New 16-cell Battery Management System (BMS)

[chas_stevenson]

When they are available I would like 4 of them as I now have 4 battery packs that need them. I have not used the batteries waiting for these units. Some one please let me know when they are ready.
Contact info: chas_stevenson@yahoo.com

Grandpa Chas S.

 

[GGoodrum]

Since the cat is about half out of the bag (http://endless-sphere.com/forums/viewtopic.php?f=14&t=2498&p=69700#p69700), I've decided to post here what Richard and I have been up to for the last couple weeks. With Bob still more out-of-action, than in, I went to Richard to see if he wanted to collaborate on getting something done for a workable BMS solution. As a design goal, I wanted something as reliable, and hopefully as simple as the proven LVC circuits, at a minimum, not requiring complicated heatsinks or time-consuming calibrations/adjustments. What we came up with, assuming it works ( ), is a clever way to still allow each cell to reach its own 100% level, without worrying about what its neighbor cells are doing, but still allows the use of a standard SLA CC/CV charger/supply.

I wanted to wait, this time, until we had a working unit, before going "public", so I'm not going to post the schematic, just yet. I will, however, be happy to show what the first boards, which I ordered this morning, look like:

There's still quite a few parts, but the assembly is pretty easy, if not time-consuming. This time, I'm going to stick to just providing kits. Those who don't have basic soldering skills will have to find someone to put them together for them, as I'm not going to go into the board assembly business. Maybe some of the members here with younger eyes can offer to put these together for a fee.

Once we've had a chance to test the first ones, I'll post the schematic, for review and comments. Basically, the way it will work is that during the first constant current (CV) charge mode, the BMS circuits do nothing, just like before. When the cells hit the cutoff point (3.65V-3.70V...), the circuits will kick in, and force the charger/supply into a pulsed mode where the cells are intermittantly give the full charge current, until all the cells are at that point. There is a bicolor LED for each channel that will blink/flicker a yellowish color, while in the "pulsed" mode. When all the cells are at that point, the total amount of current will start dropping, just like in a "normal" CV mode. Eventually, the current will drop to the point that the pulsed mode ends, and then the LED turns a solid green. When all the LEDs have stopped "blinking" yellow, and have all turned green, the cells are about as full as they can get.

We are going to start testing these new boards later this week. Please don't start bombarding us with questions, purchase requests, etc. Nobody wants this any sooner than either of the two of us do, and I promise to post our test results, good or bad, just as soon as we can. The good news is that the parts are readily available, and fairly cheap, and that the boards only take 3 days, to my door, so even if we have changes, that won't be a hangup.

More later...

-- Gary

 

[fechter]

Sorry Gary. I just can't keep a good thing quiet for long.

Don't contact Gary and ask him about the BMS until he announces it's ready to go. It will be at least a couple more weeks.

There is still room for improvement in the design, and there is a good chance I screwed up on something and will need to make changes. Things are moving along quickly though.

The design has already gone through about 10 iterations. The design objectives were: to allow high charging rates and at the same time not dissipate much heat from the BMS unit, it has to keep all the cell voltages in the desired range, LED indicators for each cell to locate weak cells and monitor charging progress, minimize parts count and use through-hole parts for hand assembly, easily adapt for any number of cells, easily scaled to higher power, use an existing charger or switching power supply, compact size (actually the list goes on, but you get the idea).

The parts count is still pretty high, but that's somewhat unavoidable if you want good functionality. I was designing for a maximum charge current of 20 amps, but heat might limit it to a lower value. Anything less would be fine.

 

[rf]

Great news, guys! Nice compact board.

Thanks for the flashing lights, gotta have those!

 

[GGoodrum]

Thanks for the flashing lights, gotta have those!

Anything to keep you kids amused...

Until we actually test the first complete board, we're not sure whether they will blink, or flicker. It depends on the pulse/oscillation rate, which we really can't predict just yet.

-- Gary

 

[Ypedal]

+2 on the LED's !!!

It's great to walk by and give a quick glance to see if all the cells are fully charged. vs wipping out the dvm 10 times per charge cycle to see what's going on..

On average, 16 of my 12ah chinese cells will ballance and go from orange to green with 10 to 15 minutes of each other if drained past 50 % dod.. less and they all stay ballanced to within 5 minutes of charge on the VP chargers.

Keep up the good work ! i'll wait patiently. :wink:

 

[rf]

Thanks for the flashing lights, gotta have those!

Anything to keep you kids amused...

Until we actually test the first complete board, we're not sure whether they will blink, or flicker. It depends on the pulse/oscillation rate, which we really can't predict just yet.

-- Gary

If we see a mushroom cloud on the horizon we'll know the rate requires adjusting ...

Good luck!

 

[Stevil_Knevil]

Maybe some of the members here with younger eyes can offer to put these together for a fee.

