LiFePO4... Why Use A BMS?

[bobmcree]

the idea of just putting a capacitor across each cell might work, if you haul a trailer full of supercapacitors. 1 farad will deliver 1A for 1 second, so if your controller draws 20 amps that is 50 milliseconds from a bank of capacitors about the size of a 12 pack. If you were using occasional bursts of milliseconds of high power like drum beats on a car stereo a huge capacitor can save wear on the battery plates by spreading out the load, but when you have a constant load the caps will do nothing for you.

there are definitely lots of the 3.3v converters on ebay and the surplus market, for the do-it-yourself guy who wants a single cell charging solution. most of these switching regulators are not isolated and not all have adjustable voltage, so one must choose carefully. 425 of the power-one units i used are still available for $10 at http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&rd=1&item=150201797488&ssPageName=STRK:MEWN:IT&ih=005 These are 20A supplies that have sense and trim and enable inputs that can be used to easily convert them to single cell chargers that will take a supply from 36-48v and turn it into an n cell charger with independently adjustable channels for each cell. Smaller modules are available for use at lower current, or the sense inputs of these units could be used to limit the current to a lower value with some added circuitry. You still need a big DC supply to power the converters but they will accept a wide input range.

putting a shunt across each cell during the charge cycle prevents overcharge, and if the charger cuts off at the proper voltage this should charge every cell up to 100% every charge cycle. Without the shunts some cells will be overcharged and some will not reach full voltage before the charger cuts off. i am sure there is a lot more to it, but that seems like a good place to start.

a comment was made that you cannot just charge these cells to 3.6v indefinitely because they might produce a lower voltage as they age? i have certainly not seen that myself or in anything i have read. my experience is that the charge current will drop to a very low value when the cells reach 3.6v and eventually the current will drop to almost zero. i would not rely on that, but if any cell cannot be brought up to 3.6v by the balancing circuit in the expected time i would want to know about it and get an error signal. I have never seen a cell that could not reach the full charge voltage and i have not seen anything indicating that is the case. In fact i have found you can charge these cells up to 4v and higher with no apparent short term damage. If the claim is that charging older cells to the same voltage as new cells is a problem, that is another statement i would like to see verified. Anyone who has information on this effect i would love to see it. there are very few "authorities" on LiFePo4 batteries, and i don't claim to be one. i am just one of the guys trying to get better batteries for my bikes and a few friends.

 

[Deepkimchi]

Yeah - when Safe gets started - why do you think they call it "Endless Sphere" ?? :lol:

 

[safe]

the idea of just putting a capacitor across each cell might work, if you haul a trailer full of supercapacitors. 1 farad will deliver 1A for 1 second, so if your controller draws 20 amps that is 50 milliseconds from a bank of capacitors about the size of a 12 pack. If you were using occasional bursts of milliseconds of high power like drum beats on a car stereo a huge capacitor can save wear on the battery plates by spreading out the load, but when you have a constant load the caps will do nothing for you.

Part One

I'm going to respond to your post with two posts because they are two totally unrelated ideas.

Do you see how PWM creates a non-constant load because it only really extracts from the battery in "pulses". Seems to me that all you need as far as capacitors is enough energy to fill a single "pulse". As I recall a typical PWM controller runs at about 20 kHz (20,000 pulses per second) so the capacitor needs to hold enough energy for:

40A / 20,000 Hz = 0.002 Farad (per pulse)

...which doesn't sound like it needs to be that big. I just imagine this little tiny capacitance that is used so rapidly because of the PWM controller that it's throughput ends up being pretty good. Sort of like a really, really, really high rpm motor, each stroke isn't very powerful, but collectively it becomes a lot of power. A small capacitor array ends up being able to deliver the entire current in small pulses.

Is my sense of scale way off?

Is 0.002 Farad really going to require a lot of capacitance?

Maybe think of it this way...

48 volts * 40 amps = 1920 watts = 1920 Joules per second

Divide that by the number of pulse cycles per second:

1920 joules per second / 20,000 pulses per second = 0.096 joules per pulse

The really cool thing about this idea is that you could modularize battery packs and be able to assemble them like lego blocks. No more worrying about cross cell connections or intricate relationships with the controller to get it to shut off early... you could always be sure that there are no "bugs" in the system as the programmers say. I'm just a big fan of solving the problem forever rather than continually patching the problems like things are done now. (it seems kind of ad hoc the way people are approaching this now... not well thought out in the "big picture" sense)

But what do I know... I'm just Colombo... :lol:

 

[safe]

a comment was made that you cannot just charge these cells to 3.6v indefinitely because they might produce a lower voltage as they age? i have certainly not seen that myself or in anything i have read. my experience is that the charge current will drop to a very low value when the cells reach 3.6v and eventually the current will drop to almost zero. i would not rely on that, but if any cell cannot be brought up to 3.6v by the balancing circuit in the expected time i would want to know about it and get an error signal. I have never seen a cell that could not reach the full charge voltage and i have not seen anything indicating that is the case. In fact i have found you can charge these cells up to 4v and higher with no apparent short term damage. If the claim is that charging older cells to the same voltage as new cells is a problem, that is another statement i would like to see verified.

