New 16-cell Battery Management System (BMS)

[PJD]

Chas,

The LVC boards are just a low-voltage cutoff to protect the slightly weaker cells from over discharge when the pack is discharged. The BMS regulates charging voltage to each cell, plus has a LVC too.

 

[chas_stevenson]

So I guess I want to wait for the BMS for my A123 batteries. I plan on 5P 12S for a 36-volt 11.5 AH pack. I have been watching TPpacks: http://www.tppacks.com/products.asp?cat=26 for them.

Thanks again,
Chas S.

 

[bobmcree]

while the parts can be laser trimmed to spec and are available in any of a large number of voltages. 2.1v, 2.4v, and 2.7v are the only ones i have found in stock anywhere, so it would take a big order to get the other parts. any voltage > 2.1v can be set by using a resistive divider on the input of the 2.1v parts, and to avoid error >1% due to the 1 microamp bias input current of the tc54 the voltage divider should use 100ua so for a value in the 2-4v range you would use a divider in the 20-30k range that wasted 100 ua instead of the 1 ua the chip needs.

Gary,

Is the cutoff point of your LVC cutoff board easily adjustable? What is the default cutoff voltage?

Thanks,

Paul D.

Yes, it is adjustable, via different versions of the TC54 voltage detector chip. There are versions 2.1V to 6.0V, in .1V increments. I think for most LiFePO4 cells, 2.1V is the right number. I've tested LiFeBatt packs with LVC's using the 2.1V chips, and run the packs down to the cutoff. At that point, the Ah used is right around 9.8 - 9.9 Ah. For a123-based packs, I use the 2.7V version, mainly because a123 cells hold the voltage up higher than any other cell I've seen, all the way to the very end. Then, the cells dump quickly. Using 2.7V as the cutoff ensures that this quick voltage drop is detected as soon as possible.

-- Gary

 

[PJD]

The generally accepted cutoff for the Thundersky LFP cells among it's users is 2.5 volts.

I assume that physically fitting the divider resistors - or maybe a 20-30K trimpot - on the board is doable?

What the status on the BMS kit availability? Thanks.

 

[GGoodrum]

24.

The generally accepted cutoff for the Thundersky LFP cells among it's users is 2.5 volts.

I assume that physically fitting the divider resistors - or maybe a 20-30K trimpot - on the board is doable?

What the status on the BMS kit availability? Thanks.

Yes, but that may be for cases where the discharge rates are fairly low, like 1C, or less. These are fairly low C-rated cells, so I think it would be quite easy for the cells to drop down to 2.5V, under typical bike loads, especially towardsthe end of the duration. I still believe the 2.1V version is the right one to use with most LiFePO4-based cells, with the exception of a123 cells. the C rating for a123 cells is so high that even 50A loads won't pull the voltage of healthy cells below 3.0-3.1V, if at least two cells are paralleled. The voltage stays this high, pretty much until the end, when it dumps quickly. I chose the 2.7V version for these so that this sudden drop could be detected as soon as possible.

Bob is still testing the latest version of the BMS boards, to verify that the the new temp controls work as expected. As soon as he is happy with it, we can finally start making them avilable.

-- Gary

 

[bobmcree]

as gary says i am debugging the latest boards right now. when considering the cutoff voltage one must realize that our system monitors the instantaneous voltage of each cell at all times, rather than sampling it occasionally as a microprocessor based system might do, so we can use a lower voltage to predict the end of useful charge life than a sampling system. we will first see the low voltage excursion at a high current drain, and provide immediate feedback to the user by limiting the power at that instant. this lets us sense the end of charge by looking at the cell voltage under load, rather than looking at the average voltage.

if you have a 2.1v sensor and you want to raise the voltage you can put on a pair of resistors quite easily. there is adequate room on the board. the only time i would use the 2.7v sensors would be for multiple parallel a123 cells.

 

[PJD]

Any new news of the status on the BMS?

 

[GGoodrum]

Any new news of the status on the BMS?

I'm still waiting on Bob, to finish the BMS testing. He needs to check that the latest boards, with the thermistor-controlled current reduction, work as advertised. I tried putting one together, but I can't get the current limiter to work at all, much less with the temp protection. I don't have the same test setup as what Bob has in his lab, which is probably why I'm haing so much trouble.

