Even Newer 4 to 24-cell Battery Management System (BMS)

[ejonesss]

i am glad you are trying to perfect the board so you then will not have to make another board or have to have us then dead bug additional updates to the board.

The idea is by having the LED light up for LVC is you can identify a weak or bad cell without having to measure each cell's voltage. For those that don't want to tie the LVC into the throttle, it would be quite easy to wire up a 12v piezo alarm that could be mounted remotely (where you can hear it).

This whole project has me pulling my hair out. Individually, the sections of the circuit behave, but when tied together, all kinds of unexpected things happen. We are zeroing in on the remaining issues, however, and I am confident that everything can be sorted out eventually. It's just taking a lot longer than anyone wants.

 

[AndyH]

Bwahahahaha...bring on the V4 thread!

 

[AbuRastas]

Hi to every one... I'm reading you for some time but, now, I need to make a charger... I'm beginning with the 2.6c version and ... there are some questions about it....

For example in the Lm431 the resistor divider that activate the shunts are 120K and 330K. if the voltage in the activation instant is 2.5V... gives 2.5V/330K= 7.5ma. when this current go through the whole divider gives 7.5ma * 450K= 3.4V...

Until I know, LifePO batteries need 3.6v to balance it so 3.4 is perhaps a little value.

And about the currents on opto leds. if the leds subtract 1v in the lvc 2.1-1= 1.1V /1K = 1.1ma that is perhaps a little value since it need 5ma to saturate and turn on the opto....

is this ok?

Anyway i'll try to wait for the V4 version but... how work the new version? have inductors with some kind of buck converter? how control now the PWM...

Well I'll wait for your responses. and thanks for the big help that you putting here. bye

 

[methods]

You guys should post up the latest schematics while I still have time to steal any clever ideas you had and work them into my BMS

-methods

 

[grindz145]

Thems fightin' words.

 

[GGoodrum]

You guys should post up the latest schematics while I still have time to steal any clever ideas you had and work them into my BMS

-methods

We are finding out that this really is a black art, of sorts, and there really are all kinds of weird things going on, "under the hood", so to speak. Richard has been "'scoping" things out all week, and I've been testing pretty much 17-18 hours a day. It has been an exercise not unlike trying to gather in a bunch of balloons. Just as you think you you've got them all, one sneaks out. We've had hysteresis problems, high frequency oscillation problems, shunt variances, capacitor leakage issues, which makes them act as resistors as well as caps, as well as other problems that were "hidden" in the overall performance.

I think we are very close, as we now seem to have all control section HF noise, hysteresis and throttling issues sorted out, but there is one balloon left to gather in, and that is related to the shunts themselves. It affects how close the cells end up reaching the same voltage level.

In any case, I don't think just slapping a big FET onto the LT pins is going to cut it. Digital or analog, the issues are going to be similar, and on the measurement end, what the LT chip sees may not be as accurate as you think. You do have the advantage in the digital realm that you can program in a sophisticated calibration and 2nd-order filtering schemes, but that won't be so easy either.

Two weeks ago, I didn't even know what I didn't know, if that makes sense. In any case, I'm hopeful that today if I can't gather in that last #@%$ing balloon, I can at least pop it. :wink:

-- Gary

 

[methods]

Thems fightin' words.

Yea - but at least I am honest :wink:

In any case, I don't think just slapping a big FET onto the LT pins is going to cut it. Digital or analog, the issues are going to be similar, and on the measurement end, what the LT chip sees may not be as accurate as you think.

You could not be more wrong Gary. I would rather not discuss it in your thread but if you are going to call out my design I will answer with a rebuttal. The measurement coming off of the LTC6802 is 12 bits and dead nuts accurate. It has a digital brick wall filter as well as a front end RC filter. The measurements are within mV even in noisy environments and they can be relied upon 100%. The chip is *designed* to "slap on a big fet" and that is exactly what I have done. I use the time domain to avoid taking readings while discharging so wiring losses and interactions are not an issue. I dont have to deal with any of the analog issues you are describing above - every action and measurement is timed and controlled. If something is screwy - I can just adjust my timing in software. Really - all I have done is copied the Linear Technology demo board design which is on its 4th iteration and not only has the bugs worked out (by hundreds of customers giving feedback) but it also has protections in place (ESD, OverVoltage, etc). There are certainly aspects of my design that I am concerned about, but the accuracy of the voltage detection and the ability to drive shunts reliably is not one of them.