Part Two

What I recall hearing is that you want to charge these cells up to 3.6v and then ideally stop and charge no more. All time spent at or above 3.6v reduces cycle life and going to 4.0v really drastically reduces cell life. It was in regard to A123 cells that this drastic damage was reported, so it might not apply to the LifeBatt 40138's, however I wouldn't get too comfortable with that idea.

Seems to me that the moment the cell reaches it's 3.6v level it's demonstrating "stress" because of it being full and so it does make sense to discontinue charging completely when this happens. Balancing systems that simply reduce the extent of the overcharge certainly help, but are not doing what should ideally be done.

 

[safe]

Increased Safety?

Another possible benefit of using the "Buffered Approach" to pack design (using capacitors) would be that the high voltage could be consolidated to within the capacitor array alone. This might mean that people could run 200 volts worth of cells and the only place where that voltage is concentrated could be placed within a safe container separate from the cells. Each cell only contributes it's 3.3 volt energy and the series capacitors build the voltage up to 200 volts. It's still going to require safety measures to be done, but it makes the idea of 200 volts seem a little more realistic. Plus the wires that come from each cell don't need to be so thick because they don't carry the combined voltage. Only the wire going from the capacitor array to the controller would need to be big.

Of course the next step (and you knew this had to be coming ) would be to combine the functionality of the controller into the capacitor array and make them one and the same thing. So in this "last stage" product you would have nothing but a controller and a bunch of thin wires (3.3v) going out to the cells. There would be nothing left to fear but the controller to motor connection... the last potential place to get electrocuted.

Using some rather simple logic it should not take any more capacitors to build this array than it does right now to build a controller... they both need to store and pass the same current.

Does that make sense?

I know that the controllers use FET's which are more like gates than capacitors, but I think they also use capacitors anyway. This is an area where I'm simply guessing based on intuition... I just don't know.

 

[bobmcree]

Increased Safety?

Using some rather simple logic it should not take any more capacitors to build this array than it does right now to build a controller... they both need to store and pass the same current.

Does that make sense?

no it does not. the capacitors in the controller are intended to keep the battery voltage constant so it will not jump around with the switching pulses due to the inductance of the battery wires. if the wires to the battery were very short these capacitors would not be needed. they absolutely do not ever contain all the energy that is passed from the battery to the motor. most of that energy comes from the battery directly to the fets and is then switched into the inductive windings of the motor. the capacitors in the controller should merely be considered an extension of the battery. the controller capacitors typically add up to 1000 uf altogether, so they can store enough energy to produce 1A for 1 millisecond. this is enough energy to stabilize the voltage so the fets switch properly but not enough to run the system.

I know that the controllers use FET's which are more like gates than capacitors, but I think they also use capacitors anyway. This is an area where I'm simply guessing based on intuition... I just don't know.

you cannot intuit an understanding of switching power supplies. get a book and read it and you will see how totally ridiculous your ideas sound. there are switched capacitor systems that can transfer energy like a bucket brigade to keep cells balanced the way a Powercheq does for lead acid batteries, but the ones I have seen only work at a small fraction of an amp, and they work by dragging the highest voltage cells down to the voltage of the rest. I suggest you may need to adjust your ratio of output to input in the hopes of adding some value to the output. try listening for awhile instead of spouting this kind of nonsense, and you might learn something.

 

[Toorbough ULL-Zeveigh]

there are switched capacitor systems that can transfer energy like a bucket brigade to keep cells balanced the way a Powercheq does for lead acid batteries, but the ones I have seen only work at a small fraction of an amp,

Interesting, I was wondering why you don't see any charge pump switchers.
The capacitor based ying to the inductor based yang if you will.

 

[safe]

...the capacitors in the controller should merely be considered an extension of the battery. the controller capacitors typically add up to 1000 uf altogether, so they can store enough energy to produce 1A for 1 millisecond. this is enough energy to stabilize the voltage so the fets switch properly but not enough to run the system.

My "Colombo Routine" seems to have worked. :wink:

You've answered with proof that controllers do in fact rely on capacitors to hold some charge between pulses. As it's done now there's no interest in holding the full charge, but in this day and age of Ultracapacitors the ability to hold a "buffer" is being taken very seriously. To suggest that the idea of a "buffer" is out of line doesn't seem to match all the other stuff going on right now.