*raises hand* :wink:

 

[karma]

same here allways willing to help if needed :wink: but ill be assembling my order first

 

[mcstar]

I can solder them together if someone can get the boards and a parts list. I'm really in intersted in getting one of these for my ebike. My BMS died yesterday. I'm charging blind now.

 

[rf]

I've been using some LVC circuits from Gary for several weeks now. They've been working wonderfully and have added much piece of mind. Much better than the controller's full-pack based LVC which cost me some battery cells before.

I'd like to suggest an improvement to LVC design, especially those connecting to the controller's brake or throttle circuitry. Currently, this LVC takes the brake or throttle signal low when it senses a low voltage on one or more battery cells. This works fine. But what happens if a cable is accidentally severed or a connector fails or becomes unplugged? The LVC protection is then compromised and the battery is at risk.

Critical external connections should probably be more fault tolerant. Similar to air brakes on a train -- if connections fail, the system doesn't continue to work without the user being made aware of the problem.

To maintain protection future designs should probably disable the throttle if cabling fails or becomes detached.

It would be good to follow this sort of philosophy throughout any design.

Richard

 

[Randomly]

This was the approach I took on the LVC design I posted for an Ebrake/throttle output. If you hard wire the components after Q7 (Q8, it's associated resistors and Q11) to the controller ebrake input, and use the connection of the collector of Q7 to the base of Q8 as your connection line, running a wire between these two points it will be fault tolerant. If the line breaks or is disconnected the ebrake input will be pulled low. The Ebrake input will only be released when it receives the proper signal from the LVC.

 

[GGoodrum]

If you wantthis to be trully fault-tolerant, you'd have to change the controller end. Right now the weakest link in the chain is the connectors between the LVC board, and the controller. It doesn't matter what you have in the way of fault tolerance on the LVC board, because if the cable going to controller comes unplugged, it's not going to work. What I do is to tape the connectors together to lessen the chance they can inadvertantly come apart. Beyond that, I'm not sure what else you could do. The wire itself is not likely to fail on its own. :?

The circuit itself is also very fault-tolerant. Each channel is completely independent of the othrs, and each output is connected together in parallel. There's no other single ponit failure. I've never had a single failure with any of these boards, once properly connected and working. There are only 3 parts per channel, and they are extremely reliable. The only failures I've ever had were related to inadvertant shorts and connection problems, initially hooking them up to the packs. In all cases I've seen, the TC54 chip got fried, and it fails pullig the opto low, so you know right away that there's something wrong.

-- Gary

 

[GGoodrum]

I've been busy, testing the new BMS board. I got the first channel working, along with the FET logic that is used to pulse the charger current.

The way it works is that the charger supplies its full current until it reaches the cutoff point. When that first happens, the green portion of the LED first starts to come on, just as the shunt transistor/resistor combo first starts to conduct. Just before the shunt gets swamped/overwhelmed, another part of the logic gets tripped that turns on an opto output, and the red portion of the LED. The opto outputs are used to pull down the gate of a FET that is inline with the charger negative lead. This cause the current to be cutoff, which then will cause the cell's voltage to dro slightly, resetting the shunt logic. This then turns off the opto, which turns the FET back on. This oscillation, or "pulsing" keeps the shunts right at their thresholds. As all the cell gets fuller, the current drops below where the shunt gets swamped, and the green portion of the LED becomes fully lit, and the red goes off.

The circuit actually worked the first time I tried it, but it needs some tweaking. I needed to adjust the values of the LM431's voltage divider resistors a bit because the cutoff/crossover point initially was at 3.85V, instead of the desired 3.65-3.70V. I changed one from 84.5k down to 75k (the other is 180k...) and that dropped the cutoff down to 3.694V. I also want to play with the values for the current limit resistors for the red-green LED. In the pulse mode, the red portion is actually going on an off but the rate is fast enough that you don't really see it "flicker", per se, but it is very dim. The green is not fully on either, but the green tends to overpower the red from a visual perspective, so the net result is just a light green color. You can't really see the color in the picture above, as the camera flash washes it out, but you can see that both portions are on. There is definitely a very noticable difference between the pulsed mode, and the point where just the green is fully lit, but still, I think it might be better if the red portion was a bit brighter. I think by tweaking the resistor values, I can do that.

Today I'm going to populate three more channels, and try doing four cells at a time. I'm going to purposely discharge four a123 cells to different levels, and see how long it will take to fully charge all four. I also want keep an eye on the shunt resistor temps. We might need to tweak the values a bit, if they start getting too hot.

Anyway, I'm pleased with the progress. Already we are way passed the point I ever reached with the old design.

-- Gary

 

[Deepkimchi]

Hello Gary,

Since the divider selection is critical, what tolerance of resistors are you using ? 1% ?

DK

 

[PJD]

But regarding the LVC, is there an easy way to test it's function?

I rigged mine to pull down the throttle signal. and shorting the "ebrake" leads does shut the throttle down, so it is good to that point. I then thought I could just disconnect a cell connection to the LVC to test the LVC itself it but this doesn't seem to work. Do I need to actually lower the voltage to just below 2.1 volts using a pot across the cell terminals?

