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

[methods]

I think the BMS circuit should be left as is for now, its simple and provides all the basic functions everyone needs. If people need active low voltage cut off or surface mount devices then that should be something in the long term. I think having this board will be great and I can't wait to buy it as it is now. I've had to go back to using lead acid after my cyclone made-in-taiwan overly-complicated BMS on my lifepo pack blew up for no reason a couple of weeks ago.

I agree with half of what you said but enlarging pads is a fairly trivial task assuming there is space (and no re-layout is required) Allowing for the use of an SMT LVC in place of the TO92 would expand the LVC range considerably making the board applicable more directly to Lipo chemistry. It is not about an SMT board, it is about the availability of the through hole LVC part.

I fully agree that an active LVC is not called for and would muddy the design. Simple is good. General is better.

I certainly dont want to be another one of those ungrateful butt heads who are always criticizing and asking for more. I would however want to offer suggestions that will improve the design.

-methods

 

[methods]

For any cells that are based on the use of bags, I would definitely do separate packs. Too many failures with these.-- Gary

I would absolutely agree that a lot of the "bag" lipo's available on the internet with very low C ratings are super shady and I would be hesitant to wire them in parallel first then series.

On the other hand I have been flying high quality Thunder Power packs for years and never had a single cell failure (that I did not cause ). Surely you own more, but I own at least 30 TP packs and many have seen hundreds and hundreds of cycles. I fly my 150mph pylon plane with them in public parks and if one of those goes out of control . . . What I am saying is that I have very high confidence in them.

That said, I did not hesitate to wire my 25C lipo packs in parallel then series. The benefits of wiring and balancing are worth the risk to me.

-Patrick

 

[GGoodrum]

Yes, I would not hesitate with high-quality RC packs either, but I what I meant was the less expensive "duct tape" variety LiFePO4-based bag packs.

I undersand about the pad/TC54 problem, but it is going to require some major surgery to get any more room on the board. I thought about adding voltage divider resistors, but couldn't squeeze them in this version. There will definitely be other features and changes will do, as this evolves, so I will add this to the list.

This morning I submitted what I hope will be the final layout, which is shown below. Richard has his working great, but it has so many hacks it looks like his board fell into the blender.

Assuming no more surprises, These should be ready-to-go soon.

-- Gary

 

[methods]

Great Work!

Wow. . . that is tight. Definitely no room for bigger pads.
I will buy whatever comes back from fab, haywire or not.

If you just put the haywiring into the instructions people will think you designed it that way

Features, not haywires. Features.

-methods

 

[ejonesss]

your design looks very good and thanks to all who worked hard on this project and i am looking forward to the new board coming out.

Yes, I would not hesitate with high-quality RC packs either, but I what I meant was the less expensive "duct tape" variety LiFePO4-based bag packs.

I undersand about the pad/TC54 problem, but it is going to require some major surgery to get any more room on the board. I thought about adding voltage divider resistors, but couldn't squeeze them in this version. There will definitely be other features and changes will do, as this evolves, so I will add this to the list.

This morning I submitted what I hope will be the final layout, which is shown below. Richard has his working great, but it has so many hacks it looks like his board fell into the blender.

Assuming no more surprises, These should be ready-to-go soon.

-- Gary

 

[Patriot]

GGoodrum,

Here's a question.

I was initially considering two 16 cell packs in parallel for my battery. However, If I was to decided to wire my battery with 2-cell paralleled packs, then 16 series to make my 48v/20ah pack, I could then use your new board, right?

If so, how well does your board handle a 20amp regen current, when using a Kelly regen controller?



However, if I was to stick to two large 16-cell packs in parallel, would I need two of your BMS's for charging?

 

[GGoodrum]

What got added was some extra logic to shut off the charge current, when the cells are full. This is handled via an SCR that turns on and pulls down the gate on the FET, after the last of the shunts are fully on. What we've learned is that by the time the LED for a channel comes on, the shunt is already almost all the way on. Now, when the last LED is fully lit, the SCR fires, and the charge curent cuts off. This will cause the main LED to turn fully green. This LED starts out red, when the charger is in the CC mode, and then turns a yellowish color as the shunt LEDs start coming on. Once they are all on, it turns solid green. With many chargers, cutting the current like this will also cause the charger to shut down. My NG1 works this way. In any case, leaving the charger on is not a problem. The BMS will balance charge all the cells, and then shut itself off.

 

[GGoodrum]

GGoodrum,

Here's a question.

I was initially considering two 16 cell packs in parallel for my battery. However, If I was to decided to wire my battery with 2-cell paralleled packs, then 16 series to make my 48v/20ah pack, I could then use your new board, right?

If so, how well does your board handle a 20amp regen current, when using a Kelly regen controller?



However, if I was to stick to two large 16-cell packs in parallel, would I need two of your BMS's for charging?

If you parallel the cells first, and then connect the parallel blocks in series, you would only need one BMS board.