I presented that last concept (the integrated buffer, controller and BMS) as some final stage where the control logic of the battery is integrated into a more sophisticated system. I'm not saying that we need to get there tomorrow but that someday down the road it's this kind of thinking that would likely be implemented.

Buffering and the BMS COULD be done at the same time... it would make a good merger. (and it allows the battery cells to be "dumb" rather than having to be "smart")

However, for the here and now we will probably muddle through with the usual slapped together solutions. (it's always like this) Things will happen slowly.

 

[Jeremy Harris]

Bob,

Those eBay 3.3V DC-DC converters look like a bargain, but checking the spec sheet I've found that they have a minimum current output of 10% of rated output, which means that they have to deliver at least 2A all the time. I guess a shunt could be used to take the excess, but that seems a bit wasteful.

Had you got any bright ideas for getting around this snag?

Thanks in anticipation.

Jeremy

 

[safe]

To BMS or Not to BMS - Part One

Let's say you decide to get in on the "Group Purchase" idea through Patrick and go ahead and buy some 40138 cells at a discounted price. (like $30) Without the BMS the only way to get long life from your 40138's is to watch them carefully and rotate out the weak one's when they appear to be falling behind. This demands that you buy an extra one or two 40138 cells in order to be able to swap them. If you do your swaps really well and keep the overall system balanced over time you should do as well (or close to as well) as the BMS in your result.

The question becomes one of economics and effort.

In the big picture is it more work to implement and then trust that a BMS can actually achieve better results than careful monitoring of the cells yourself?

The 40138's are big and they come with bolts that make them very easy to remove and swap out. This is even easier than dealing with the smaller SubC and D cells because you have far fewer things to keep track of. In a small system there might be as few as 15-20 cells. So all you need to do is buy a couple spares and rotate.

Tough call? Easy Call? What would you choose?

 

[safe]

To BMS or Not to BMS - Part Two

Here's another thought that might be junk, but here it goes...

In a rotation scheme there is the possibility that eventually the spare cells become so bad that they fall off the scale and can't be brought back into the rotation.

Is it possible at this point to take the two weakest cells and make them in parallel so that they are now understressed? (the rest remain in their series configuration)

That forces the next in line to carry the load and will gradually wear it out enough so that it too starts falling off the bottom. After a while you could get some rather creative configurations based entirely on the idea of deflecting the stress off of the weaker cells and forcing it back onto the stronger ones.

 

[safe]

To BMS or Not to BMS - Part Three

Okay this is even stranger...

I LIKE having manual control over every aspect of my bike. I would like the idea of being able to TUNE my cells manually and the way I want to do this is to use the shunt idea (or something like it) that has been mentioned elsewhere.

The idea would go like this...

Each cell would have a corresponding bypass shunt that would redirect a certain amount of energy past the cell. When the cell charges a certain amount is diverted depending on the setting, so too when discharging would a certain amount of the energy that it might normally be forced to produce now be lessoned based on this setting.

So imagine a trim pot for each cell.

You keep constant track of how your cells are behaving and when one needs some help you turn the trim pot so that it gets the help it needs. The really strong cells are left so that they get no bypass, the weak ones are given a lot of bypass.

8) Cool idea... or has insanity taken hold?

 

[Deepkimchi]

Here - try this

http://www.healthyminds.org/expertopinion2.cfm

 

[safe]

Just Take A Look... I might come off like Colombo, but I do "get there" more often than not. This might allow use of some very low cost components that are more durable. With no need for control logic (since it's manual) it simplifies things a lot.

Maybe this is all you really need?

 

[TylerDurden]

I LIKE having manual control over every aspect of my bike.

You could make special hollow spokes and hubs to monitor and control your tire pressure...

...altering the pitch of your bike by sucking snd blowing.

 

[safe]

Throw in a few Resistors?

Seems to me a resistor on either side that was of high enough resistance to reduce the current (which can be rather large) would allow you to use a lower priced pot control. An issue to consider would be power losses, so you would want to do everything possible to calculate how much of a bypass you need and not to use any more than you need because more bypass could translate into power losses. (I would guess since resistance means losses)

That pot control costs 50 cents each.

Also, from this shunt circuit you could even have your measurement of the individual cell voltage. You could literally monitor each cell right down to it's voltage... now that's extreme control of your cells.

You could even watch the cells as they charge and when one reaches it's maximum you turn down it's pot setting a little. The next time you charge you might adjust it some more. You could really fine tune your pack. It would give you something to do while it was charging...

 

[EMF]

To BMS or Not to BMS - Part Three

8) Cool idea... or has insanity taken hold?

The bad news is insanity has taken hold. The good news is since you asked...your not too far gone yet. However, I am not a mental health professional, I just play one on the Internets. :wink:

 

[safe]

The bad news is insanity has taken hold.