 

[GGoodrum]

Hello Gary,

Since the divider selection is critical, what tolerance of resistors are you using ? 1% ?

DK

Yes, with 1% resistors you can control the range between 3.65-3.70V. What I hadn't counted on was a built-in offset for the LM431. The output of the divider needs to be 2.606V, instead of 2.50V. By using 180k and 73.2 for the two values, the 1% range is 3.645V-3.687V. The actual values of what I have now are 180.4k and 75.0k, and the cutoff point is 3.694V.

Tomorrow, Richard will build up the same configuration and then put it up on a scope to see what sort of oscillation rate where seeing. There was a concern that if it was too fast, switching losses would cause the FET to heat up, but this is clearly not the case, as the temp never got above 94F.

-- Gary

 

[GGoodrum]

But regarding the LVC, is there an easy way to test it's function?

I rigged mine to pull down the throttle signal. and shorting the "ebrake" leads does shut the throttle down, so it is good to that point. I then thought I could just disconnect a cell connection to the LVC to test the LVC itself it but this doesn't seem to work. Do I need to actually lower the voltage to just below 2.1 volts using a pot across the cell terminals?

Yes, that's right. You have to have the voltage go below 2.1V for the opto to trip. I test each channel with a 0-5V supply, and then just monitor the opto outputs with a multimeter set to the diode tester mode. The resistance mode will work fine as well.

One way you could test is to disconnect a cell and simply use a single 1.5V AA or AAA battery. The minimum the TC54 chip needs is 0.7V. Anyway, the TC54 is active low, so it needs to see the actual voltage below 2.1 in order to pull down the opto LED, turning it on.

-- Gary

 

[jeffkay]

Hey Gary, I ran this by Richard but maybe you will like it because of your "4 pack" methodology: If you have to do another board run, please consider a mod/mode for using 4 "12v" Lead acid chargers? This would be pretty handy and most of them are isolated outputs...

Jeff K. "Deep Cycle" project

P.S. Hurry up and get this done, will ya!
LOL

 

[Patrick]

Hey,

I've missed the latest circuits, but I have a question. No problem if I want to use this with a higher voltage system, right? I can series the charger inputs, and parallel the opto outputs, and all is good, yes? Yes, may need a few extra jumpers, and an extra board (or two if am VERY lucky), but less parts overall and it's all good?

Thanks for the great work!

Patrick

 

[GGoodrum]

Hey,

I've missed the latest circuits, but I have a question. No problem if I want to use this with a higher voltage system, right? I can series the charger inputs, and parallel the opto outputs, and all is good, yes? Yes, may need a few extra jumpers, and an extra board (or two if am VERY lucky), but less parts overall and it's all good?

Thanks for the great work!

Patrick

Yes, it should scale up fine. The LVC circuits don't draw but microamps, and the main pack leads don't go through the board at all. The optos from multiple boards can be paralled as well, regardless of whether the boards/packs are in series or in parallel, or both.

The charging portion should also scale up. I think the single FET and the board traces we're using ought to be good for a 20A charger, but we'll have to test that, of course. I've got a 48V Zivan NG1 I got from you guys I'm going to try, and it'll put out 16-17A, if the cells are drained down enough.

-- Gary

 

[fechter]

You could string any number of them in series up to the voltage limitation of the charge control switch (100v). If you could interface the charge control opto couplers directly to the charger you could possibly bypass this limitation.

I'm interested in seeing if it can interface to a Zivan through the temperature probe connector.

Virtually any switching mode charger uses an optocoupler inside to control the output. These are usually easy to identify. It should be fairly easy to tap into one and use it for current control.

If the onboard FET switch works good enough, then this would be unnecessary and you could use any CC CV power supply for charging as long as it has the right output voltage.

 

[rf]

I understand the BMS is targeted for use with lead-acid type chargers since they're somewhat plentiful. It will also apparently work with a simpler power supply. My question is: how simple?

I suspect A123 cells wouldn't have a problem with a very simple, cheap, noisy, power supply. What about the BMS? Will it require a higher quality supply? Will it tolerate a lesser supply? Or will it help make up for a cheap one?

Just curious.

Wondering if a high power, low cost, simple supply could be built by mere mortals. Most everyone seems to shy away from attempting to build a welding machine/lethal weapon (high power) supply. But the prices of high output chargers are pretty high themselves. Building our own would also seem to expand the selection of voltages -- good for those of us with less standard pack sizes.

Richard

 

[GGoodrum]

I don't know about noise, but as long as there is some sort of current limiting, and a max voltage that is in the range needed (i.e. 3.7 x the number of cells...), I think it would work fine. There's also a ton of supplies on ebay that would work as well. I got an HP 6274B 0-60V 0-15A supply for $125.

-- Gary

 

[mcstar]

Can someone post a link to the latest circut schematics, parts list and board layout? I'm itchin to build one of these.

Thanks