Regen is not a problem at all, with this BMS design, as there is no current limiting being done, like with the Chinese BMS variants. During discharge, only the LVC portion is active, and it simply looks at the voltage levels for each cell. No large current loads ever go through this board, in either direction, so regen is not an issue. Since all the current has to go through all the cells, all the time, there is no need to do current limiting at the cell level. It can be done at the pack level (i.e. -- by the controller, or via a CA.). The only reason the Chinese BMS designs include a current limit function is typically the cells/packs these boards are used with can't take moderate loads (30-50A...) that might be common with a lot of setups.

-- Gary

 

[PJD]

Gary or Richard,

This is a bit of a minor thing, but in my old versions (and presumably this version too) the shunt indicator LED's get awful bright, which in conjunction with the flickering, strobe-like in unison, can be vertigo/headache inducing. (Both me and my wife experience this). Someone with epilepsy could have trouble with it.

Can the brightness be reduced with a change in resistor(s) value - or an alternate LED used?

Thanks!

 

[GGoodrum]

Here's the latest schematic;

I'll let Richard describe the circuit details, but basically, when ALL SHUNTS ACTIVE gets tripped (logic low...), Q1 turns on, which turns on the SCR. This clamps the gate driver in the mode that keeps the FET turned off.

 

[Tom_D]

Gary & Richard,
Thanks for carrying the BMS design forward with the new ability to be able to turn off the charge circuit! This will be great for having my charger in the battery box, and not worring about generating unnecessary bypass heat (like all night) after the batteries are done charging.

2 Questions for either one of you:
1. How much heat are you experiencing from this new version of the BMS board ?
2. How many channels did you test for when observing this heat output ?

I have the box already built with a thermostatically controlled fan to come on at 105F and off at 90F. I am just hoping this will work out OK.

Thanks

-- Tom_D

 

[fechter]

With 16 channels running at 500ma the total dissipation could get up to 24 watts so it gets pretty toasty, but will not require a fan if the board is in the open. If you run a fan, even a small one, it should be fine anywhere.

When it's running full blast, the surface temperature of the resistors and transistors gets close to 100C, but not quite hot enough to sizzle spit.

If you use a lower shunt current, like 200ma, the dissipation gets down to where you might get away with sealing the circuit inside a box, but I have not tested this.

 

[wrobinson0413]

.

 

[GGoodrum]

Gary or Richard,

This is a bit of a minor thing, but in my old versions (and presumably this version too) the shunt indicator LED's get awful bright, which in conjunction with the flickering, strobe-like in unison, can be vertigo/headache inducing. (Both me and my wife experience this). Someone with epilepsy could have trouble with it.

Can the brightness be reduced with a change in resistor(s) value - or an alternate LED used?

Thanks!

Hmm, I haven't seen that. I think it is somewhat cell-dependent, and/or relative balance-related. All I've seen is that the LEDs, when they first come on, start out bright, but dim down after a bit. Near as I can tell, when the LEDs are bright, the shunt for that channel is in full bypass mode, which means the cell is about as full as it is going to get. The reason they dim is I think because the charger/supply's CV limit is being hit, so the total current is being reduced. That tells me the cells are not only full, but fairly well balanced as well. In any case, when a cell's LED first cmes on, the transistor and shunt resistor are already in the bypass mode and the cell is close to being full (95%?). When the LED is on bright, the cell is in full bypass, which means the cell is as full as it is going to get. Left unchecked, the shunt circuit would oerload and get swamped. A bit after the LED is on brightly, and the shunt is in full bypass, about to be overloaded, the opto for that channel turns on, which causes the ANY SHUNT ACTIVE to turn on, which causes the gate driver to turn off the FET. With the current temporarily removed, the voltage at the cell will drop a bit, which cause the opto for that channel to go off, which causes the FET to turn back on. This oscillation, or "throttling" is what causes the shunts to keep from being overloaded/swamped, and is what causes the "flicker" you are seeing.

Before this latest change, once the last LED came on and got bright, it would eventually go dim as well. All I would do, though, is wait for the last one to come fully on, and then I disconnected the charger. With this new logic, once the last opto trips, which means the LED is fully on and the cell is full, the ALL SHUNTS ACTIVE signal is tripped, which causes the SCR to fire and the FET to be turned off, and kept off, until power is cycled and the circuit resets.

The main LED has two colors, red and green. These are driven by the complementary outputs of the gate driver. When the FET is on, the LED is red. When the FET is off, the LED is green. Durring the CC phase, the FET is always on, so the LED is red. When the optos start tripping, the duty cycle of the FET gradually changes from on mostly, to 50-50, to off mostly, so the LED color gradually shifts from red, to redish-orange, to orangish-yellow, to a yellowish green. Once the SCR trips, the color snaps to solid green.

