Open source BMS for 48V to 400V lithium-ion battery pack

ENNOID said:
Thank you Flippy.
This is the kind of request that are useful & will be implemented. This will serve everyone using the product.

i just hope you dont make the mistakes i made when i designed my boards. you will have a solid product when this is all said and done.
 
Solarsail said:
"Does not generate heat" is not the same as 0%.
Yes, it is. You can generate a small amount of heat, of course. A straight wire from the input charge terminals to the cells would generate a small amount of heat. A DC/DC converter would generate a lot more.
Tell me, how much will the temperature of a 20 kg metal box of volume 12 liters rise due to a 5A switching mosfet, where the 60V DC supplied power is rectified, DC-DC converted and conditioned OUTSIDE the box? Numbers please.
Impossible to answer. To answer I'd need the thermal impedance from junction to ambient. If it was perfectly insulated, then the temperature would keep rising indefinitely until you exceeded the 175C limit of the FET and then it would fail.
Have you built such a circuit? I have.
Yes. It's my job.
Take any standard cheap 24V switching power supply. It does not allow for current regulation and current limitation except for short and exceptional conditions. It consists of AC to DC rectification and filtration, then a switching DC to DC step-down converter/isolater both of which generate heat. There is no CCCV function.
Many people here use Meanwell power supplies - cheap Chinese power supplies that convert AC to DC. They have a CCCV function. The CV function is robust and tightly regulated, since most power sinks require tightly regulated voltages. The CC function is less well regulated since ordinarily it is merely circuit protection. But it works well enough for ebike batteries, since most people are not trying to push the charge limits as hard as they can. Batteries don't care much if you charge them at .4C instead of .35C.
The supplied voltage will stay approximately constant 24V and the current will vary according to ohm's law (unless the rating is exceeded). So it has a CV function but not a CC function or the ability to drop the voltage.
Every modern supply you get will have a current regulation function. It will be one of two types - current limiting (the type you want for a charger) or hiccup (shutdown and retry.) This, again, is because they have to do this for circuit protection and reliability. The only challenge for someone attempting to use one as a charger is to ensure that it's the current limiting type. Meanwell spells this out in their data sheet, which is one reason they are popular for ebike chargers.
In addition, the CV function lacks tolerance and may not be suitable for a li-ion pack in CV mode even if your cells add up exactly to 24V at 100% SoC.
The Meanwell supplies I have used are good to within 1% when trimmed, and will hold that regulation even when they heat up.

A switching mode DC to DC step-down regulator with a low Rds at 5A and associated circuitry generates negligible heat. A IRFZ44 at 100% duty cycle will generate .0175 * 25 = 0.43 W of thermal energy.
If it's at 100% duty cycle you don't need the FET. Just use a wire. That's a LOT cheaper and more efficient.

How much switching energy will you lose at 99% duty cycle? That's the loss required to pump the gate capacitance up and down, as well as charge and discharge the D-S capacitance and any snubber capacitance across the switchnode.

How about core losses in the inductor? Depends on core material. Maybe you find a good inductor with Fair-Rite type 76 material. Nice and small but gets a little warm. No problem - go to type 73 material, which has 30% less loss. Uh oh - permeability just went down. Now you need more turns for the same inductance. And all the turns you need don't fit, so you have to go up one size. Now ohmic losses are starting to increase (more wire) so you have to go to a heavier gauge. But with the heavier gauge skin effect becomes more of an issue. Do you do a bifilar wind? Maybe but now you need twice as much thinner wire, and now it doesn't fit . . .

That's the reality when you start with the idea "I'll just make this simple super efficient DC/DC."

Real world DC/DC's run at about 95-96% efficiency if you do a good job in design, and if you keep control of their use cases (i.e. input voltage close to output voltage, max load.) So let's look at your 60V 5A DC/DC. That's 300 watts. A 95% efficient DC/DC will generate about 15 watts of heat. Not a lot - but enough to significantly warm one side of the pack.