Anyway, as soon as he checks it all out, we can move forward. I've gone as far as I can go, and I'm working without a net.

-- Gary

 

[PJD]

Gary,

I hope the bugs get worked out soon, I may have moved on my project a bit prematurely.

A question: I am charging my individual cells (Thundersky 40 AH) to 3.7 volts each in order to get each cell to an equal state of charge. I notice that upon reaching 3.7V at 3A, the CV stage, down to 0.4A, takes about 90 minutes.

What was the default timer setting going to be for the BMS? Was the CV charging stage duration taken into account?

Thanks again!

 

[chessir]

"These were at our request replaced with a new BMS circuit that would have virtually no quiescent current, causing some of the original delays to the shipment. As a downside though, it turns out that the rate of cell balancing with this new BMS is many orders of magnitude less than with the prior circuit. When we began testing the packs from this shipment we were typically only getting between 10 to 11.5 amp-hours out of them, due to the cells not being balanced. By opening up the packs and individually discharging those cells that are more charged than the rest of the pack, and topping up those cells that are running low, we've been able in almost every case to restore a full 12 amp-hours of available capacity. The images below shows a case in point."

Quote from ebikes.ca concerning quiescent BMS current which ultimately can destroy the cells.

 

[rf]

[...]
By opening up the packs and individually discharging those cells that are more charged than the rest of the pack, and topping up those cells that are running low, we've been able in almost every case to restore a full 12 amp-hours of available capacity.
[...]
Quote from ebikes.ca concerning quiescent BMS current which ultimately can destroy the cells.

That discharge part sounds odd? Charging all cells to their maximum would seem the best approach.

Discharge `balancing' is a technique used with matched, small cell-count R/C packs to try to get by without a BMS. The intent is to cause all cells to deplete simultaneously to keep from damaging them. Probably not appropriate for ebike packs -- especially not for LiFEPO4 cells with their steep discharge curve.

`Balancing' cells on ebike battery packs seems wrong, in almost all cases. Full charge is what we want.


Richard

 

[GGoodrum]

Gary,

I hope the bugs get worked out soon, I may have moved on my project a bit prematurely.

A question: I am charging my individual cells (Thundersky 40 AH) to 3.7 volts each in order to get each cell to an equal state of charge. I notice that upon reaching 3.7V at 3A, the CV stage, down to 0.4A, takes about 90 minutes.

What was the default timer setting going to be for the BMS? Was the CV charging stage duration taken into account?

Thanks again!

The time constant is adjustable, from about 15 minutes, to about 2-3/4 hours. With large capacity packs, you'd obviously want a higher time, or no time limit at all, which is my preference. In any case, I think we can make the time constant used as a selectable option.

The way the current version works, once all the shunts are active, meaning all the cells are at some point in the CV profile, the timer starts. In your case, two hours would seem to be an appropriate value. This would allow all the cells to receive a full charge. At the end of this time period, a second timer starts, which can also be varied from about 15 minutes to about 2-3/4 hours. At the end of this period, the whole system resets and the charge cycle starts again.

My preference actually, is to have no timers, and no automatic shutdown. In this version, the LED is red, during the initial CC mode, orangish when at least one of the shunts are active, and green when they are all operating. This way, if I do have one cell that is significantly out-of-whack with the rest, it can have every opportunity to "catch back up". I dooubt this will be an issue with any of my LiFeBatt-based packs, but I have a number of a123 packs that have some "stressed" cells that can benefit from a little extra "balance" time.

-- Gary

 

[GGoodrum]

[...]
By opening up the packs and individually discharging those cells that are more charged than the rest of the pack, and topping up those cells that are running low, we've been able in almost every case to restore a full 12 amp-hours of available capacity.
[...]
Quote from ebikes.ca concerning quiescent BMS current which ultimately can destroy the cells.

That discharge part sounds odd? Charging all cells to their maximum would seem the best approach.

Discharge `balancing' is a technique used with matched, small cell-count R/C packs to try to get by without a BMS. The intent is to cause all cells to deplete simultaneously to keep from damaging them. Probably not appropriate for ebike packs -- especially not for LiFEPO4 cells with their steep discharge curve.

