[jhuebner]

I don't want to steal the other thread and be all negative about the ideas there. So here is a new one. I'm thinking a lot of over-engineering is going on there, both in diy and commercial

Observations both first hand and by others

• Lithium batteries vary slightly (belowlike 3%) in capacity
• They vary even less in self-discharge
• They also vary slightly in internal resistance

Now, capacity and internal resistance differences is nothing that a BMS should try to fix. Only varying self discharge is what makes cells drift apart in SOC over prolonged periods.
The period can be so prolonged that some do fine with no BMS at all after careful initial balancing. I don't want to promote that as it can (and has) go(ne) wrong.


Next option is to ignore SOC drift and just mitigate its effect. That is monitoring only solutions that shut down operation once certain voltage limits are hit on any cell. I'm using that but I also don't want to promote it as I find myself topping up cells that have drifted away too far once a year.


Now a very popular approach is top balancing on charging by shunting a given current over the cell. That solution will kick in even if the cell just passes the voltage limit because of higher internal resistance. In that case what should actually be done is give cells a cutoff voltage proportional to their higher resistance. Top balancing also often happens due to component drift especially on analog systems. I'm inline with Fraunhofer institute if I say: forget top balancing.


Balance when idle
A better time to balance is when there is no current flow at all and when the dielectric effects have settled down. That is 1-2 hours after the last current has gone through the pack. Now differences in internal resistance will no longer bias your measurement. What you do need now is an accurate (say +-2mV for lfp and +-10mV for '3.7V chemistries'). Per month maximum self discharge is often specified 3%. In practise it is lower. And the difference in self-discharge is lower still. Say 0.5%. With that, 100Ah cells would drift apart by 0.5Ah a month! So you have one month time to suck away 0.5Ah from the cells that self discharge the least. That results in a whopping balancing current of 0.6mA!
What this means is that you can do away with a cheap 0603 resistor and matching transistor instead of large wirewound devices.


Don't try to eliminate parasitic draw
Instead minimize it. Modern controllers can do away with 1mA or less. A multi channel measuring module will work just fine with megohm divider resistors. This all results in an approximate loading difference of about 50-100micro amps. Much less than the self discharge! Forget individual opamps or expensive high voltage multiplexers.


Don't even think about active balancing
Or individual cell chargers. With the above said thats a solution to a problem that does not exist. Say you have a 100 cell pack and you have to drain 0.5Ah from half of the cells every month. That is 92.5 Wh per month. Can the efficiency of your single cell chargers or active balancers keep up with that?

 

[Jerry Liebler]

I fail to see a reason for all the antagonism. Perhaps it can be explained. The BMS that I'm developing happens to incorporate charge only active balancing but that is a bonus that comes with a very effective isolation method, using tiny low cost high frequency transformers. These transformers allow extremely low parasitic losses when powered down. By using active "rectification" on the secondary (cell) side very accurate measurement of cell voltage can be done on the primary side of the isolation.
You take a very strong stand against active balancing yet you acknowledge a 3% variation in cell amp hour capacity. Without active balancing the capacity of the pack (amp hours) is that of the weakest cell and no more. With active balancing that is 100% efficient and "up to the task at hand" the pack capacity is the arithmetic average of the cell capacities. Without active balancing a single weak cell WILL reduce the pack to it's capacity (amp hours). In addition, with "conventional BMS technology, either top balancing or bottom balancing, the depth of discharge will be greater on the weak cells shortening their life relative to the stronger cells. Here again active balancing forces the depth of discharge to be the same for all cells lengthening the life of weaker cells. The "holy grail" of battery management is full time active balancing that is "up to the task" and can transfer charge fast enough to keep up with the needs! Most active balancing technologies are far from "up to the task" especially with high charge and discharge rates and very "flat" capacity curves. With LIFEPO4 most all the charge transfer has to take place in the nearly full and nearly empty ends of the cycle between 3.3 volts and 3.0 volts very little balancing can be done.