If that's OK for you, and you're willing to accept the tendency for that to unbalance the pack, then go for it. But there's no reason to take a simple design and complicate it by adding another switchmode converter.
The rest of the circuit even less. Please let me know when you have calculated the temperature rise.
Sure, as soon as you tell me theta JA.
Cheap off the shelf power supplies are not suitable for charging li-ion packs without a CCCV function. In fact they are unable to drop the voltage.
At this point I'd suggest you get some real world experience with lithium ion charging.

How do you suggest simultaneously charging 18 modules each 14s11p, with the modules hard wired as 3p6s, and with no internal contactors, with just one charger?
I'd get a 60V supply that did current limiting for overload protection, trim it to 58.8V and use that.
Maybe you know how Tesla does this, and I am all ears. I would like to avoid timesharing the single charger and multiplexing the power.
Tesla uses ~400VDC battery packs. When DC charging they use an external 400VDC charger. They also have very extensive battery monitoring and balancing systems inside the pack.
 
ENNOID said:
Yes, SS-24 fits on 100x60mm board, even with a fuse for USB GND :p. I had to be creative for making this fit on a 4 layers board:
Untitled.png

Very nice. Like the "logical" design so far.
Would the board be damaged if you were to hook up the connectors wrong? As you are using the same number of pins on several connectors with (relativly) very little to no visible markings. People are dumb.
 
All balance/temperature connectors have different pin qty. 10 for temperatures (they are signal ),12 & 13 pins for balance.

The actual design prevent the user to connect the balance connector into the temperature receptable (opposite would not damage anything) & can't mix 12 & 13 pins balance connectors as well. but...I cannot build a BMS that can handle wrong balancing cable order made on the connector itself. This is simply not possible to prevent this with the actual IC used on the BMS. There is simply no monitoring IC or simple design approach that can handle mixed voltage level on balancing inputs as far as I know.

I will try to put some labeling on the board, but space is getting limited
 
No problem Ennoid. I am out of here. I will just have to reply to that self-appointed thread pillock though.

Haha, you are wrong. Put six 24S BMS in series. Now try to charge all with one 100V charger. You can't. It will blow up. If you knew how, you would have explained it.

With a relatively small modification to the BMS, any number of modules can be charged simultaneously with one charger, no matter what the S/P configuration. But of course you don't care for the adult world of EVs. Just remain in the skateboard bubble.
 
ENNOID said:
I will try to put some labeling on the board, but space is getting limited
do not worry too much, even is there was enough, people are dumb.

if you can make the space: great. if not, dont worry too much.


Solarsail said:
No problem Ennoid. I am out of here. I will just have to reply to that self-appointed thread pillock though.

Haha, you are wrong. Put six 24S BMS in series. Now try to charge all with one 100V charger. You can't. It will blow up. If you knew how, you would have explained it.

With a relatively small modification to the BMS, any number of modules can be charged simultaneously with one charger, no matter what the S/P configuration. But of course you don't care for the adult world of EVs. Just remain in the skateboard bubble.


please go and make your own thread, stop posting in this one. trust me, you need help. you will kill yourself or worse: someone else if you dont.
 
Been playing with the idea of a 3dprinted PETG plastic case recently…
I thought that some folks might want a case with the SS BMS. It also looks more like a finished product with a case, but I won’t use this myself for my packs.

I don’t know what to think about it honestly. It should at least survive heating from balancing & discharge. It can’t be perfectly sealed with 3d printing & without molding, but it hides most sensible electrical parts… This adds approx. 2mm all around the existing PCB

Anybody have something to say about this? Idea? Use of some metal plate for cooling?
I will put the 3D STL files directly on the website page once the design is good enough.

Here is a pic of an initial attempt for a 3dprinted case for SS-15 board:

6fd7b96d468f17e1fbc03bb1f0d0b4c1783521ff.jpeg
 
looks nice. good addition.

a metal backplate with thermal pads for the balance resistors would be a very useful addition. also adds safety if someone decides to bolt it too close on a metal plate and short out the terminals on the back side when putting bending forces on the cables for example.

but a cover that also covers up the terminals would be more useful imho. stops stuff from falling/dropping on the terminals.
 