`Balancing' cells on ebike battery packs seems wrong, in almost all cases. Full charge is what we want.


Richard

I agree. I think the new BMS boards are typical Chinese variants, with discharge-type balancing circuits. What makes me think that is because he says "the rate of cell balancing with this new BMS is many orders of magnitude less than with the prior circuit." This resistive-typ cell balancers only discharge cells at a max of 100-150mA, typically. Individual CV charging for each cell is the only way I know of to allow each cell to reach its own 100% level, whatever that happens to be. You can do this with shunt-type voltag regulators, like we are doing, or you can use individual cell chargers like many here have done.

-- Gary

 

[justin_le]

That discharge part sounds odd? Charging all cells to their maximum would seem the best approach.

Discharge `balancing' is a technique used with matched, small cell-count R/C packs to try to get by without a BMS. The intent is to cause all cells to deplete simultaneously to keep from damaging them. Probably not appropriate for ebike packs -- especially not for LiFEPO4 cells with their steep discharge curve.

`Balancing' cells on ebike battery packs seems wrong, in almost all cases. Full charge is what we want.
Richard

I agree. I think the new BMS boards are typical Chinese variants, with discharge-type balancing circuits. What makes me think that is because he says "the rate of cell balancing with this new BMS is many orders of magnitude less than with the prior circuit." This resistive-typ cell balancers only discharge cells at a max of 100-150mA,

I am a bit confused as to what is being deduced here. The BMS circuits that are presently on these battery packs will dissipate a small amount of current (<20mA) from the cells that have reached their full charge voltage, allowing other cells time to catch up in the charging process, so that ideally the end of charge voltages all end up pretty similar. Unlike a proper transistor system that clamps the cell voltage though, in this case it is either on or off for each cell, and the current bled away is dissipated through LEDs. There is no balancing that takes place on the discharge cycle or at any time when the charger itself isn't connected.

The balancing BMS that we had in the sample packs last fall used an active switched capacitor system, and if one cell was low over an AMP of current could flow into it from all the other cells, regardless of whether a charger was connected or not, by simply shuttling the charge around through the capacitors. This system was always on, so if the pack was fully charged or connected to a charger then it would guarantee 100% charge on all cells. Likewise, if you let it sit for a while when the pack was flat, it would balance the cells to all have the same end-of-charge voltage. If all the cells are well matched for capacity then these two states are basically equivalent, and in either case (all 100% charged or all matched at the end of charge) the total available capacity you could get from the battery as a whole is still going to be limited by the lowest capacity of any one cell.

typically. Individual CV charging for each cell is the only way I know of to allow each cell to reach its own 100% level, whatever that happens to be. You can do this with shunt-type voltag regulators, like we are doing, or you can use individual cell chargers like many here have done.
-- Gary

Of course, the shunt-type voltage regulator is infinitely better from a connector and wiring perspective!
-Justin

 

[PJD]

I know impatience isn't virtuous, but Bob has been absent from all the ES fora for more than a week now. Anyone know if he is OK?

And, if he is having probelms with the lifecycle BMS, does anyone have any leads on a suitable alternate 16 cell BMS?

 

[GGoodrum]

I am a bit confused as to what is being deduced here. The BMS circuits that are presently on these battery packs will dissipate a small amount of current (<20mA) from the cells that have reached their full charge voltage, allowing other cells time to catch up in the charging process, so that ideally the end of charge voltages all end up pretty similar. Unlike a proper transistor system that clamps the cell voltage though, in this case it is either on or off for each cell, and the current bled away is dissipated through LEDs. There is no balancing that takes place on the discharge cycle or at any time when the charger itself isn't connected.

Actually, this is not how the clamps work, Not in a digital on-off fashion. What happens is that as the cell first hits the cutoff value (i.e. - 3.65V), it is still absorbing current at the max rate put out by the charger. With the voltage held at 3.65V, the cell starts gradually reducing what it wil let in and the shunt starts to bypass the rest of the charge current. The next cell in series will see the full amount of current, which is important if it hasn't yet reached the cutoff point. Each cell is allowed to reach its own completely full state, at its own pace. The cells may never be balanced, but it doesn't matter. As long as you also protect each cell from being discharged too low, it doesn't matter.

