high power discrete switched capacitor balancer

C

curious

1 kW
Joined
Dec 29, 2007
Messages
445
Location
NJ, USA
I think it should be pretty straightforward to do using discrete mosfets. It can be made to run slow enough such that generic (cheap) optoisolators will suffice for the gate drive. Then you have N-1 capacitors for N cells switched back and forth (alternating offset of 0 and 1) by 2N mosfets. The control circuit should run two mosfet groups with a dead time interval to avoid quiescent current. Caps should probably be of MLCC type for longivety. Mosfets for this switching network are low voltage (cheap).

The benefit of such balancer is that it can be run continuously or periodically during discharge as well. If it has enough power then per cell LVC becomes unnecessary. Moreover the battery capacity can even be slightly extended - the pack capacity is no more limited by the weakest cell.
 
Build it! I will fund it. It sounds like it would work, and I like the idea of transfering energy to balance rather than just wasting it as heat.
 
Switched capacitor balancing is not 100% efficient (unfortunately) but certainly better in terms of efficiency than shunt balancing. Another alternative for gate drive is custom xformer with 1 primary and N secondaries (two xformers needed for a 4-phase drive for dead time interval). Custom job but should be even cheaper.
 
I would be more interested in the cap setup than a x-former setup. For a 20s setup, it would only need 19caps, 19fets, 1 uControler, and some misc driver chips, resistors, caps etc right? That shouldn't be expensive. Do you have the skills to make a prototype?
 
Or maybe 1 cap and 20fets, and 1 ucontroller...
 
I mentioned pair of xformers as alternative mean to drive mosfet gates (control path only). It is still a switched capacitor topology.
 
bahh... I'm not digging the transformer idea at all though. I'm sure it could work, but custom parts always seem to make projects into nightmares. Can you do it without that?
 
I think this would be a fine idea. There is already a good reference shunt regulator design on the board. This would be a good compliment to it. I'd just recommend using transistor packs and multi-channel parts where practical. High part counts make assembly a chore. I would be fine with a home made gate drive transformer IF it's based on a readily available core (i.e. available from Digikey, Mouser, or equivalent) and a lot of pictures/directions on how to wind it are provided.

Personally I'd still want to include a per-cell LVC as a backup and to provide a measure of redundancy.

Lawson
 
liveforphysics said:
I would be more interested in the cap setup than a x-former setup. For a 20s setup, it would only need 19caps, 19fets, 1 uControler, and some misc driver chips, resistors, caps etc right? That shouldn't be expensive. Do you have the skills to make a prototype?
Click to expand...

For 20s you would need something like 40 Fets, 19 caps, 1 control circuit and either 40 optoisolators (there are assemblies of multiple units per chip) or two DIY gate drive xformers. The idea for gate drive xformers is surely to use off-the shelf core, or perhaps core+primary. But secondaries need to be DIY. Perhaps a pack of N wires can be wound together in one pass to form N secondaries.

I plan to build one but perhaps a bit later. I've been sidetracked from my ebike project for almost a year (due to job and moving) and my first priority is to get the bike running with minimal custom effort (I am using a pair of RC chargers and LVC-only circuit). Then I'll get back to this idea.
 
How about the alternative of a switched single cell charger as a balancer?

Might be an interesting approach, and would certainly ensure that all cells were charged to exactly the same terminal voltage.

Jeremy
 
That will certainly work but there is a limit on balancing power, plus only one pair of mosfets will be utilized at a time with high peak to average current ratio (low utilization of silicone cost).
 
Check out the charge transfer battery controller from Texas Instruments, the bq78PL114 and bq76PL102. Using the PL114 would save an anyone a LOT of time designing a charge-transfer balancer. :D
 
The number of parts or cost of parts doesn't matter much. Parts are so cheap these days, it just doesn't matter at a hobby level for making a handful of something. What does matter to me, is making something compact that can balance effectively. I charge a 40Ah pack at 58amps. My balancers do a pathetic 700-300mA, which is not adquate to balance a 40Ah cell in a timely mannor. If you can make something that can do cell transfer balancing, and make it in a high current format, people will use it.
 
Actually, for charge-transfer balancing, you don't need a very high balance current rating at all. Since you can balance during charge, when idle, or during a discharge, it allows you to use much smaller components and still get the same balancing benefits of shunt balancing (resistor bleeding of higher voltage cells). Existing charge transfer chips max out at about 1A. And that's a lot for being able to balance continuously. :mrgreen:
 
The Ti chipset is interesting, but oddly it is limited to 12 cells. It also looks like you need to add a ton of parts to get the transfer current up over about 100mA.

About a year ago, there was a thread or two about doing a switched cap balancer. I looked into using the LM2663 chip, which allows a balance current of up to 200mA. Here's what the circuit for two cells looked like:

Switched%20Capacitor%20Balancer-v1.2.png


The op-amps were used to drive a bi-color LED that indicated which direction the current was flowing by the color.

I did a proto board for a 16-channel version. Here's what it looked like:

Switched%20Capacitor%20Balancer-v1.2-PCB.png


The big problem I had was trying to solder those tiny little surface mount parts. Unfortunately, there's not a DIP version. Anyway, after making a mess of a couple of these, i back-burnered the whole idea.

Around the same time, Justin started using a switched cap-based BMS, but ended up recalling them and went back to a more traditional shunt-based design. I think the reason was that somehow it would drain the cells too far down. Can't remember exactly.

-- Gary
 
You can easily expand the number of cells that are handled by using multiple blocks of 12 cells and digital isolators to the bq78PL114's comm lines. This is the method recommended by TI Tech Support (spoke to them last week). Similar to using the Linear Tech LTC6802-2 chip.

