New 16-cell Battery Management System (BMS)

GGoodrum

1 MW
Joined
Sep 20, 2007
Messages
2,908
Location
South Orange County, CA
As many here are aware, Bob and I have been working on a Charger Management System (CMS) as an external board/unit that would sit between the output of a standard SLA charger, or a regulated power supply, and the pack, connected to the latter via a special 18-pin plug. The reason for splitting off the CMS function from being resident in the pack was because we wanted to use fairly high power shunt regulators on each channel, in order to ensure that each cell can get a 100% charge, at its own pace. Once Bob actually prototyped and tested the CMS circuits, it became clear that the heat was much more manageable and by simply throttling back the charge current a bit, once the cells hit the cutoff (i.e. -- 3.65V) in the final phase of the charge, the heatsink requirements were drastically reduced. Now all we need is a simple plate.

We then both agreed that if we could fit the LVC circuits in with the CMS channels, and stll fit within the same 3" x 6" footprint that would still mount on the front of a LiFeBatt pack, we'd end up with a full blown BMS, with features usually only found in designs used in larger EV applications. It was a bit of a chore to get it all to fit, but we finally got it. Here's what the layout looks like:

16-Cell%20BMS%20v1-03.gif



In normal operation, the LVC circuits for each channel will make sure the voltage for any one cell doesn’t drop below 2.1V. When that happens, a pair of FETs will cutoff power to the negative lead. The charger is now connected directly to the BMS, using two separate wires. This can be any suitable CC/CV charger that can deliver about 3.7V x the number of cells. For a 12-cell 36V pack, that would be about 44.4V, and for 16-cells, the voltage needs to be about 59.2V. Many 36V SLA chargers typically output 44-45V and many 48V models output 59-60V, so this is fine. This can also accept any regulated DC supply that can be set to output the voltage needed (i.e. – 44.4V/59.2V…). because this design doesn’t use the CV portion of the charge profile. Instead, the shunts control the CV portion individually, and lets each cell taper the current off at its own pace.

The way this works is that a depleted pack will start charging at whatever the max current the charger/supply can deliver. Once any one cell’s voltage hits 3.65V, its shunt circuit will start bypassing whatever current the cell doesn’t need. At this point a current limiter circuit kicks in, which uses a FET to throttle back the charge current to about 2-3A (this is adjustable via a resistor, so we will set it so that the heat that needs to be dissipated can be handled by the heatsink/plate...). There is a red LED for each channel, to indicate when that channel’s shunt circuit is operating. Once all the shunts are in full bypass, which typically takes about 10-15 minutes after the first channel starts going into bypass mode, the current limit FET is turned off and the charge process stops. When this happens, there is a separate green LED that lights up, to indicate charging is complete and each cell has received a 100% charge. There is also a temperature-controlled switch that mounts to the heatsink. If the temp gets too high this switch will ground the gate on the FET, cutting off the charge current.

The plate, which mounts on the top side of the board, will serve as the front of the LiFeBatt packs. The heavy-duty shrink wrap goes just over the edge of the plate, so the circuits are protected. Two sets of wires will come through the plate, one heavy set (10-12 gauge) for the main pack leads, and a smaller set (14-16 gauge) for the charger.

Most of the existing SLA CC/CV chargers that are readily available, are limited to 3.5-4A. There are a few others, like the ZiVan NG1, that can output as much as 15-20A, but these are $500 (Patrick has them at http://www.thunderstruck-ev.com). LiFeBatt will also be coming out with some 7-8A "Speedy Chargers", but I'm not sure when those are going to be available. Also coming is a new LiFePO4-specific model from Soneil that can deliver 6A. Another option is to use any one of a number of regulated DC supplies available on ebay for very good prices. Bob found a 0-60V, 0-15A HP supply that normally would cost upwards of about $2k, but I only had to pay $125. For 12-cell/36V setups, you could also use the Mastech 5020, which is sold lots of places online. It outputs 0-50V and 0-20A. In any case, with a 10A charger and this BMS, you could recharge a fully depleted pack in just a little over an hour. With a typical 2A Chinese "duct tape" charger it takes 5 hours to fully recharge an empty 10Ah pack, and then another 10 hours to have the RC-styled balancer circuits used in the companion duct tape BMS board, bring down the voltage of the stronger cells to that of the weakest.

