you need a thick wire exiting. For my 10s pack I'd need 11 powerpoles to do 20A charging, but I kinda like this idea, because you can change the charging circuit without changing the pack, and it's much easier to dissipate heat externally rather than internally.

Is a shunt regulator the best design, or could we design a PWM-based dissipator or charger?

Is the above correct? Either way, let's keep up this discussion.

[GGoodrum]

For those that still want the LVC modules, I have made them available now on my RC website. You can find them here: http://www.tppacks.com/products.asp?cat=26. Two 10-cell versions are provided, one that includes the board, all the parts and a detailed set of instructions, and a "deluxe" version that adds two RC-type balancer plugs and a small JST plug for the output. The deluxe kit also includes detailed instructions.

Here's what is included in each kit:





The deluxe/complete kit also includes an option for matching male balancer and JST plugs that can be used to wire into existing packs and the controller.

I'm working on finishing up the instructions for the a123 10s4p pack construction kit, and will make these available soon, as well. There will be a version that includes the LVC circuits on the same board that is used to make the balancer connections (as shown earlier in this thread...).

I'm also working on doing a number of kits for the LiFeBatt cells that will at a minimum include the LVC function. If this "clamp" idea pans out, I will probably include this with LiFeBatt boards, and also update the a123-based kits as well.

-- Gary

[Malcolm]

Great work Gary! I'm definitely interested in a version for the LifeBatt cells (12s please!).

Regarding charging: As the extra time/expense needed to add balancer plugs is relatively small compared to the cost of these packs I definitely think it makes sense to add them right from the start. It leaves options open for the future.

The clamp idea sounds good, but what's wrong with an existing solution such as the Astroflight Blinky: http://www.astroflight.com/store/store- ... 74RtU2x0b7
Would this not be able to dump enough current to cope with balancing a pack of more than 1p (2.3 Ah)?

[GGoodrum]

Some RCers already use CC/CV supplies to charge their A123s, and there's even a YouTube video showing how it's done with a Mastech 5020E supply.

I think there's an issue with your overvoltage-threshold-in-series idea. I'll assume that a CC/CV charge is being used, with voltage set to (#batteries*3.6v). In an out-of-balance pack, one cell will need longer to reach full voltage. That lagging cell might be at 3.4v, while the other cells are at the 3.6v target. Since the charger starts in CC mode (not at Vlimit yet), the other cells rise in voltage until the total of their voltages plus the lagging cell voltage equals the charger limit (#batteries*3.6v). At Vlimit, the charger switches to CC mode, with current through the string dropping quickly - many cells won't accept as much current. The cycle stops when all cells have reached at least the 3.6v threshold. But while the lagging cell voltage is rising, the other cells are all being overcharged! The cycle does nothing to balance the pack.

This is the whole reason for the shunt regulator "voltage clamp" design; it enables each cell to rise in voltage to exactly 3.6v, then accept no more current. The pack comes out of every charge cycle balanced.

I probably have as much experience with RC chargers, as anyone on RCGroups, especially as they apply to use with a123 cells, and I can tell you I think you are not right to think that all you have to do is charge a cell until its voltage reaches 3.6V, and then cut it off. The "resting" voltage will end up only being around 3.4V, and it will only be charged to about 80%. What you described is simply a CC charging profile. The whole point of adding the CV mode is so that the cell can be held at that voltage for awhile, while the current gradually reduces down. This is what "tops" off the cell to get it close to 100% charge. BTW, the best voltage to use for this cutoff is 3.65V, not 3.6V. It doesn't really hurt the a123 cells to go over that, to say 3.7 or even 3.8V, but going under won't get you to 100%.

About a year and a half ago, when the a123s first hit the RC world, we did quite a bit of testing with various chargers, and charging profiles. We started simply using regular LiPo modes, which uses a CC/CV cutoff of 4.2V per cell. We monitored the voltage with a PC, as the cells were charged. What happens is that voltage rises steadily at a fairly slow rate until about 3.7V is hit. Then, the rate changes drastically, and it will rise quickly to the 4.2V cutoff, where it stays until the CV portion is finished. We asked a123Systems about this and they said they have done similar tests. They said it doesn't really hurt the cells, but if you did this all the time, it might reduce the cell life. I still have a pack with a number of cells that went through these early torture tests, and it is still going strong with close to 800 cycles on it now and one of my helicopters.