Also, the LEDs have nothing to do with the process, other than to indicate that each shunt is active. In the version I'm still testing, they aren't even installed. The current is bypassed via the large TIP105 Darlington-pair power transistors. With a big enough heatsink, these could handle as much as 8A. With the heat sinks used below, 1-2A is the max, so we have an op-amp - controlled current limiter that throttles back the charge current to that level, once the first shunt starts to work. In the latest version, we've added thermistors on each heatsink that will reduce the current necessary to keep the heatsink temps to < 150F.

The balancing BMS that we had in the sample packs last fall used an active switched capacitor system, and if one cell was low over an AMP of current could flow into it from all the other cells, regardless of whether a charger was connected or not, by simply shuttling the charge around through the capacitors. This system was always on, so if the pack was fully charged or connected to a charger then it would guarantee 100% charge on all cells. Likewise, if you let it sit for a while when the pack was flat, it would balance the cells to all have the same end-of-charge voltage. If all the cells are well matched for capacity then these two states are basically equivalent, and in either case (all 100% charged or all matched at the end of charge) the total available capacity you could get from the battery as a whole is still going to be limited by the lowest capacity of any one cell.

Actually, this sort of "bucket-brigade" type of balancing idea intrigues me, but you are right, though, that they will always balance to the level of the lowest capacity cell. With resistive-type discharge balancing circuits, however, the same thing happens. The weak cells don't really ever "catch-up", but instead the stronger ones are eventually bled down to the level of the lowest cell.

The fact of the matter is that no matter how carefully the cells are initially matched, for capacity and internal resistance, eventually, over time, they will drift apart. Most of this is due to temperature differences that each cell sees, over the life of the pack, based on where the cells are located within the pack. The worst case is for the inside cells in prismatic "bag"-type packs. The inner ones will run hotter than the ones that are on the outside. Even the highly regarded a123 cells are prone to major temp variations between cells. The point is that over time, cell capacities can and will vary. Maybe it is only 3-4%, and if so, maybe you don't care, but with our approach, differences don't matter. We just let each cell charge to its own 100% level, whatever that might be.

typically. Individual CV charging for each cell is the only way I know of to allow each cell to reach its own 100% level, whatever that happens to be. You can do this with shunt-type voltag regulators, like we are doing, or you can use individual cell chargers like many here have done.
-- Gary

Of course, the shunt-type voltage regulator is infinitely better from a connector and wiring perspective!
-Justin

Yes, not having to bring out all the junction wires is certainly easier, but even this doesn't have to be such a big deal. For my own a123 and LiFeBatt-based packs, I'm using LVC-only boards, like the one shown below. Each of these have a board-mounted multi-pin connector. I have taken one of the new BMS boards, left off the LVC parts, and create a "CMS" board that mounts in a small extruded aluminum case. The charger plugs into one end of the CMS, and there is an output cable with a matching multi-pin connecter that then is plugged into the one on the LVC board.

My main reason for going this route is that I have multiple 12-cell and 16-cell LiFeBatt packs, as well as several 12s and 16s a123 packs, all that already have the LVC boards installed on them. I can use a single CMS unit to charge all of these packs, using existing 36V and 48V chargers. Actually, with the 12s/12-cell packs, which I always use as pairs in series, in a 72V configuration, I'll use two CMS units and two chargers, just so I can do the pair together. The 48V setups I don't use near as often, so I'll just use one of the CMS units and a my 48V Zivan NG1 to charge them.

The other reason for doing the CMS version is that by using the extruded aluminum case as an additional heatsink, the charge current used during the CV/shunt phase can be cranked up higher. I've also had numerous requests from those that also have existing setups with the LVC boards already installed, and a couple that have multiple pack setups, like I do, and want the separate CMS for the same reasons. Anyway, I will post some pictures of the completed unit later, after I take some and get them edited. Here's a shot of the board, before the wires, etc., were added:

-- Gary

 

[GGoodrum]

I know impatience isn't virtuous, but Bob has been absent from all the ES fora for more than a week now. Anyone know if he is OK?