Bob has done a great job of not only designing a feature-rich, "big boy"-styled BMS, he was able to accomplish this using a minimum number of fairly inexpensive, readily available parts. My value-added is limited to doing the PCB layout. :) I'm not sure of the exact costs yet for this new BMS, but it will be less than what separate LVC and CMS units together would cost. Well under $100, in any case.

We will have the first boards and the parts for these by Wednesday. I'm hoping that we will be able to complete the testing by the end of next week. We will then finally start selling LiFeBatt 12-cell and 16-cell packs with this integrated BMS, by the beginning of the following week. We will also offer this BMS separately, with heatsink/plate, for those that would like to use it with other LiFePO4-based packs, duct taped, or not. :)

As soon as I get the first on together, I'll post some pics in this thread.

-- Gary
 
Excellent Posting

Not only is the title of the thread appropriate, the description provided of what you are doing is well done, the project does a great job of covering all the bases and you've even stretched out into the issue of thrid party energy sources for the charger.

If everyone starts posting this well there will be nothing left to complain about. :D


(and now that I've spent some time grappling with LVC logic processes I can appreciate the work it takes to get it right)

This should elevate the LifeBatts to a higher level in that your BMS will provide better full lifecycle performance while allowing the user to have no clue what is going on. (the way it really needs to be eventually)
 
That sounds really great Gary!

The only thing that would concern me is the dissipation on the current limiting FET. I can see somebody packing that thing inside a bag where it could build up some heat. The overtemp safety is a very good idea. It might be possible to do a switching mode current limiter, but inductors tend to be expensive, so the linear approach might be most cost effective.

I'll look forward to some test results.
 
Great news!

What if we wanted two 16 cell packs, one for each side of pannier pair Can we get away with one electronic circuit, CMS with a schotky (sp) wired to the pack. Seperately charging them if needed.
I realize I'd have to open the pannier when charging, but would i need two CMS boards?
 
Sweet design Gary,

I wanted to clarify one aspect of the BMS. You mentioned that a pair of FETs shut down the negative lead when LVC is reached. Does this mean that the battery pack would simply refuse to supply current to my controller when a cell LVC condition were reached. If so, this is very good news. I was assuming that the pack BMS was going to supply a signal lead that I would connect to my controller telling the controller to stop asking for current. The former approach makes the battery pack less dependent on its external environment, but I was assuming that it would take too much circuit board real-estate (e.g. FETs) to do efficiently in the pack.

--Joey
 
Gary,

According to Ping . None of his packs can be ran in a series. I am gussing this new bms would allow that.

Arbiker.
 
Most excellent work :D! I'll have to snag one when I get some cash for lithium.
 
fechter said:
That sounds really great Gary!

The only thing that would concern me is the dissipation on the current limiting FET. I can see somebody packing that thing inside a bag where it could build up some heat. The overtemp safety is a very good idea. It might be possible to do a switching mode current limiter, but inductors tend to be expensive, so the linear approach might be most cost effective.

I'll look forward to some test results.

The current-limiting FET is also mounted to the main heat sink, so it should be okay. Bob decided to go the linear route because the same FET could used to also cut the charge current completely, when the shunts are all in full bypass. Space was at a real premium. :)

-- Gary
 
recumbent said:
Great news!

What if we wanted two 16 cell packs, one for each side of pannier pair Can we get away with one electronic circuit, CMS with a schotky (sp) wired to the pack. Seperately charging them if needed.
I realize I'd have to open the pannier when charging, but would i need two CMS boards?

You can use one BMS if you connect all the cell junctions together from each pack That will parallel each cell with its "mate" in the other pack. The BMS will treat both cells as one. No need for a Schottky at all.