Anyway, the clamp circuit needs to simply limit the max voltage a cell can get to during a charge to 3.7V, and then a normal CC/CV charger can be used, and then you will end up with a balanced pack.

-- Gary

[brandonh]

Thanks for the explanation Gary! That really helps.

I see now, there are three basic ways to charge with balancing:
-charge entire pack in (mostly) CC mode, cut off cells at threshold by dissipating current as heat, as described above
-charge each cell individually with CC/CV chargers
-alternate charging to a simple voltage threshold with balancing (but has issue if one cell is really out of balance and charge never stops)

I guess the next questions are how to dissipate the (3.65v * pack current * # cells) watts, and whether and how you can dissipate that much in or out of a pack. With a perfectly balanced pack, you'd dissipate no heat at all.

How out-of-balance has anyone seen an A123 cell go? This could get us a bound on the time where we'd need to dissipate the heat, and then we'd have some data for the internal/external argument.

Thanks,
Brandon

[GGoodrum]

Great work Gary! I'm definitely interested in a version for the LifeBatt cells (12s please!).

Regarding charging: As the extra time/expense needed to add balancer plugs is relatively small compared to the cost of these packs I definitely think it makes sense to add them right from the start. It leaves options open for the future.

The clamp idea sounds good, but what's wrong with an existing solution such as the Astroflight Blinky: http://www.astroflight.com/store/store- ... 74RtU2x0b7
Would this not be able to dump enough current to cope with balancing a pack of more than 1p (2.3 Ah)?

Yes, I plan on doing LiFeBatt versions for 6 cells, 8 cells, 10 cells and 12 cell. Maybe 16 cells as well, but it depends on the configuration. I need to see if 16 cells will be easy enough to mount on a typical rack.

The "Blinky", and the Thunder Power 10-Cell TP-210V autobalancer work exactly the same way. They simply look for the cell with the lowest voltage, and try to bring the others down to this level. The problem is the high cells are only discharged at about 150mA, so that's not really going to work here.

-- Gary

[Malcolm]

The problem is the high cells are only discharged at about 150mA, so that's not really going to work here

Suspected that might be it. Thanks for taking the time to explain.

[rf]

There's much going on in the marketplace addressing more efficient equipment everywhere. Shave off a watt or two here and there and soon it adds up. Efficiency is becoming more important and it makes sense.

This is why the shunt balancer bothers me. You put energy in and then you bleed it off as waste heat. I think there are better ways.

I just ordered 12 of those single cell chargers that Gary is using. I think charging cells individually may be more efficient than charging enmass and bleeding off to balance. We need to find the clever ways to do it that require fewer parts and simpler designs.

Richard

[Malcolm]

Getting a little obsessive now. Wide awake at four o'clock this morning I came up with another daft idea for balancing cells. But in the clear light of day it doesn't seem so daft.

Am I right in thinking that if I could find a fairly simple and practical way of switching cells in a pack from series to parallel that A123 cells and LifeBatt cells should balance themselves perfectly well without any help?

The idea is to run wires from each cell to a multi-pin socket (possibly two). The cells themselves would not normally be connected to each other. Instead, you have two plugs that fit each socket; one is wired with jumpers to give a parallel (balancing) configuration, and the other wired to give a series (run) configuration. If you fit the parallel plug whenever the pack is not in use you can keep the pack perfectly balanced until you need it.

You also have the option of charging the entire pack as a single parallel string at 3.7V, although I'm not sure how useful that would be.

Granted it's not a very practical idea for A123 cells, but for LifeBatt cells with their screwed terminals it seems workable. For a 12s pack I could use two 12-pin trailer plugs and sockets. As long as I use 10 gauge cable, solder the connections to the sockets and keep the runs as short as possible it shouldn't add too much resistance.