You do have the advantage in the digital realm that you can program in a sophisticated calibration and 2nd-order filtering schemes, but that won't be so easy either.

All done in the chip. Like I said - there is a digital brick wall filter backed up by an RC filter on the front end. Nobody does A/D better than Linear Technology. They have been on the cutting edge for years. I interviewed there (and received an offer) 10 years ago and they already had a working 14bit A/D running (that was a big deal back then). If you follow the design spec on LTC A/D's you get sold performance.

When the rubber hits the road what you really need is accurate data. You need to know exactly what the voltage is (because that is what we are trying to do right?). Simple problem - prevent the cells from going over voltage while maintaining the maximum charge rate.

None of that really matters though right? Just talk.
What matters is who has a well tested product in hand to sell to the needy - and that is you

-methods

 

[__Tango]

...
The CellLog units connect into a PCB that has an opto-coupled buffer for the alarm outputs of each CellLog unit. This keeps each of the alarm outputs isolated, which they need to be as each is referenced to the cell 1 ground connection for its pack.The opto outputs are all ganged together, like they are in the BMS, and sent to the control board, which is identical to the latest ones used with the full BMS. The CellLogs have been programmed to have the "HVC" alarm point at 4.15V. The control board uses this signal to make sure no cell can go over this point, using the same PWM-based "throttling" circuit. The shunt circuits are completely separate, in their own box, and only get plugged in when the balancing is actually required. The box shown, when completed, will hold four 6s shunt boards, which are also set to be fully on at 4.15V. The box has four fans, two at each end.

Right now, like I said, this is a parallel, lower priority effort, but one that will be great for those already using, or contemplating using, the nifty little CellLog units. I'm not sure how these will be packaged, but I may end up combining the control board onto the interface/opto board the CellLog units are attached to, so that only one board is required. Also, since these units can do 8s, I'm going to do the connections in a way that two CellLog units can be used with 12s or 16s, and three can be used with either 18s or 24s, using jumper blocks. I need the latter for my own setups, which are either 18s or 24s.

Oooooooh. I'm very interested in this. I'm doing a 36s LiFePO pack for my motorcycle (yeah, i know it's a lot), and have been patiently waiting for the v4 boards to be completed and built by Andy. "In the mean time", i've gone off and purchased 5 CellLogs 'cuz i thought it'd be cool to be able to do some manual monitoring.

But this is great! The best of both worlds!

So with it, you'll be able to do high and low cutoff, right, and the shunts will be separate. Very interesting.

Please keep us posted on the developments to this setup.

Thanks!

 

[DaveAK]

I really hate to post this question, but any news on availability? This is going to be the last piece to my project and I need to decide whether to build it with my limited skills or buy a pre-built one.

 

[alangsam1@me.com]

seems like they are trying to fix some difficult bugs in their close to final design. until they fix them and have a run of the boards that are defect free i dont think they know. their are a lot of us on this board that have projects that need this yesterday.

 

[DaveAK]

seems like they are trying to fix some difficult bugs in their close to final design. until they fix them and have a run of the boards that are defect free i dont think they know. their are a lot of us on this board that have projects that need this yesterday.

I appreciate that. Just looking for an update. I don't need mine yesterday though, tomorrow will do fine.

 

[fechter]

Unfortunately, it seems like as soon as one bug gets squished, two more pop up.

This whole project has Gary and I pulling our hair out.
Anyway, we are still making 'progress', but I won't declare success until everything passes testing in a fully assembled system.