The Cell Balancing section of the datasheets for the two chips mentions that they're spec'd for up to 1A in balancing current. But, even 100mA is a LOT of current when you can balance 24 hours a day. The cells just don't have a chance to become unbalanced and don't need the high current levels that other balancing methods might require.

I don't know how switched-cap circuits compare to switched-inductor circuits but I was impressed by just how few components were needed for the high-current balancing circuit in the bq76PL102 datasheet. Maybe I was just expecting it to be so much worse. :mrgreen:
 
The TI chips are also surface mount, and a lot more complicated. I think you still need a bunch more parts, even to do basic 100mA current transfers. The LM2663 is completely self-contained, except for the small 10uF cap, and you can use any number of channels, without any glue chips.

-- Gary
 
So, are you saying with a pair of bq78PL114, and some other parts, you could design a cap balancing setup with no transformers that will work well? If you can get a circuit that works, I will pay to have some boards CNC'd, and order the parts and build them. I have a hot air work station at work, and I can work with surface mount IC's Though it would be nice if the rest of the parts were through-hole just to ease assembly.
 
I've glanced at 114 datasheet. Nice part but I am not convinced that we need all its functionality. Plus I think for DIY project the package would be too cumbersome to work with. I am all for SMD, in fact I even prefer larger discrete SMD packaging to through hole parts for prototyping work. But anything non-socketable in fine pitch packages, pins underneath etc. - forget it unless you have a binocular and masochistic character.

BTW 114 reference schematic shows even simpler gate drive - capacitor isolated. So this is yet another option. Basically needs one zener/TVS for gate protection, and a series RC network, three parts total for a gate drive. One thing to be careful is about pre-charge transient with this gate drive arrangement. It can potentially cause quiescent current on start-up/shut-down/pack load transients. Needs some thought.

As far as C vs L switching balancers - there are some tradeoffs in either one. I think C is easier for DIY. No need for fast recovery diodes etc. When C balancer reaches equilibrium there are no recirculating currents, the only energy sink is gate drive. C balancer power can be adjusted simply by changing drive frequency.
 
liveforphysics said:
So, are you saying with a pair of bq78PL114, and some other parts, you could design a cap balancing setup with no transformers that will work well? If you can get a circuit that works, I will pay to have some boards CNC'd, and order the parts and build them. I have a hot air work station at work, and I can work with surface mount IC's Though it would be nice if the rest of the parts were through-hole just to ease assembly.
Click to expand...
The bq78PL114 is an inductor-based charge-transfer balancing chip. It handles up to 4 cells on its own and uses the bq76PL102 for each additional 2 cells, up to 12. Repeat in 12-cell blocks as necessary.

But, I'm thinking why use the bq78PL114 at all if I already have a microprocessor that can handle comms with just the bq76PL102 chips? They handle the balancing functionality and that's all I need. The micro will handle the protection functions. I'll be playing with both chips in about a month when the eval kits are back in stock.
 
GGoodrum said:
The TI chips are also surface mount, and a lot more complicated. I think you still need a bunch more parts, even to do basic 100mA current transfers. The LM2663 is completely self-contained, except for the small 10uF cap, and you can use any number of channels, without any glue chips.

-- Gary
Click to expand...
I've looked at the schematic and am still confused as to how the LM2663 handles bidirectional charge transfers. Is there a overview of the circuit posted anywhere?
 
CamLight said:
I've looked at the schematic and am still confused as to how the LM2663 handles bidirectional charge transfers. Is there a overview of the circuit posted anywhere?
Click to expand...
Basically similar to this:
http://www.maxim-ic.com/appnotes.cfm/appnote_number/725
 
One more neat idea. Since switched capacitor circuit is not sensitive to gate drive shape a sinewave can be used from LC tank for gate drive allowing recovery of gate charge and very low power operation. Moreover sinewave gate drive avoids the need for dead time in switchover and second xformer. Only a single LC tank with 2N secondaries is needed for gate control. And the entire control circuit in this case is a single transistor ! ;)

Sounds too simple but it may actually work. Hmm... maybe I should order some parts ;)
 
LC tank secondaries can be wound with a single flat ribbon cable. Makes pairing easy.
 
I have very limited electronics experience, but read the http://focus.ti.com/lit/ds/symlink/bq78pl114.pdf anyway. Looking at the shematic of chip plus the reasons for its design criteria, it looks like a very sensible approach. I see sophisticated capabilities - not all need be implimented. I see scalability. I see heat efficiency. I see the potential for resin potting and a robust solution. With very limited heat dissipation required. That means lots of packaging possibilities exist. and its small.

In fact the more I look at the overview the better I like this approach.

Repeat in 12-cell blocks as necessary.
Click to expand...
I would be concerned a little in very large packs you would have 12 cell blocks become "unbalanced" compared to other 12 cell blocks. I doubt in practice this would be an issue with the odd overnight on the charger.

I'll be playing with both chips in about a month when the eval kits are back in stock.
Click to expand...
I will follow that with great interest. I think it has huge potential.

With regard to a kit assemblly I found http://www.youtube.com/watch?v=vZ1qisX52rI&feature=rec-HM-r2 and Norman Meir. (Well worth watching the rest of his collection of video.) A specialized skill indeed, but not impossible. Need a stereo scope though. Perhaps a kit could include a chip done professionally, The rest descrete components?
 
You must log in or register to reply here.

Similar threads

methods
    • Like
How to turn a high voltage DC-DC converter ON/OFF
Replies
17
Views
5,207
methods
methods
Back
Top