And, if he is having probelms with the lifecycle BMS, does anyone have any leads on a suitable alternate 16 cell BMS?

As of an hour ago, Bob ws still alive and kickin'. He's just been busy, trying to run down the last of the issues with the BMS boards. The problem is that whenever a major change is made, like we did recently, to add the temp controls on the current limiter, we both feel it is important to fully test the complete circuit, just to be sure we idn't break something that used to work. This just takes a lot of time, but I'd much rather do this than have to worry about fixing a bunch of boards down the road, for some problem that crops up later.

The other issue is that Bob has a bad back, and can only spend so many hours a day on doing anything. Some days are beter than others. Yesterday, we made a lot of progress, and right now I believe that he is ttacking down the final issue on the list, which is making sure that the current limiter kicks in when any of the shunts first start working.

Another reason it is taking so long is that we are trying to make this work with the widest variety of setups that we can, using a very wide range of chargers/supplies with charge currents from 2A up to 10A, or more. Maybe this level of flexibility is trying to do too much, but I think not. We certainly could've narrowed the operating range, and finished sooner, but we both felt the extra flexibility was worth the extra time, and expense. I'm afraid to add up what I've spent on PCB runs, for boards that will never get used. I think we are up to our 5th "production" run. I'm encouraged by our most recent round of testing, because so far, at least, neither Bob or I have found anything that would require a change to the latest board layout.

In any case, most of the time that Bob spend on the forums is when his back is not letting him do "real" work, so I, for one, am okay with his temporary absence here. :wink:

-- Gary

 

[rf]

Any reason the new 'active cutoff' shouldn't also become main power-on?

Power-on surge with most controllers causes problems, from wearing out connectors and switches to requiring half-on arrangements with resistors, etc.

Any reason not to connect a switch to the `active shutoff' on the BMS as master power?

Thanks

Richard

Sorry, Richard, I missed this earlier...

I agree, I think this is totaly doable. I believe all you would need to do is ground the gate signal on the FETs, to prevent any current flow, and then release the ground to let in the "surge".

Right now, for my 72V setups, I'm using using a couple of big power resistors and a momentary pushbutton in order to "pre-charge" the controller caps, before switching on main power. This has completely eliminated the big "zaps" I was getting before, which were especially bad after the bike/packs had not been used for a couple weeks. This, however, seems even simper. I might give it a try on my Mariner folder, which is slated to get two new 12-cell LiFeBatt packs, with these LVC boards.

-- Gary

I thought about this some more and realize that a delayed turn-on instead of or in addition to the switch would be better. If the user connects the battery system with the switch inadvertently set to the on position then the connectors will still suffer surge damage. If power-on was delayed a second or two whenever load is connected that would go a long way to `fool proofing' that aspect of the system.

Richard

 

[PJD]

Gary,

My existing 48 volt chargers I hope to use are 7-8 amp with 59.5 volts CV mode.

If the BMU cuts the charge amperage immediately down to 2 amps upon the first cell reaching the voltage cut setting, This will cause the cell's charging voltage to drop considerably, resulting in a long time to get the voltage of the cells back up to the voltage cut value, then some more time to allow the curent to taper during the CV stage to the charger's setting. This extended charging time may be tolerable for the smaller packs used in e-bikes, but may be very long for 40ah or larger pack as are used in scooters or MC's.

What is the largest batttery pack Bob will be testing the BMU on? I'd be a willing guinea pig to test a prototype BMU on the large pack sitting on my bench if Bob has a spare BMU to loan.

Also, this is from an electronic neophyte, but what is the source of all that heat from Darlington pair? is there another type of transistor with lower on-resistance and power disapation that would be available to solve the heat problem? Attach those heat sink bars to a large finned aluminum case?

 

[lawsonuw]

Also, this is from an electronic neophyte, but what is the source of all that heat from Darlington pair? is there another type of transistor with lower on-resistance and power disapation that would be available to solve the heat problem? Attach those heat sink bars to a large finned aluminum case?