 

[CKidder]

It appears that Tesla agrees with you to some extent. The Model 3 has 1k balancing resistors and no external FETs. It just bleeds about 4ma of current from each group of parallel cells to balance. Quite obviously it would take a long time to do much of anything to 230AH worth of cells at a 4ma discharge rate. But, as you said, you've got time. If you only need to balance over the course of, say, a month then 4ma is enough balancing current. That is, so long as your cells are reasonably balanced in the first place.

 

[arber333]

I can say first hand now after a year of using LiPo cells with only crude balancers at 4.1V per cell. I only charge up to 4V per cell, so balancers dont even turn on. I was very rear that i got so high that BMS started to drain charge.

1. Cells stay perfectly lined together.
2. EOC is always caused by the lowest SOC cell that trips the high voltage alert at 4.05V.
3. On empty that cell allways trips low voltage alert at 3V.

There were other observations, such as at both sides of SOC internal resistance suddenly changes. It is only then you are effective with balancing. Theoreticaly you would have to continuously measure IR of the cell and balance only when this rises. I know REC BMS does this, but i never tried since i dont go for the top. I find last 0.1V gives only some 6km additional range and vs cell life i dont even consider it.

So there is no way around this. Glass is only so big and you cant fill it with more. If one cell has less volume it cant store more, or will store more but at higher tension, which will slowly erode its capacity to retain this same charge. This defeats the purpose of BMS to protect cells, not to extend range at any cost.

BMS algorithm may be complex and self learning, but we learned OTJ. Cant force more charge into battery with lower capacity.

 

[jhuebner]

Jerry, the reason for “antagonism“ is that people with little knowledge come to this forum. I don't want them to go home thinking “I have to do active balancing/need high balancing current otherwise I waste money/range/...“. Years ago I was the noob and all those papers by IC manufacturers made me think I had to do active balancing. I respect your work as science or as a piece of art but it is certainly not economical nor DIY friendly. You can proof that wrong by showing that the price difference between active balancing and 4mA passive balancing is less than the extra cost of adding 3% more capacity to your pack. And you'd have to show that you can shift the required energy from all above average cells to all below average cells during discharge time (say 2-3 hours in an EV).
Edit: maybe I should add that I'm not specifically attacking you or you design but any design that puts inappropriate effort into balancing.

I guess I have a different understanding of “holy grail“. To me it is the simplest, cheapest, satisfying solution.

So Tesla uses 1k bleed resistors thanks for the info, thats amazing!

 

[Jerry Liebler]

Jerry, the reason for “antagonism“ is that people with little knowledge come to this forum. I don't want them to go home thinking “I have to do active balancing/need high balancing current otherwise I waste money/range/...“. Years ago I was the noob and all those papers by IC manufacturers made me think I had to do active balancing. I respect your work as science or as a piece of art but it is certainly not economical nor DIY friendly. You can proof that wrong by showing that the price difference between active balancing and 4mA passive balancing is less than the extra cost of adding 3% more capacity to your pack. And you'd have to show that you can shift the required energy from all above average cells to all below average cells during discharge time (say 2-3 hours in an EV).
Edit: maybe I should add that I'm not specifically attacking you or you design but any design that puts inappropriate effort into balancing.

I guess I have a different understanding of “holy grail“. To me it is the simplest, cheapest, satisfying solution.

So Tesla uses 1k bleed resistors thanks for the info, thats amazing!

YES, the IC manufacturers have expensive and far from "up to the task" active balancers. Complex systems with abundant software and computer power "featuring" balancing currents in milliamps & not many of them to.
My thought is that's simply not worth while!

What balance current is needed to be "up to the task" in any given situation is not something well researched.

What I have designed may not be "up to the task" in the EV world of 1 hour charge and discharge cycles with 100 AH batteries. But for my use on a solar energy storage battery which has day long charge and all night discharge cycles I expect that even 6 amps will make a substantial difference even on a 400 AH pack. BTW the 6 amps is a result of what is available at low cost (I'm building a minimalistic BMS sort of along the lines of the Ckean Power Auto mini BMS but adding the feature of being able to shut it off and reduce parasitic losses to microwatts or less and active balancing with 6 amp currents. The added parts to do the added features include a transformer, a controller IC and a power FET which together cost less than $5 in modest quantities. and I save about $1.50 of optoisolator. My 4 cell module could easily sell for under $60.