-- Gary
 
Arbiker501 said:
Gary,

According to Ping . None of his packs can be ran in a series. I am gussing this new bms would allow that.

Arbiker.

I never understood this. I can't imagine how a BMS/pack would even know if it was connected in series with a second one. In any case, that's not the case with ours, either the BMS, or the pack. I use two 12-cell LiFeBatt packs in series with no problem.
 
Joey said:
Sweet design Gary,

I wanted to clarify one aspect of the BMS. You mentioned that a pair of FETs shut down the negative lead when LVC is reached. Does this mean that the battery pack would simply refuse to supply current to my controller when a cell LVC condition were reached. If so, this is very good news. I was assuming that the pack BMS was going to supply a signal lead that I would connect to my controller telling the controller to stop asking for current. The former approach makes the battery pack less dependent on its external environment, but I was assuming that it would take too much circuit board real-estate (e.g. FETs) to do efficiently in the pack.

--Joey

Yes, in this design, the LVC circuits have an active cutoff, so that it can be used in a wider selection of setups. There are two high-power FETs use for this functioln, and they together are good to 100A.
 
GGoodrum said:
Arbiker501 said:
Gary,

According to Ping . None of his packs can be ran in a series. I am gussing this new bms would allow that.

Arbiker.

I never understood this. I can't imagine how a BMS/pack would even know if it was connected in series with a second one. In any case, that's not the case with ours, either the BMS, or the pack. I use two 12-cell LiFeBatt packs in series with no problem.

If I remember correctly, it had to do with the MOSFETs that worked the LVC. If they weren't adequately rated for it, two packs in series would fry them.
 
Indeed, on the point1 lipo's the fets would fry under certain circumstances. Mainly, it was OK whilst running series untill one of the BMS's tripped (LVC or overcurrent), at which point one set of fets would die horriibly.
 
I guess the linear charge current regulator won't see a very large voltage drop if the supply voltage is in the right range, so maybe dissipation won't be a big deal.

I was thinking, however, the if the charging current was PWM'd (with NO inductor) rather than linearly limited, the average current to the cells could be controlled without the FET needing to dissipate any significant heat. I'm not sure what the response of the chemistry would be to a pulsed charging current. If the PWM frequency was high enough, the voltage on the cells would have some ripple, but only a few millivolts I'd guess. The shunt regulators should still function with the ripple.
 
EXCELLENT! I'm especially impressed that you fit all of this on that board using through hole components! (btw. great idea to include the current limiter FET. )

If I wanted say, a 10-cell BMS, could I just cut two of the Shunt/LVC circuits off of the right of a 12-cell BMS board and jumper a trace or two? I also hang around over at the Parallax Forums And it's not unusual to see someone carving up a Propeller Proto Board or one of it's brothers to shoehorn into small projects. :twisted:

Now someone just needs to find a bike size 10A+ PSU for this BMS. That'll really get the electric meter spinning when opportunity charging. :mrgreen:

my 2 coulombs,
Marty
 
lawsonuw said:
If I wanted say, a 10-cell BMS, could I just cut two of the Shunt/LVC circuits off of the right of a 12-cell BMS board and jumper a trace or two?

Yes, you can use this board with 10 cells, just by not populating the six channels on the right, and then running one jumper wire.
 
thanks to everyone for all the positive comments. i do listen and you all will all help us to refine the features. i needed a bms for a customer i promised it to last year, and it looks like the LifeBatt bms will not be quite appropriate for ebike applications, since most ebikers don't need the pack to call the factory by cell phone and report failed subsystems. I decided the path of least resistance was to build the ebike bms myself, and gary and i have worked together on it. he deserves all the credit for the neat pc layout. Gary wanted to make the first system with through hole components for easier assembly by anyone who wants to just buy the boards or buy it as a kit. We will likely convert the design to surface mount if the volume justifies it.