All the heat coming from the Darlington pair is due to it carrying any and all of the extra power provided to the cell it's managing. Since the current circuit doesn't provide a way to move this extra power to someplace it can be used, it's just converted to heat. Just changing to a different type of transistor isn't going to make the heat go away. Adding additional heat sinking is an excellent idea, and will allow higher currents to be used when topping off cells.

Marty

 

[GGoodrum]

As Marty points out, the bigger the heatsink, the more total current that can be bypassed.It should be noted that the way it works in a battery pack, all the current has to go through all the cells, so if you just limited the voltage to each cell, and didn't bypass whatever the cell doesn't use, the strongest cell would start limiting the current for the whole pack, which means the lowest ones may not even get to the cutoff, much less, receive a full charge.

I haven't tried the BMS/CMS on a 40Ah pack, but it should work fine. Most of the charge cycle time is during the initial constant current mode, when the BMS/CMS doesn't do anything. Once the cells first hit 3.65V, they are at the 85-90% level, and once there, the current going into the cell starts to reduce. When the shunts kick in, the BMS/CMS will reduce the current dwn to about 1-2A. Initially, this drop in current might raise the voltage up higher than the cutoff, whch could cause the current to oscillate back and forth a few times but this won't hurt anything.

 

[justin_le]

I am a bit confused as to what is being deduced here. The BMS circuits that are presently on these battery packs will dissipate a small amount of current (<20mA) from the cells that have reached their full charge voltage,

Actually, this is not how the clamps work, Not in a digital on-off fashion. What happens is that as the cell first hits the cutoff value (i.e. - 3.65V), it is still absorbing current at the max rate put out by the charger. With the voltage held at 3.65V, the cell starts gradually reducing what it wil let in and the shunt starts to bypass the rest of the charge current.

Oh sorry, I thought the original poster was referring to the somewhat problematic BMS circuit that we just got which came with the batch of LiFe packs from China since it had quoted from our homepage update. I wasn't talking about the 16cell BMS you guys are developing here which is by all means is a more intelligent design. In the ones we got, there are three blue LED's per cell channel and in this particular case the LEDs are indeed the dissipating element rather than a shunt transistor on a heatsink.

The fact of the matter is that no matter how carefully the cells are initially matched, for capacity and internal resistance, eventually, over time, they will drift apart. Most of this is due to temperature differences that each cell sees, over the life of the pack, based on where the cells are located within the pack.

For sure. And it seems even for packs stored on the shelf and not even used, the internal self discharge rates vary enough that by the time you get around to using it the cells are often out of balance again. I'd be interested to see if other people here have statistics on the deviations, but our own experience is that in almost every case that we have received a shipment of lithium battery packs, by the time they get here (usually ~6 weeks after pack assembly), then there are already measurable balancing issues in a large number of the packs, even though the companies assembling the batteries swear all cells were originally at the same state of charge.

Anyways, it's quite encouraging to see the development of a local ebike-sized BMS circuit take place by you guys, since almost without a question that is the main impediment to lithium packs getting used with PEVs right now. There are lots of companies that can produce pretty great cells, but none seem to have done the requisite BMS circuit for 10-16 cell batteries very well.

As an example, another problem we have often encountered is that the supplied BMS circuits have a built-in short circuit protection, but the surge current that flows into the capacitors on some of the motor controllers is enough to trip the BMS. This is especially true with the larger 35-50 amp crystalyte controllers, as soon as you plug in a lithium pack the BMS shuts off. These parts just haven't been designed with the whole picture in mind.

Here's a shot of the board, before the wires, etc., were added:
-- Gary

It is very tight and pretty! -Justin

 

[Jozzer]

As Marty points out, the bigger the heatsink, the more total current that can be bypassed.It should be noted that the way it works in a battery pack, all the current has to go through all the cells, so if you just limited the voltage to each cell, and didn't bypass whatever the cell doesn't use, the strongest cell would start limiting the current for the whole pack, which means the lowest ones may not even get to the cutoff, much less, receive a full charge.

I haven't tried the BMS/CMS on a 40Ah pack, but it should work fine. Most of the charge cycle time is during the initial constant current mode, when the BMS/CMS doesn't do anything. Once the cells first hit 3.65V, they are at the 85-90% level, and once there, the current going into the cell starts to reduce. When the shunts kick in, the BMS/CMS will reduce the current dwn to about 1-2A. Initially, this drop in current might raise the voltage up higher than the cutoff, whch could cause the current to oscillate back and forth a few times but this won't hurt anything.