There have been several weird problems along the way that I just never anticipated, and a few that I should have anticipated but didn't. The ver. 2.x boards could be tested without cells attached, for example, but since ver.4 shares the HVC and LVC bus, you cannot test without at least having some big capacitors across the cell circuits or else the LVC triggers during the throttling PWM and screws everything up.

Another bizzare behavior was when attempting to test more than one cell circuit in series, they could go into an oscillation where one cell would shunt fully and it's neighbor would go open and oscillate back and forth like that. This never happened with the ver.2 stuff and it took a while to figure out what was going on.

Individually, the 'building blocks' of the system work, they just don't play nice when connected together.
I'll have to try behavior modification with high voltage next. That usually works

 

[DaveAK]

I certainly don't envy you or Gary having to sort this stuff out! And of course we all appreciate your efforts.

Gary had earlier mentioned that he might release a version that would work with the larger fans. Is this now not the case because of the other bugs that have popped up?

 

[alangsam1@me.com]

Gary, any updates?

 

[GGoodrum]

Gary, any updates?

Yes, we at least know what is causing the weirdness I've been seeing all week. The shunt transistors are getting overloaded/swamped, when the current gets close to 1A. This is because we've always used 3.0V across the shunt resistors as the set point for the throttling, which puts as much of the heat generated as possible through the resistors, instead of the power transistors. This works fine, under an amp, but as we got close to this, the voltage drop through the power transistors (Vce) we were using was too much. Before (i.e. -- v2.x...) we never saw this as a problem, for two reasons, the BD136s have about half the Vce drop of the TIP105s TO220-based parts, and because we never really did 1A shunts with these boards.

Now, with a more heat-friendly layout, and TO-220-based power transistors, it is okay for more of the heat to be "shared" between the transistor and the shunt resistor(s), so we are now lowering the voltage across the shunt resistors, when the throttling kicks in, to about 2.5V. We've also picked a new power transistor (KSA473...) that has about the same Vce drop as the BD136s, which is about half what we see with the TIP105s.

Anyway, we are testing the fixes. so it shouldn't be too much longer.

 

[Indubitably]

Gary, any updates?

Yes, we at least know what is causing the weirdness I've been seeing all week. The shunt transistors are getting overloaded/swamped, when the current gets close to 1A. This is because we've always used 3.0V across the shunt resistors as the set point for the throttling, which puts as much of the heat generated as possible through the resistors, instead of the power transistors. This works fine, under an amp, but as we got close to this, the voltage drop through the power transistors (Vce) we were using was too much. Before (i.e. -- v2.x...) we never saw this as a problem, for two reasons, the BD136s have about half the Vce drop of the TIP105s TO220-based parts, and because we never really did 1A shunts with these boards.

Now, with a more heat-friendly layout, and TO-220-based power transistors, it is okay for more of the heat to be "shared" between the transistor and the shunt resistor(s), so we are now lowering the voltage across the shunt resistors, when the throttling kicks in, to about 2.5V. We've also picked a new power transistor (KSA473...) that has about the same Vce drop as the BD136s, which is about half what we see with the TIP105s.

Anyway, we are testing the fixes. so it shouldn't be too much longer.

Awesome, its a good thing the community has people like you guys around that refuse to rest until they get to the bottom of this sort of thing, even when it must obviously eat up god knows how much of your life. We're all pissy and impatient because we have decided to wait for your work to be finished in order to do our own projects, but what you do and the way you do it benefits us all far more than any one of us could possibly be inconvenienced. The fact that projects like this are floating around is exactly what will ultimately make widespread acceptance of EVs an inevitable reality.

No pressure or anything though, you know, its just that the fate of humanity as we know it rests on your shoulders. XD

At any rate, I just wanted to say thanks for all the hard work you guys do.

 

[amberwolf]

Yep...no pressure. ::watches pooka and fygar inflate and pop:: No pressure at all. :lol:

 

[fechter]

And even more weirdness happens in the thottling circuit. As the highest cell's shunt approaches it's maximum, the charging current gets reduced by the throttling circuit, which is much like a PWM motor controller, only without any inductor. The PWM ripple on the cell circuit interacts with the shunt and causes sort of a feedback oscillation. This same oscillation can cause problems in the charger supply as well. By increasing the PWM frequency, I can get most of these problems to go away.