Will add a thin top cover on top of the actual cover to protect the power terminals.
I could possibly use aluminum for the flat bottom plate.
 
The original ENNOID-BMS GEN 1 master board & the associated slave boards are using the same firmware & software stack as the newest SS lineup. They are oriented toward high voltage/high power applications which is well beyond e-bikes power level. This forum is mostly e-bike related. So that is why I'm now mainly focusing here on the SS lineup.

GEN 1 master & slave boards are available for ordering on the ENNOID website.
 
cant remember if it has been asked before.

how fast is the current/power measurement? with very "pulsy" loads like controllers or inverters (or basically anything that switches) you can get current spikes that are too short to properly detect wich causes the SoC or consumed power readings to be off over time and causes the readings to get skewed. or is the power readings not really a priority but more of a nice-to-have?
 
how fast is the current/power measurement?

As fast as the power monitor IC can do, I'm normally using the ISL28022 if you want to look into the details. The sampling rate is not handled by the BMS firmware directly, but by the power monitor IC. I'm using the highest number of sampling per reading as possible for better accuracy. I don't think there is much I can do on that front. Current measurements are quite relatively precise honestly. Of course you loose some precision with lower shunt resistor value or with an isolation layer like on master-HV board, but this is to be expected
 
Even fancy dedicated SoC% meters that cost hundreds

only get within 3-8% accuracy, most are much worse and still need to regularly get manually reset at 100%, have the current SoH% manually updated etc.

So no BMS based in basic ICs where SOC is a secondary function will do much better, until you get up to the very large and expensive EV ones, maybe.
 
within 3-8% accuracy

It depends on many factors. Seems reasonable to say within 10% accuracy and probably more like 5% in most cases based on the coulomb counter with the SS boards. Some fancier algorythm as discussed earlier based also on cell voltage could help to improve the precision a bit. This as yet to be implemented...

SOC resets are done every full charge on the BMS. Otherwise, large offsets on real vs calculated SOC would be visible after a few charges.
 
Is the BMS tracking CV stage trailing amps, or just defining 100% from voltage?

The difference would be pretty significant at higher C-rates, say over 0.5C

Although I guess that does not matter too much so long as the stop-charge is consistent each cycle.
 
Is the BMS tracking CV stage trailing amps, or just defining 100% from voltage?

I answered a very similar question already asked by Flippy in a previous post
 
ENNOID said:
Yes, SS-24 fits on 100x60mm board, even with a fuse for USB GND :p. I had to be creative for making this fit on a 4 layers board:
Untitled.png

hey ENNOID, i want to ask how did you manage to connect 2 ltc6811-1 in daisychain without isolation transformer. i am looking for similar approach to make a 24s bms using this ic but want to keep cost low.
 
It is well documented in the LTC68XX datasheets.

Internet transformers are actually quite cheap. I moved away from transformers on the 24S EBMS board mostly for keeping it small & compact.
 
very excited for 24s lite, any updates? the accessoires xt60 connector. is already crimped to lug? stupid question lol. Do you have pics of 24s version? Full vesc BMS data integration?
 
Hey ENNOID,

A question regarding your board layout for the LTC slave boards, what is the purpose of these 'islands' on the nodes of your discharge resistors? Are they there for heat dissipation somehow? They aren't exposed copper so just curious as to why they're in there on all the slave boards. Is there any issues with breaking up the ground plane with these islands? I was under the impression that ground planes should be generally undisturbed.

LTC6811SS.png

Great work on your system, look forward to seeing more.
 
I pulled the PCB files from your GitHub a few months ago. Was just curious to see if they were for any real calculated purpose, as I couldn't really think of a purpose for them other than potential heat dissipation
 
@Ennoid, did you checked this Renesas ISL94203EVKIT1Z for Slave Boards? As they are in 8S model, it will be more better to setup for LFP & NMC chemistries. And also they are cheaper than LTC68 series as well.
 
Back
Top