The design is as usual full of compromises that make it imperfect, like most things. Anybody wanting a perfect bms should not use it, and design their own. The reason i am using a linear mode fet in the current limit portion is mainly because i want the lowest possible drop, to be able to use chargers that might kick out if they see the load voltage going up or changing rapidly, as some might with anything other than a linear device with minimal voltage drop. It was also of course a space and cost issue. We will consider going to pwm charge current regulaton when we take the board to surface mount if it looks like there is a good reason. On the first systems we are using a .005 ohm shunt in the charge path that permits adjustment of the current so it can be adjusted to the heat sink and the projected cell mismatch. Our experience to this point indicates it should only be a matter of a few % mismatch between cells, or there is a problem that needs attention. We are also considering the charge cycle as an episodic event, so the heat sink is not designed to handle indefinite operation. It will get hot in operation but not hot enough to be a problem, and the system will shut down on over temperature and resume the cycle when it cools down.

We can adjust the time after the last cell hits 3.65v before the charge circuit is interrupted; there is also an adjustable time for an automatic cycle reset. The system will also reset by disconnecting and reconnecting the battery. Those of us who have played with this chemistry know that there is not a big voltage difference between 8A or 2A charge rates, so the fet should not need to have much voltage across it and should not dissipate too much power. It also lets us work with supplies that produce only 3.65v per cell, since we are using a fet with very low on resistance.

The question has arisen about using the bms in series. The reason this is a problem for any design which uses fet(s) to shut off the power on an error condition is that the fet(s) in the cutoff circuit needs to withstand the full battery voltage when they are in the off state. When the fet is off, the load is still at battery voltage with respect to ground. High voltage fets are more expensive and have higher on resistance, so nobody would want to use a higher voltage fet than they needed. This is likely the problem with stacking some bms systems if for example they use a 55v fet that works fine for one pack but blows when you use two packs. On the first systems we will be using the 4110 fets which are rated at 100v. These fets will handle much more than 50A but the pcb traces may not so i would put the limit there, and recommend using a 50A fuse or smaller. If you need more than 100v pack voltage you can replace the cutoff fets with 150v parts.

The shunt regulators are independently connected to each cell, so that part of the system is expandable almost indefinitely, as is the low voltage detector. Both functions use optic isolators to OR the low voltage signals and shunt active signals from each cell. the output signal is still available to shut down a controller with an ebrake, and it is possible to use this function in addition to the cutoff fets in the bms. For cases where people want to use more than 16 cells in series we can make it work with a couple of tweaks.

Expected current drain is under 5 ma. If a separate wire is provided from the charger to power the charge management circuitry this could be cut to <1 ma for the lvc only.
 
I'm going to purchase 16 X 2, Lifebatt's in a few weeks time like many ebikers i'm sure that live in the northern hemisphere.

For those of us that are "tinkerer's but not electronic experts, can we assemble this BMS kit ourselves if we have soldering experience? I understand the USA have a sueing dis-order and you have to protect yourself from other peoples screw-ups, but rest assured that most of the developed world cannot, and would not do this, if we screw-up.
OR, would you like us to order completed boards. Is the difference really that much?

I still don't understand clearely if we actually need these devises if we have a "Cycle Analyst" meter, and moniter the Lifepo4 batteries ourselves.
 
recumbent said:
I'm going to purchase 16 X 2, Lifebatt's in a few weeks time like many ebikers i'm sure that live in the northern hemisphere.

For those of us that are "tinkerer's but not electronic experts, can we assemble this BMS kit ourselves if we have soldering experience? I understand the USA have a sueing dis-order and you have to protect yourself from other peoples screw-ups, but rest assured that most of the developed world cannot, and would not do this, if we screw-up.
OR, would you like us to order completed boards. Is the difference really that much?