 

[sholland]

It appears that Tesla agrees with you to some extent. The Model 3 has 1k balancing resistors and no external FETs. It just bleeds about 4ma of current from each group of parallel cells to balance. Quite obviously it would take a long time to do much of anything to 230AH worth of cells at a 4ma discharge rate. But, as you said, you've got time. If you only need to balance over the course of, say, a month then 4ma is enough balancing current. That is, so long as your cells are reasonably balanced in the first place.

The Model 3 has 37.4 ohm balancing resistors. I think you are confusing them with the 1k input sense RC filter resistors.

 

[CKidder]

The Model 3 has 37.4 ohm balancing resistors. I think you are confusing them with the 1k input sense RC filter resistors.

Indeed, I had taken a picture of that part of the board and it wasn't very clear because I was lifting the BMS board up while it was still connected to the wires on the other side. I thought that the 37.4 resistors said "4703" on them. Yes, that's a bit far off but my picture was really bad/blurry. I could clearly see the "103" marking on the other resistors. So I jumped to the conclusion that the "47k" resistors were for measurement and the 1k resistors were for balancing. That's wrong. In reality the 1k resistors are for measurement and the 37.4 resistors are for balancing. Oddly enough, not all of them are 37.4. It looks like some are 47 ohms and some are 37.4 but the 37.4 also run through a resistor that isn't marked so I think that resistor is probably around 10 ohms (didn't measure) and thus all balancing is really around 47 ohms.

At any rate, yes, the Model 3 does use much smaller resistance for balancing.

 

[jhuebner]

So 80 mA balancing current. Still much less than the typical 1-2A currents of charge-end balancers. Thanks for looking into it and providing a relevant practical example.


Speaking of cost: my next design will have a BOM cost of 4€ for a 6 cell module with 50mA balancing current. About 15€ manufactured in small quantities in German fab.

 

[sholland]

Yes, correction, 37.4 + 10 ohm.

 

[PStechPaul]

I posted a more detailed reply in the other thread, but my research and preliminary design experience has brought me to the conclusion that a simple measuring BMS for as many as 16 cells per module is probably most cost effective and practical. Low current shunt bottom balancing can be added for little extra cost. And for charging, unless you absolutely must have top balancing, you can just shut down or throttle back charging when one cell reaches its maximum voltage point.

 

[steveob]

I fail to see a reason for all the antagonism....You take a very strong stand against active balancing yet you acknowledge a 3% variation in cell amp hour capacity. Without active balancing the capacity of the pack (amp hours) is that of the weakest cell and no more.

The reason is as I and others have explained repeatedly, is that you NEVER go anywhere near %100 unless you are uneducated, doesn't matter if it is a motorcycle or a ship or whatever. Even OEMs probably have a safety factor when they tell the user it is %100 charged. You are killing your batteries if you do (well that and heat), you seem ironically unaware of how analog batteries are and all the caveats involved in managing them, and have the classic solution in search of a problem syndrome.

http://www.diyelectriccar.com/forums/showthread.php/interesting-thesis-paper-lifepo4-cycle-life-86517.html

With active balancing that is 100% efficient

Lol, have you ever made a circuit before?!?

 

[Tony Bogs]

Don't forget statistics. The presumption of all cell parameters within a certain percentage of each other is absolutely false !!!!


But I guess that the TS is suggesting that it is all about finding the sweetspot between porobability of a catastrophic failure,
the total cost of the system and the statistical reliabity data of a BMS when the complexity increases.


And where is that sweetspot? Well, that all depends on the set criteria (for instance the accpetable chance of a battery fire) and factors like quality of the build,
the number of parts etc. Great stuff for experts in applied mathematics.



I'd like the cnance of a battery fire to be less than the chance of winning the jackpot in the National Lottery.