Even Newer 4 to 24-cell Battery Management System (BMS)

[biohazardman]

i was talking about the cyclone ones not the ones you linked to they looked great. I was asking if you are including the case with the purchase of the BMS?

In the past only the bare boards were offered and a BOM Bill of Materals/parts list. The rest is DIY. I can't wait to get one of these myself although I am not looking forward to all the soldering again. Good for many that Andy stepped up to the plate for that part.

 

[reagle]

I've dealt with some battery boards in the past where we had to provide a mechanism for discharging before disposal (those were primary packs).
When we started, we used 2512 size SMT resistors which were indeed too expensive for the application and concentrated heat in one spot. We ended up replacing them with a whole bunch of smaller (1206) size and significantly cheaper resistors. That dispersed heat much nicer and cost less. Pick and place costs did not change much,a s most of these machines pick up something like 8-10 parts at once anyways

 

[ejonesss]

with the talk of resistors couldnt a resistor be emulated by using some pulse width modulation technique.

if you vary the switching speed of a transistor you emulate the load of a resistor.

basically you are shorting out the cells thousands of times per second.

of course transistors are pretty expensive too i think.

 

[fechter]

I've seen the 'swarm of little chip resistors' approch on some other boards. Anything that increases surface area is good. I guess the price breaks on the smaller sizes are due to higher volume, so it's good to minimize the cost it it works out that way.

Our new layout has themal vias under the power resistors to move heat to the other side of the board better. With a fan, heat should not really be an issue, but it would be nice to not need one. Board real estate is another issue, so making the thing as small as possible is also a consideration.

ejonesss, somebody posted about doing something like what you're talking about. A FET switch was used to breifly 'short out' the cell and run at a high frequency. The duty cycle was very low. This essentially uses the cell itself as a resistor, which would increase cell heating. Depending on the charge rate, this may not be a significant problem. It's more of a problem to make a FET short out the cell without blowing up, but apparently it can be done.

You can also run transistors in the linear mode and make all the heat come off of them. This takes a really big heat sink.

Ultimately, I have an idea for using tiny little flyback transformers to pump the excess charge back into the pack instead of dissipating it as heat. It works on a breadboard :wink: This adds quite a bit of cost and complexity, but if the boards were machine made, it might not be that bad. I have a transformer about the size of 5 dimes stacked that can handle 3 amps. The ones they use to charge camera flashes are even smaller and can do about 1 amp.

 

[reagle]

Something like TI PowerPump where each cell has a DC -DC capable of pumping charge to cells one level up or down? Or boosting to pack voltage and injecting it there?

 

[ejonesss]

in a switching power supply i believe that the transistor is switched so fast that there is no need for a limiting resistor between the transistor and transformer the chopper chip i think controls the switching.

self oscillating power supplies like those cheap ones used in vcr's and dvd players use resistors to limit the current.

in a garden light you know the solar charged ones that use a single aa or aaa battery there is an inductor coil that looks like a resistor and usually yellow or green and is slightly bigger than the other resistors on the board is the heart of the charge pump to pump the 1.2 volts up to enough to power the led.

it may be possible to wind your own coil or use the inductor coils used in the rfi,emp filter section of a power supply.

i have seen in older computer supplies ( during the transition from non atx to atx) there would be a separate regulator board that has a very large coil of wire and would produce the 3.3 volts needed for the memory from the 5 or 12 rail.

another possibility ( you and ggoodrum would know more about if it can be done and cheaply enough) is to have 1 power supply and use an electronic switching system like used on electronic distributers used in cars.

the ignition coil makes the spark and the distributer indexes to the next spark plug.

i think the electronic ones do the same thing using transistors instead of a mechanical switching system. (sort of like what was used in the hamlin cable boxes ( remember the bulky 2 piece slider boxes from the 80's and early 90's?) ( and like hollywood does to set off a rapid sequence of small flash charges to depict machine gun fire)).



of course all of these may add cost to the board.

I've seen the 'swarm of little chip resistors' approch on some other boards. Anything that increases surface area is good. I guess the price breaks on the smaller sizes are due to higher volume, so it's good to minimize the cost it it works out that way.

Our new layout has themal vias under the power resistors to move heat to the other side of the board better. With a fan, heat should not really be an issue, but it would be nice to not need one. Board real estate is another issue, so making the thing as small as possible is also a consideration.

ejonesss, somebody posted about doing something like what you're talking about. A FET switch was used to breifly 'short out' the cell and run at a high frequency. The duty cycle was very low. This essentially uses the cell itself as a resistor, which would increase cell heating. Depending on the charge rate, this may not be a significant problem. It's more of a problem to make a FET short out the cell without blowing up, but apparently it can be done.

You can also run transistors in the linear mode and make all the heat come off of them. This takes a really big heat sink.

Ultimately, I have an idea for using tiny little flyback transformers to pump the excess charge back into the pack instead of dissipating it as heat. It works on a breadboard :wink: This adds quite a bit of cost and complexity, but if the boards were machine made, it might not be that bad. I have a transformer about the size of 5 dimes stacked that can handle 3 amps. The ones they use to charge camera flashes are even smaller and can do about 1 amp.