We've built a BMS thats very similar, with mostly the same spec, and use it on my aprilia's 45AH pack. After the first cell trips the voltage drops for a few mins, but as gary says, since the pack is very nearly charged at this point, the cells are soon back at the cuttoff and trickling nicely to full charge.
I had considered variable current limiting to speed the topoff, but it hardly seems worth it TBH.

 

[Doctorbass]

I am a bit confused as to what is being deduced here. The BMS circuits that are presently on these battery packs will dissipate a small amount of current (<20mA) from the cells that have reached their full charge voltage,

Actually, this is not how the clamps work, Not in a digital on-off fashion. What happens is that as the cell first hits the cutoff value (i.e. - 3.65V), it is still absorbing current at the max rate put out by the charger. With the voltage held at 3.65V, the cell starts gradually reducing what it wil let in and the shunt starts to bypass the rest of the charge current.

Oh sorry, I thought the original poster was referring to the somewhat problematic BMS circuit that we just got which came with the batch of LiFe packs from China since it had quoted from our homepage update. I wasn't talking about the 16cell BMS you guys are developing here which is by all means is a more intelligent design. In the ones we got, there are three blue LED's per cell channel and in this particular case the LEDs are indeed the dissipating element rather than a shunt transistor on a heatsink.

The fact of the matter is that no matter how carefully the cells are initially matched, for capacity and internal resistance, eventually, over time, they will drift apart. Most of this is due to temperature differences that each cell sees, over the life of the pack, based on where the cells are located within the pack.

For sure. And it seems even for packs stored on the shelf and not even used, the internal self discharge rates vary enough that by the time you get around to using it the cells are often out of balance again. I'd be interested to see if other people here have statistics on the deviations, but our own experience is that in almost every case that we have received a shipment of lithium battery packs, by the time they get here (usually ~6 weeks after pack assembly), then there are already measurable balancing issues in a large number of the packs, even though the companies assembling the batteries swear all cells were originally at the same state of charge.

Anyways, it's quite encouraging to see the development of a local ebike-sized BMS circuit take place by you guys, since almost without a question that is the main impediment to lithium packs getting used with PEVs right now. There are lots of companies that can produce pretty great cells, but none seem to have done the requisite BMS circuit for 10-16 cell batteries very well.

As an example, another problem we have often encountered is that the supplied BMS circuits have a built-in short circuit protection, but the surge current that flows into the capacitors on some of the motor controllers is enough to trip the BMS. This is especially true with the larger 35-50 amp crystalyte controllers, as soon as you plug in a lithium pack the BMS shuts off. These parts just haven't been designed with the whole picture in mind.

Here's a shot of the board, before the wires, etc., were added:
-- Gary

It is very tight and pretty! -Justin

Justin, have you any pics of these LEDs BMS ?

 

[GGoodrum]

My experience with cells in packs sitting on the shelf pretty much mirrors what Justin has seen, except less so with a123 cells. they seem to even keep up the surface charge for a considerable period of time. I don't really have enough data on the LiFeBatt cells just yet, but one difference I've noticed with these is that they will "self-bleed" the surface charge, right after a charge, so that within minutes, all the cells are down to about 3.40-3.45V. Once there, however, they seem to be able to hold this level for weeks, without any change, whatsoever.

I also agree that none of the Chinese BMS designs really seems to have it right, just yet. Ours is the other end of the spectrum, but it is quite expensive to make. Maybe someday there will be something in the middle that maybe doesn't quite have the charging robustness of our current design, but at least won't shut off at the mere threat of a higher current pulse.

I finally finshed the first CMS unit. Here's what it looks like:

I'm going to do some final "real world" tests with this on a 16-cell 48V/10Ah pack, and then I will finally make these available on my site, along with 36V and 48V 10Ah packs, and 36V/48V Soneil chargers.

Bob is testing the final change to the full BMS units, and has managed to eliminate a couple of parts, which is always good. As soon as he verifies that everything is working the way it should, we will make these available as well.

-- Gary