 

[GGoodrum]

Hi Gary/Richard

Just a question about the design. From what I have seen, the only hole in the protection of the cells is the case where someone has come off charge and lives on a hill, and uses regen to go down it. The shunts will eat ~0.5A of regen, but the rest will raise the cell voltage. I did not read through the whole thread, just tried to hit the last couple of pages. Do you feed the "any shunt on" signal to an opto to disable the controller in that case, or is it left up to the controller to monitor bus voltage for an over voltage case? Probably for the majority of people the regen case is null point, but if it can happen, it will probably happen to someone.

Wayne

Interesting problem, for sure, but I thought the controllers with regen were supposed to look at the battery level and not try to pump current back in, if the battery was full. Wishfull thinking, perhaps?

I can think of a couple of "workarounds" One is that you could just stop charging when the first few LEDs come on. That would leave the pack at about 85% full, so that big downhill regen would not overload the pack. Another would be to change the value of the voltage divider resistors, in order to lower the shunt cutoff value a bit. That could limit the max charge to about 90-95%, again leaving room for regen.

The problem with having the controller cut out, if the bus voltage gets too high, is that you'd lose the braking force, right? If a setup used this as its main source of braking power, it would certainly make for a wild ride down the hill.

 

[methods]

So assuming that the offset is constant on the LM431 I calculate the following R101 values to get me 4.2V at each cell

To get 3.69V at the cell
R101 = 75k
R102 = 182k
V = 2.613 @ LM431

So logic follows that:

To get 4.2V at the cell (Lipo)
R101 = 111k
R102 = 182k
V = 2.609 @ LM431 (close but slightly low)

To get 4.2V at the cell (Lipo)
R101 = 107k
R102 = 182k
V = 2.645 @ LM431 (close but slightly high)

Yes?

Odd way to use the LM431 but it works.

-methods

 

[fechter]

Hi Gary/Richard

Just a question about the design. From what I have seen, the only hole in the protection of the cells is the case where someone has come off charge and lives on a hill, and uses regen to go down it. The shunts will eat ~0.5A of regen, but the rest will raise the cell voltage. I did not read through the whole thread, just tried to hit the last couple of pages. Do you feed the "any shunt on" signal to an opto to disable the controller in that case, or is it left up to the controller to monitor bus voltage for an over voltage case? Probably for the majority of people the regen case is null point, but if it can happen, it will probably happen to someone.

Wayne

Even when the charger is disconnected, the shunt circuits will still be active and start shunting whenever a cell reaches the set point. If the controller is smart enough to limit the overall battery voltage, the cell shunts should do a fairly good job of equalizing the regen charge.

You could possibly use the any shunt signal to limit the regen if you could interface to something in the controller that would stop the regen. This would make it fairly bulletproof from a regen standpoint.

 

[fechter]

Gary or Richard,

This is a bit of a minor thing, but in my old versions (and presumably this version too) the shunt indicator LED's get awful bright, which in conjunction with the flickering, strobe-like in unison, can be vertigo/headache inducing. (Both me and my wife experience this). Someone with epilepsy could have trouble with it.

Can the brightness be reduced with a change in resistor(s) value - or an alternate LED used?

Thanks!

Boy, I never thought that would be a problem.. How about some shaded plastic?
You could increase the value of R105 to make the LEDs dimmer. This should not affect the operation of the circuit much otherwise. If you use 1k, they should be about 10x dimmer.

 

[michaelplogue]

.... or, just go low-tech and put some paint over the LED's, scratching off a small enough spot to give you the amount of light you want......

:wink:


.

 

[GGoodrum]

So assuming that the offset is constant on the LM431 I calculate the following R101 values to get me 4.2V at each cell

To get 3.69V at the cell
R101 = 75k
R102 = 182k
V = 2.613 @ LM431

So logic follows that:

To get 4.2V at the cell (Lipo)
R101 = 111k
R102 = 182k
V = 2.609 @ LM431 (close but slightly low)

To get 4.2V at the cell (Lipo)
R101 = 107k
R102 = 182k
V = 2.645 @ LM431 (close but slightly high)

Yes?

Odd way to use the LM431 but it works.

-methods

The value of the 1% resistors we are using now are 75.0k and 180.0k, but the technique you are using should work fine.

-- Gary

 

[methods]

The value of the 1% resistors we are using now are 75.0k and 180.0k, but the technique you are using should work fine.
-- Gary

Ok. I got the 182k from the schematic.

-Patrick

 

[GGoodrum]

The value of the 1% resistors we are using now are 75.0k and 180.0k, but the technique you are using should work fine.
-- Gary

Ok. I got the 182k from the schematic.

-Patrick

Sorry, too much resolution reduction, I guess. Attached is a full-sized version, suitable for framing.

 

[wrobinson0413]

.

 

[ejonesss]

i was wondering if the new bms will be expandable?

meaning are there going to be 1 or 4 cell modules that are sold separately so additional cells can be added later?

 

[fechter]

.... or, just go low-tech and put some paint over the LED's, scratching off a small enough spot to give you the amount of light you want......

:wink:
.

Actually, changing the resistors would be a real pain. Paint is not a bad idea. You could also use short pieces of black heat shrink tubing and shrink them down so only a little opening is left.