 

[fechter]

Something like TI PowerPump where each cell has a DC -DC capable of pumping charge to cells one level up or down? Or boosting to pack voltage and injecting it there?

I'm thinking boost to pack level and dump it back into the main terminals. This reduces the complexity greatly and the transformers I'm looking at can go from 3.7v to over 300v. Efficiency of the transformers is not particularly important other than the heating factor. The transformers don't seem to get very warm compared to the switch transistor. Pumping back into only the low cells would be better in theory, but I think it adds way too much to the complexity and invites all kinds of failure modes. Pumping back into the pack will work nearly as well unless you have a really out of balance cell.

Switching 3.7v into a flyback is an interesting problem, since logic level FETs can barely turn on at that voltage. One of my test setups used a bootstrap arrangement to get a higher gate voltage. The other approach is to use a low saturation bipolar transistor. Heat wise, it seems to be nearly the same.

ejonesss, swtiching a FET across a battery cell is not anything like switching an inductor. With an inductor, the instantaneous current is near zero, then it builds with time. Across a cell, there is not much inductance. If you use an inductor, the stored energy has to go somewhere. I designed a circuit a long time ago that could pump excess charge to the next cell, something like a PowerCheq. There are issues with this approach if the cell count is high and the end cells only have neighbors on one side. The 'multiplexed' charger idea has been discussed here before. The stickler there is all the switches and how to drive them at various voltage levels. The other problem with it is the average current in each cell is only a fraction of the charger current since they all take turns.

 

[ejonesss]

i was thinking more and here is an example.

lets say you gut a cordless drill with variable speed.

you have the speed controller switch with a pwm chopper and small fet and a motor assembly.

you pull the trigger haflway then the motor runs at half speed and you also hear the pulse width modulation as a high pitch whistle or something.

lets say you short out the motor wires with the trigger at half.

now you have a very heavy load on the battery and you will drain the battery faster.

you will need a higher current fet possibly to handle several hundred amps witch may raise the cost (unfortunately) very high.


this just a theory i came up with so i dont know you may just fry the fet or it may cause a short that feeds back into the controller (pwm chopper chip) and blow the controller.

same if instead of the motor you used it as a light dimmer and you shorted the bulb outputs.

probably the best is to make each channel a separate charger.

Something like TI PowerPump where each cell has a DC -DC capable of pumping charge to cells one level up or down? Or boosting to pack voltage and injecting it there?

ejonesss, swtiching a FET across a battery cell is not anything like switching an inductor. With an inductor, the instantaneous current is near zero, then it builds with time. Across a cell, there is not much inductance. If you use an inductor, the stored energy has to go somewhere. I designed a circuit a long time ago that could pump excess charge to the next cell, something like a PowerCheq. There are issues with this approach if the cell count is high and the end cells only have neighbors on one side. The 'multiplexed' charger idea has been discussed here before. The stickler there is all the switches and how to drive them at various voltage levels. The other problem with it is the average current in each cell is only a fraction of the charger current since they all take turns.

 

[reagle]

Fechter,
dumping back into the pack is probably easier, though without playing with one it's hard to tell how the currents will end up distributing, Ti's solution is very involved, and while it works well, it requires significant amount of effort (and a lot of parts!). On a related note- if you still need help with converting things to a machine build-able SMT version (and maybe building them in volume), I may be able to help

 

[alangsam1@me.com]

Keep it simple and keep the price low.

 

[fechter]

Fechter,
dumping back into the pack is probably easier, though without playing with one it's hard to tell how the currents will end up distributing, Ti's solution is very involved, and while it works well, it requires significant amount of effort (and a lot of parts!). On a related note- if you still need help with converting things to a machine build-able SMT version (and maybe building them in volume), I may be able to help

We're still trying to move to a surface mount version of Ver.4, so stay tuned. We want to thoroughly test the through hole version first.

The transformer thing is way off since there are quite a few details to work out and it adds considerably to the complexity and parts count. The testing I did was very encouraging however, and I'm sure it can be made to work. I was pumping around 3A from a single cell voltage into a small 120v light bulb and lighting it up. The transformer wasn't even getting warm, but the transistor was toasty.

 

[dnmun]

i wanna clarify what jonesy said about not needing a power resistor on the switching FET.

every switch mode power supply i have seen uses a power resistor on the source leg of the switching FET. i don't have one open now but seems like they are 2 ohms.

you can add a resistor in parallel to drop that value and increase the current into the inductor, and that will push up the power you can deliver to the back end through the inductor at max power feedback from the IC pwm controller..

 

[reagle]

Sure, makes sense. Let me know when you need help.
3A 120V boost from a single cell is some crazy boost! 360W on the output translates to way over 120A on the input side. Or was the voltage a bit lower on the bulb? Either way- quite an impressive achievement

We're still trying to move to a surface mount version of Ver.4, so stay tuned. We want to thoroughly test the through hole version first.