Designing a complex circuit like this is much like playing a game of chess. Every move (change) has hundreds of possible consequences in other parts of the circuit and it is difficult to analyze all of them. Sometimes the only thing is to make a move and see what happens. Sometimes you kill a bug, sometimes you lose a piece (smoke).

The other thing that makes this so difficult is we are always trying to do this with the fewest number of parts possible, which ultimately translates to lower cost and hopefully less chance for Murphy's law to kick in. We are also trying to use the least expensive parts that will adequately perform the function. If we didn't care how much it would cost or how hard it was to build, we could have had something by now, but I don't do things like Microsoft. :wink:

We made so many changes from the previous version that the number of bugs was overwhelming, but we are stomping them as quickly as possible. I think the first models we roll out will not have all the features we have been working on simply to expedite the production. Eventually we will work the additional features in.

For now, I'm toast. Gary, Andy and I have been working double overtime on this thing for weeks, but I think this will all work out eventually and we'll have something worth waiting for.

 

[AndyH]

Will sleep help me get these songs out of my head? Or more coffee?

La Cucaracha

"who let the smoke out (woof woof woof woofwoof)"


we blow them up so you don't have to? ZZZZZzzzzz.....

 

[amberwolf]

Will sleep help me get these songs out of my head? Or more coffee?

Sleep is irrelevant, and coffee is pointless. :lol:

"who let the smoke out (woof woof woof woofwoof)"

I think that last part is "poof poof poof poofpoof".

 

[GGoodrum]

Actually, I think the weirdness is gone. I have a configuration now that works quite well. After we figured out the erratic behavior was caused by the shunt transistors being overloaded. the "fix" actually over-corrected the problem, and the shunt current ended up too low. After a couple rounds of tweaking today, I now have a setup that seems rock solid. I charged 3 different packs today, each with varying levels of balanced cells, and all three times, the packs ended up with the max delta between the lowest and highest cells of around 20mV. This is about the limit you can get with 1% parts, without hand-matching the values, which I did not do. Tomorrow, I will build up a couple more boards, to make sure everything is the same, and then I think I'm done.

We also have some controller card changes that will hopefully keep Andy from frying resistors, trying to put over 100V into the 12V regulator. He also found that the "Start" button didn't need to be a button at all. Just touching the pad where the switch would go was enough to get things going. Yet another resistor change has decreased the sensitivity, so the button actually has to be pressed now.

Finally, Richard is looking into using the current measuring logic we've had in some of the recent versions to implement an adjustable current limiter. This will allow virtually any supply to be used, including those, like many of the MeanWells, which either have current limits set too high (130%...), or just have a front-end "hiccup" mode. These inexpensive supplies will be able to be used unmodified (i.e. -- no hacks required...). The reason we are looking to add this option is because the controller already has most of what is required (PWM oscillator, gate driver, dual FETS, etc...), so all we need to add is a dual opamp chip and a shunt resistor.

More tomorrow...

 

[rscamp]

"I just want to tell you both good luck. We're all counting on you."

Airplane! 1980

 

[fechter]

"I just want to tell you both good luck. We're all counting on you."

Airplane! 1980

"I picked a bad week to quit sniffing glue"

 

[etard]

I think I heard Angels singing as I read that.

Finally a safe, inexpensive way to charge up in the time it takes to eat a burrito!

 

[MitchJi]

Hi,

Finally, Richard is looking into using the current measuring logic we've had in some of the recent versions to implement an adjustable current limiter. This will allow virtually any supply to be used, including those, like many of the MeanWells, which either have current limits set too high (130%...), or just have a front-end "hiccup" mode. These inexpensive supplies will be able to be used unmodified (i.e. -- no hacks required...). The reason we are looking to add this option is because the controller already has most of what is required (PWM oscillator, gate driver, dual FETS, etc...), so all we need to add is a dual opamp chip and a shunt resistor.

Great idea!