Obviously I said that as a joke.

Seriously... why WOULDN'T this work?

Maybe the shunt needs to be configured differently?

Each cell should have an internal resistance that is something like 50 milliohms so ideally you would want to balance your bypass circuit with this resistance value. You would want to set things up so that whatever needs to pass through the shunt would pass and it's range would be within expected parameters for that cell.

When charging the extra current would reroute to avoid the cell because the resistance would be low enough to extract a small amount away.

When discharging the bypass would lower the amount of contribution by the cell because the current that went through the bypass would act to reduce the voltage difference. With a lowered voltage difference the cell would not pull as hard.

Anyone?

 

[safe]

Either I've made some kind of conceptual mistake or people are getting that sinking feeling like:

"Oh shit... safe has another good idea... I HATE that!"

 

[Link]

So, effectively, you're balancing the cells by throwing energy out the window?

I'm not even going to try to figure out what happens to cells when you touch every terminal to every other terminal through a resistor. Make that lots of resistors. All with different values. That change.

 

[Ypedal]

Not quite.. more like:

... ...or... ....

"Oh shit... safe has another... idea... I HATE that!"

 

[safe]

So, effectively, you're balancing the cells by throwing energy out the window?

There would be some mild losses due to resistance, but those would be kept as low as possible. A better way to look at it is each cell is given a certain "utilization level". So let's look at an example of a 10 cell 40138 pack. The control pots "might" be set with the following settings:

Cell 01 - 75%
Cell 02 - 100%
Cell 03 - 100%
Cell 04 - 100%
Cell 05 - 90%
Cell 06 - 100%
Cell 07 - 80%
Cell 08 - 100%
Cell 09 - 100%
Cell 10 - 100%

...so cells 01, 05 and 07 would have a certain amount of bypass in effect (25%, 10%, 20%) and this means that those cells would be ASKED to do less in both the charge and discharge situation.

This is a MANUAL way of restoring civilization in the LiFePO4 world and it could be done while you are sitting around and waiting for the cells to charge. You really wouldn't even need to have voltage meters on the bike because you could check this stuff with your multimeter while it was charging.

It would give you something to do while you wait...

 

[Link]

That would get really boring, really fast.

What about Zeners? Instead of pots, stick two Zeners oppositely biased on each cell. Once the cell reaches a certain voltage, the Zeners start conducting and dump off all the excess power.

Ignore me if this sounds like nonsense. I can design a circuit, no problem, but things get start to get a little iffy when voltage is applied.

Oh, yeah. This reminds me to ask something: Does anybody know where I can get some electronics SIMULATION software? The only things I can find are like $3000 (eep!).

 

[safe]

What about Zeners? Instead of pots, stick two Zeners oppositely biased on each cell. Once the cell reaches a certain voltage, the Zeners start conducting and dump off all the excess power.

Apparently you haven't been following all the subtle problems that have been cropping up with automated BMS systems. Let me review...

One of the big problems with LiFePO4 is that when people try to do things like set a cap voltage this still allows that cap voltage to grind away at a full cell for a longer period of time that it ideally should endure. You are in a sense "beating up the runt kid" when you do this and in the cruel, cruel world of LiFePO4 that runt dies pretty fast.

What I'm suggesting is that the "runt kid" is given an easier path both on charging and discharging. So during the charge this "runt kid" is given less to handle THE WHOLE TIME and so he arrives full at the same time as the "big strong jocks" that are pulling the big energy loads. During discharge the "runt kid" doesn't have to work as hard so he doesn't tire himself out as early compared to the big boys.

Using the "equalizer approach" means that everyone works as hard as they can, but they all don't work at the same level. They all win their "personal best" awards and all the politically correct liberals cheer that the LiFePO4 world is now "fair".

Make sense?

Our human brains can analyze what's going on pretty easily. We can figure out which cells are stronger and which are weaker and we can set the controls to adapt to this situation really easily. To get that level of "abstract thinking" into the BMS circuit (you need to in effect "learn" the behavior of the cells) is not easy to do.

 

[safe]

Struggling to Comprehend the Math

From what I've found so far a typical LiFePO4 cell is going to have an internal resistance somewhere in the 20 milliohm range. This means that the internal resistance is 0.02 ohms.

Using the idea of a current divider (a shunt) if we want a small amount of current to be bypassed to the next cell we need to create a resistance shunt that will give us the appropriate amount of resistance to get the current we desire.

http://www.ee.ucl.ac.uk/~amoss/java/cdiv.htm

...aw heck, I've got one eye on the tv (election results) and am trying to do this at the same time. Enough for now, I'll do this later. Anyone who wants to jump in and solve the math on this go ahead.