The transformer thing is way off since there are quite a few details to work out and it adds considerably to the complexity and parts count. The testing I did was very encouraging however, and I'm sure it can be made to work. I was pumping around 3A from a single cell voltage into a small 120v light bulb and lighting it up. The transformer wasn't even getting warm, but the transistor was toasty.

 

[GGoodrum]

We are still inching closer to finishing this. Below is what the latest version looks like, a variant for a standalone 12s setup. This one fits fits nicely in the 3.07 x 4.7 x 1.69 Hammond box, and as you can see, I've added replacement end plates. One is for the LEDs and the other is for the connections plus mounts for two 30 x 30mm fans, which are also inside the box.

View attachment 12s LiPO BMS-v3.9y.png


With this variant, I am trying Patrick/methods' idea about using two 3W shunt resistors in place of one larger 5W. That, coupled with some other changes we are making to the cell circuits, allowed me to trim a full half-inch off the length of the boards.

The order for these are in, and I should get these back by Monday. If all goes according to Hoyle, we'll be able to start offering these by the end of next week. I know I've said that before, but now I'm starting to believe it.

-- Gary

 

[jonescg]

So a 32S BMS is no trouble then?

And how do they go with discharge rates? Or are these for charging management only?
Cheers,
CHRIS

 

[GGoodrum]

So a 32S BMS is no trouble then?

And how do they go with discharge rates? Or are these for charging management only?
Cheers,
CHRIS

Yes, 32s is no problem with this version (or the last, for that matter...). I think the practical limit is something like 400-500V, but even higher is possible, with a couple part substitutions.

There is no limit on discharge current, because the individual cell low voltage protection function doesn't directly limit the current. Instead, we have an opto-coupled output that can be used to interrupt the throttle signal, sound an alarm, or trip a relay

-- Gary

 

[jonescg]

Awesome! I'll take one

 

[graemebc]

So I don't have to go back through a bazillion posts, how much for a complete 24 please? Do you sell them complete & tested ? I want to combine 2 x 36/10 psi & headway packs on one bms without having to worry about the a bms from china going pop

cheers

 

[ejonesss]

i like that design .

i am not sure what the 12 circles are in each square on the first segment if they are an array of capacitors that are used for the capacity based balancing bms's or if it is 2 mini transformers made from an emi filter used to make a charge pump so the board can run from a lower voltage higher amp power supply.

i also like the idea that there is a gap between the board segments so they can be cut and stacked.

are these going to be.

1. ready build (just connect wires).

2. ready built except for the large parts like resistors, some capacitors, leds and such.

3. completely diy like the version 1 and 2 boards.

We are still inching closer to finishing this. Below is what the latest version looks like, a variant for a standalone 12s setup. This one fits fits nicely in the 3.07 x 4.7 x 1.69 Hammond box, and as you can see, I've added replacement end plates. One is for the LEDs and the other is for the connections plus mounts for two 30 x 30mm fans, which are also inside the box.


With this variant, I am trying Patrick/methods' idea about using two 3W shunt resistors in place of one larger 5W. That, coupled with some other changes we are making to the cell circuits, allowed me to trim a full half-inch off the length of the boards.

The order for these are in, and I should get these back by Monday. If all goes according to Hoyle, we'll be able to start offering these by the end of next week. I know I've said that before, but now I'm starting to believe it.

-- Gary

 

[ejonesss]

ggoodrum i see another problem with the bms i dont know if it is problem or not.

it is missing the fan 2 + terminal.

i see a fan 1 + and - and a fan 2 - but not a fan 2 +

 

[GGoodrum]

i like that design .

i am not sure what the 12 circles are in each square on the first segment if they are an array of capacitors that are used for the capacity based balancing bms's or if it is 2 mini transformers made from an emi filter used to make a charge pump so the board can run from a lower voltage higher amp power supply.

These holes are for two 30mm x 30mm fans.

i also like the idea that there is a gap between the board segments so they can be cut and stacked.

are these going to be.

1. ready build (just connect wires).

2. ready built except for the large parts like resistors, some capacitors, leds and such.

3. completely diy like the version 1 and 2 boards.

3, for sure, but Andy will offer completely assembled versions as well.

 

[GGoodrum]

ggoodrum i see another problem with the bms i dont know if it is problem or not.

it is missing the fan 2 + terminal.

i see a fan 1 + and - and a fan 2 - but not a fan 2 +

It is on the left side of the board, connected to the large power resistor.

 

[liveforphysics]

Very cool stuff Gary. I look forward to using my first ever BMS!


Thanks for all the hard work!

-Luke

 

[Arlo1]

Yeh Im sick of manualy balancing my batteries as I charge them all the time. Cant wait.

 

[Indubitably]

I'm thinking about using this BMS for a 16s setup with 40ah thunder sky prismatic cells in an e-moped project. They do about 3c max continuous but I figure my application will pull more like .5 to 1c continuous most of the time, and I'd want to cap it at around 2c to be on the safe side. How many of these modules would I need, what would it probably wind up costing me if I bought it tested and pre assembled from tppacks or what ever. Do you guys think my application would be a pretty good fit for this BMS?