[TylerDurden]

Malcom,

It is a fine idea, IMO. As others have discussed, 10 gauge can be stiff... using double connectors and smaller gauge wire may be easier and provide lower resistance.

[fechter]

The connector thing would work fine in theory. I can't imaging how hard it would be to pull apart a 24 pin Anderson block though. Finding the right kind of connector would be the trick.

[Jozzer]

I had been considering this idea Gary, but had given up, mainly due to the fact that a 10AH 33v pack would become a 100AH 3.3v pack...most chargers we have would take a month to charge it up Also, I am of the belief that the fewer connectors needed the better.

I am sure that for my application the best bet is a BMS that sits with the pack and handles all balancing and low/high voltage cutoff. Other ebike vendors here in the UK are after tha same thing, most are only interested in selling a boxed solution that "seems" simple to the end user.

[fechter]

If you use the shunt regulators on each cell and one full voltage charger, the shunt resistors won't dissipate any power unless a cell is out of balance. Normally, I think the shunts wouldn't be doing much since only a slight difference in current should keep the cells in balance.

[rf]

...

Am I right in thinking that if I could find a fairly simple and practical way of switching cells in a pack from series to parallel that A123 cells and LifeBatt cells should balance themselves perfectly well without any help?

The idea is to run wires from each cell to a multi-pin socket (possibly two). The cells themselves would not normally be connected to each other. Instead, you have two plugs that fit each socket; one is wired with jumpers to give a parallel (balancing) configuration, and the other wired to give a series (run) configuration. If you fit the parallel plug whenever the pack is not in use you can keep the pack perfectly balanced until you need it.

You also have the option of charging the entire pack as a single parallel string at 3.7V, although I'm not sure how useful that would be.

Granted it's not a very practical idea for A123 cells, but for LifeBatt cells with their screwed terminals it seems workable. For a 12s pack I could use two 12-pin trailer plugs and sockets. As long as I use 10 gauge cable, solder the connections to the sockets and keep the runs as short as possible it shouldn't add too much resistance.

Similar ideas have been bouncing around in my head too. From connectors and switches to a completely flexible array of cells connected by solid-state relays. Rearrange cells in any configuration on the fly ... even take exhausted cells out of circuit.

If only those extra parts didn't cost money and consume energy ...

[rf]

If you use the shunt regulators on each cell and one full voltage charger, the shunt resistors won't dissipate any power unless a cell is out of balance. Normally, I think the shunts wouldn't be doing much since only a slight difference in current should keep the cells in balance.

In my limited experience with A123/Dewalt packs a common out of balance scenario seems to be one cell lower than the rest. So balancing via resistor shunts is going to bleed all cells in the pack but one.

Putting charge into those cells, then bleeding it off contributes to their eventual demise like any other discharge. (Whether the impact is significant, I don't know. Not enough coffee yet this morning.)

One charger per cell or circuitry to modulate charge to each cell would be more efficient and contribute to longer cell life.

Just thinking out loud.

[GGoodrum]

Like Jozzer, I'm looking for a simple solution that only requires plugging in one charge connector. My current 10-charger setup has two connectors and only does one 10s4p pack at a time, so I'm not quite there yet. It also charges at a 2A rate, so it takes awhile. I know from personal experience that a123 cells can be safely charged at rates up to 25A per cell, which is about 10C, and I've heard from a few of my RC friends that they have tried 30A, with no problems. Practically speaking, limiting it to 4C or about 10A per cell, ensures a long cell life and is still reasonably quick.

The problem is that I don't know of any CC/CV single-cell chargers with a max rating of 40A and with the proper 3.7V cutoff, so going the multiple single charger route is not very practical. There's also the problem of haw to connect these, etc.

I also agree with Richard that the shunt regulators won't be doing much for cells that are reasonably well balanced in the first place. These would work different than the the typical RC balancers, which all try and constantly bring down the level of the highest voltage cells to that of the lowest. They are working all the time. In this application, they would do nothing until the cell reaches the 3.7V limit. In the worst case, for a 10-cell pack, you might have 9 of them working to bleed off the max charge current while the "weak sister" catches up. Once it does, the CV mode in the charger kicks in, and it does the job of limiting the voltage. Then, the cell shunt regulators would be off. You would just need to make sure the cutoff for the shunt regulators is just higher than 1/10th of the charger's CC/CV cutoff voltage, so that it is doing the "heavy lifting".