I still don't understand clearely if we actually need these devises if we have a "Cycle Analyst" meter, and moniter the Lifepo4 batteries ourselves.

yes kits will be available to build them yourself and it is all easy soldering. as to whether you need it or not, that may be debatable. if you are very careful about your low voltage cutoff and your pack stays balanced you don't need the bms. the problem is the pack may not stay balanced. we just don't know what will happen 1000 cycles down the road, but lab testing is indicating the cells themselves can do several times that. the problem with charging cells in series is that as soon as the first cell gets fully charged its voltage starts to rise, and the charger may hit its voltage limit and cut off well before the last cell is charged. this could be a cumulative effect resulting in a single cell being discharged beyond recovery. In a system of 16 or even 12 cells it is very difficult to tell by the pack voltage the difference between normal discharge and a cell dropping out until it is too late and an expensive cell is destroyed. Sometimes the cells recover, but they are never the same again.

by using a clamping circuit across each cell we prevent them from going above 3.65v or whatever voltage is desired, thus giving the other cells a chance to catch up. Testing on the LifeBatt cells indicates that eventually the current will drop to about 6 ma. if they are held indefinitely at 3.65v. The pack performance will always be limited by the weakest cell, but if we are sure to charge every cell to 100% each time and watch each one to make sure we do not discharge it too deeply we can get the best performance possible out of that weakest cell and all the rest. there are lots of much more sophisticated systems involving pulsed charging and fancy microprocessor control, and they may be worthwhile, but most just cut off the charge when the first cell hits 4.0v and cut off on a pack low voltage rather than watching each cell. These batteries are expensive, so it seems worthwhile to do all we can to get the best performance out of them. This system is the most general and simplest configuration i could come up with for now. I am sure we can do better but i don't know how much difference it would make. Only time will tell.

as for discharge cycle balancing, where the pack shuttles charge from one cell to another until all are dragged down to the same level; we figure it is better and easier just to charge them all to 100% in the first place, hopefully making that kind of balancing unnecessary. If we become convinced that is necessary we could add that capability. We could probably get the bms to mix margaritas if there was enough interest, but it would not make the cells last any longer.
 
Are there any unique features or issues in paralleling these battery packs? I have been considering using two packs with same voltage in parallel. Are Shottky diodes required?

-- Joey
 
Joey said:
Are there any unique features or issues in paralleling these battery packs? I have been considering using two packs with same voltage in parallel. Are Shottky diodes required?

-- Joey

in operating principle they are the same as the a123 cells, and many people have been using those cells in parallel clusters successfully. The most important thing is to never let the cluster drop below the critical 2.1v level or all the cells will die. This is my major concern with using cells in parallel; the possibility that one may develop a short or discharge faster than normal for some reason on the shelf, dragging down its parallel neighbors until all are damaged. Connecting the cells in parallel makes it more difficult to isolate a bad cell, but as i said lots of people seem to be doing it.

For myself, I usually isolate strings of batteries with .55v drop diodes. If the voltage of one string is just tens of millivolts higher than the other these cells will move charge from one string to the other. They are very efficient in terms of charging losses, so maybe that does not matter, but it is a source of loss we can prevent. Let's say a bike with parallel packs in the panniers is parked in the sun and half the battery pack is in the shade. If the subsystem voltages differ diodes will prevent any cross current. If i had two serial strings connected together with diodes and one of these bms systems on each, the first string to have a cell drop below 2.1v would cut out but the second string would continue to provide power. In this case one bad cell does not shut you down. I fear that with parallel groups of cells one bad cell can be a show-stopper, if it drags down its fellows in parallel. That does not seem to be a real problem with these cells if one is careful. The bms is intended to shut down the bike before the batteries are damaged no matter who is riding it, and to charge each cell fully every time.

I am not sure if there is a warranty issue with connecting cells in parallel. I will ask Don about that one.
 