I'm not sure how hard it would be to find a suitable 40A CC/CV charger with the right cutoff voltage (37V for a single 10s4p pack, or 74V to do the whole 204p setup...), but what worries me more is what it would take in terms of a heat sink, etc., for the individual shunt regulators to be able handle this much current. Maybe scaling things back to about 1C, or 10A would be more practical? In the schematic previously shown, how much will the parts listed handle, and what sort of heat sink would be required? My hope would be that a big chunk of metal would not be required.

-- Gary

[Jozzer]

I think 10 amps is more than enough for most Ebikes, and more to the point, it is more than most of the chargers that are suitable for our use (old SLA/NiMh chargers) put out.

For motorcycle type applications, where more charging current would definitly be desireable, the board could be fitted with a heatsink, and mounted in open air or fanned, or a new board made with components chosen for higher loading...

[rf]

I think 10 amps is more than enough for most Ebikes, and more to the point, it is more than most of the chargers that are suitable for our use (old SLA/NiMh chargers) put out.

For motorcycle type applications, where more charging current would definitly be desireable, the board could be fitted with a heatsink, and mounted in open air or fanned, or a new board made with components chosen for higher loading...

Charge current will likely be different for different folks. Being able to charge in 15 minutes is wanted by most everyone, but not necessarily needed. The cost of fast charging versus overnight, or somewhere inbetween will probably make the difference.

Even if you have fast-charge capability you may not be able to use it if you're away from home. It could blow fuses. Something you especially don't want to do if you're `borrowing' power.

Richard

[brandonh]

Before choosing a charge method, I want as many opinions as possible, so I emailed John Reid from BattleBots fame, who has a page with A123 info: (http://www.terrorhurtz.com/a123/). Conversation copied with permission:

---

I'm looking at building a charger for a 10s4p A123 pack, to replace the stock DeWalt charger that doesn't seem to balance well (sometimes charge stops at 3.3v, sometimes at 3.6v), for a full-custom Segway. My question is "How far out of balance have you seen an A123 pack get?"

---

Hi Brandon,

Well I suppose it is possible that if one cell develops a micro-short (not unheard of) and they are left for a while, then one cell may be flat while the others are more than half charged. I have never actually seen a pack get anything like that bad.

Note that the current clamp technique does not necessarily result in a balanced pack, especially at high charge currents. Because of the internal resistance of the cells, there is a voltage rise due to the charge current - around 0.2V at 20Amps. So the current will not just shut off as soon as they reach the clamp voltage – as the current drops, the voltage will drop too, so the current will only slowly tail off. So the ones that started clamping first, will still be the most charged at the end, unless you leave the them all on charge long enough for the last one to catch up. But then, at 20 Amps, I guess that won’t take too long.

The way the DeWalt does it is just to do it at a very low current, so that the resistive voltage rise is small.

But then, as you have current clamps built in, maybe it is not too important that they are perfectly balanced, because they can never over-charge. The only problem is on discharge, when one cell may go low, so it might be wise to be conservative on the cut-off voltage.

Good luck

John Reid

---

Thanks for the answer! Two more questions:

-Since I have four cells in parallel for each segment, only 5A is going through each cell at 20A pack charge current, so the voltage rise should be pretty small, right?
-Do you mind if I share your answer on then Endless Sphere ebike forum? I'm sure others would like to hear this.

Thanks
Brandon

---

Ah, OK, didn’t realise it was 5 Amps per cell.

Even at 5 amps, the voltage rise is large in relation to the voltages differences you look for in normal peak balancing, but in your setup where you protect from over charge and are only worried about the balance on discharge, the actual difference in charge is small.

Basically you only have to balance the cells well if you are charging them without any individual cell protection.

Yes, feel free to share my ramblings.

Cheers

John

[rf]

Basically you only have to balance the cells well if you are charging them without any individual cell protection.

Or if you hope to realize most of their discharge potential.