There's no issue with paralleling cells. Lifebatt already does this for the 20Ah and 30Ah configurations they are doing for their other applications. As for whether or not it it is better to parallel cells first, or use separate strings, I guess a case could be made either way. In my experience, however, the only way I've been able to kill/damage an a123 cell is by over-discharging it. This happened with one 48V setup, where I had 4 separate 16-cell strings, that were connected in parallel at the string level. One day I forgot to hook up two of the four 16-cell strings and then went off on one of my normal rides, which usually consumed about 7Ah out of the combined 9.2Ah available. When I got about 2/3rds into the ride it was like a fuse blew, or power was cut. It was then that I discovered that only two of the four strings were connected. It turned out I had 2-3 cells in each of the two 16-cell strings that had voltages well under a volt. the rest were up around 2V, I think, and they fully recovered. The bad ones did not recover.

Had this setup been constructed with 16 groups of four cells in parallel, I would not have killed any cells. Having LVC circuits on each cell would have prevented this as well, but it was actually this event that started my on a quest to add low-voltage protection. As I said, I've never in two years, seen an a123 cell fail, or even lose capacity, unless it was over-discharged. I've seen a123-based packs survive numerous horrific crashes in RC helicopters, and work fine for hundresds of cycles after being regularly abused with peak discharges well over 100A, which for a 2.3Ah cell is about 40-45C. A lot of these very cells are in some of my ebike packs right now. The point is, I highly doubt any LiFeBatt cells will fail sitting on a shelf, or in normal use. Maybe the cheap "duct tape" variety might have longer term reliability problems, but I think even for these, failures that aren't induced by misuse are going to be rare.

Anyway, I certainly don't see a problem with paralleling two 16-cell LiFeBatt packs, as long as they both have BMS boards. I also don't think you need to worry about using Schottky diodes to isolate them. Sure they will "equalize", but it will be at the pack level. The voltage for the higher one will be lowered slightly, but this reduction is spread evenly across all 16-cells. The differences are going to be small, in any case. I have two 12-cell LiFeBatt packs I use in series and even though they are charged separately, with different chargers, I'm amased that both packs end up within about .02V of each other, every time I use them. I still use one setup with 5 16-cell sub-packs that are connected in parallel at the string level, without using Schotly diodes, and it still works great.

-- Gary
 
one of the reasons gary and i started to work together on the bms for the lifebatt systems was that the combination of my engineering experience and his practical experience abusing this chemistry in rc helicopters for a couple of years seemed like a good match. i have said before that i just don't know about using the cells in parallel, but he now has more experience with the lifebatt cells to add to years of a123 experience, so if he says parallel operation is not a problem i have no basis to disagree.

if the cells are connected in parallel first before being combined into strings then only one bms board could be used, so it would still need to be fused at 50A. the bms would prevent any parallel group of cells from draining below 2.1v, and since we have no evidence these cells do short or self discharge at high rate this should be fine. It would seem that even charging these cells in parallel should not present a major problem, as the ones that have the highest capacity will keep the lower capacity ones from reaching full charge potential early. The bms will cap the voltage of each parallel group of cells to 3.65v, so they should all reach full charge at the same time by charging at slightly different rates. Lab tests on the lifebatt cells indicate that they will drop to .006A charge current with the voltage capped at 3.65v so there seems to be little danger of overcharging the parallel clusters if the bms is used to monitor each group of cells. If serial strings are connected together they could use separate bms boards each fused at 50A to provide greater current output. The limiting factors are the cutoff fets rating and the actual pcb traces.

One advantage to using serial strings connected in parallel is that you could probably charge the separate strings at a higher rate than you would want to charge the string of paralleled cells, so the latter would likely take longer to charge. It also makes it much easier to locate a defective cell, but we aren't expecting too many of those. I tend to err on the side of caution, and i am still learning, so i prefer to separate the strings with schottky diodes so i can independently monitor each one. This is probably unnecessary. The diodes do however let me hook in an auxiliary pack of a different voltage and chemistry than my main pack without any problems. Most people will never need to do that.
 
It would certainly make it easier for using two packs in parallel if they each have their own BMS board, because they could be connected together at the pack level. Adding Schottky's wouldn't hurt, though, especially if you ever did want to use two different types of packs.

-- Gary
 
Back
Top