[dimitri]

All,

as I have been toying with the idea of simple and cheap BMS I realize that it would be best to make it an open source project. I'm pretty good at reading schematics and making electronics, but certainly not as smart and/or experienced as some folks on this board. I'd like this to be super technical thread with input from anyone who can contribute specific details, we can keep general discussion at the other threads.

I have researched various BMS types and designs for many months, I have contributed a little to Brian's Volt Blochers project, I made my own version of VBs, which is currently installed in my EV and works great, I have a couple of satisfied customers using my modules as well. Since VBs aren't available anymore and there seems to be a market for cheap, simple and reliable BMS ( see poll results in the New BMS poll thread ), I figured why not take the best ideas and try to implement it all in one universal product.

I especially liked the idea of simple BMS proposed by Johnny on Australian forum, so I have taken his schematic and applied few changes based on input from other BMS discussions. I also done some research on key components and their prices at Digikey, to make sure we keep total parts cost around $5 per cell.

In past few days I prototyped this schematic on a bread board using basic Radio Shack parts and it appears to work great. I really like how comparators allow precise and fast switching when voltage gets to trigger levels. It switched from on to off with 0.01V precision and very little hysteresis.

I'd like some input from pros out there, my goal is to minimize number of parts, improve anything that may need improvement, pick best resistor values and minimize different values so there is less different resistors to buy. For example, I think we can collapse R6 and R7 if we find LED with similar voltage drop as in optorelay, so they can share same current.

I'd like to use Clare CPC1014 optorelay for NC signal loop, since its the only suitable low on-resistance relay I found which can be purchased for 80 cents in reasonable quantity.

http://www.clare.com/home/pdfs.nsf/www/CPC1014N_R01.pdf/$file/CPC1014N_R01.pdf

Using 4 comparators may seem complicated at first, but they are all integrated in one IC chip which only costs 22 cents in quantities

I'd like to use these trimpots, which can be bought for 85 cents, yet very reliable for automotive environment.

Since there are so many needs for various HLVC levels, I think trimpots are pretty much a necessity, and they reduce number of resistors a little.

http://www.bourns.com/data/global/PDFs/3361.pdf

Later in this thread I will post initial schemaitic for head end controller, where I will also need some input from the pros.

According to the poll, many people still want shunting function in their BMS, despite recent ideas against top balancing. I'd like to keep this as an option on the PCB and simply skip these components if one doesn't want them. I'd like to keep shunting current low, like 0.5A, to avoid heat sink on the transistor and use 5 Watt resistor, which is reasonably small to fit on PCB.

I will be designing a modular PCB for this, which will allow a number of modules on one PCB, yet be able to cut it in pieces if one wants per cell installation. This will reduce PCB cost and assembly labor. However, for this and other reasons I abandon an idea of PCB under terminal bolt, so I will just have holes for either wires or rigid bars as in my current BMS modules.

I am planning on SMD components to minimize PCB size and further reduce costs.

Once we have working design I can produce them for DIY community at best price possible. This will probably take some time, so if someone needs BMS today, you might want to explore other options.

So, please, look at the schematic and question every component, every connection, everything...

 

[dimitri]

I think I found the answer for relay capabilities at low output voltage.

This is from Clare datasheet.

Seems that it can pass 200 mA at 0.12V , which would happen if 100 units in series are driving 12V head end controller.

Did I get it right?

 

[Tesseract]

A few quick observations for now:

* restrict the range over which the LVC and HVC can be adjustable with additional resistors. We know about where each should be so make the adjustment window maybe 0.4V or so.

* Bypass the pot wipers to both + and - so they are always at AC ground - important to prevent switching ripple from the controller tripping the circuit.

* Add hysteresis (small amount of positive feedback) to the comparators so they don't oscillate. Oscillating comparators are really annoying.

* it might be more costly in parts, but much cheaper in labor, to use multiple SMT transistors in parallel with thermal vias to a copper heatsink/ground plane. Try to avoid going 4-layer if at all possible.

I did not really look at the circuit, just giving a few quick suggestions before jetting off to the anodizing shop to find out if we will once again be able to offer colors.

 

[Dave Koller]

A few quick observations for now:

* restrict the range over which the LVC and HVC can be adjustable with additional resistors. We know about where each should be so make the adjustment window maybe 0.4V or so.

* Bypass the pot wipers to both + and - so they are always at AC ground - important to prevent switching ripple from the controller tripping the circuit.

* Add hysteresis (small amount of positive feedback) to the comparators so they don't oscillate. Oscillating comparators are really annoying.

* it might be more costly in parts, but much cheaper in labor, to use multiple SMT transistors in parallel with thermal vias to a copper heatsink/ground plane. Try to avoid going 4-layer if at all possible.

I did not really look at the circuit, just giving a few quick suggestions before jetting off to the anodizing shop to find out if we will once again be able to offer colors.

All great comments - I will put in my schematic that has built in time delays for low and high and pulses the normal on led in the opto.. But Dimitri has a simple circuit... I just put in the delay and pulses so there was a delay under load times to prevent false open of the NC and which way things were going ..

 

[Russco]

All,

So, please, look at the schematic and question every component, every connection, everything...

View attachment 4947

I have done considerable work on a lithium LV HV shunting BMS.

I feel using 1960 comparators with a series zener reference can be accomplished much more precisely with the common 431 precision IC. Two 431's would provide LV and HV monitoring. If the HV monitor is set higher than the shunting voltage, such at 4.2 volts and 3.70 volts, a third 431 would be required for the shunting monitor. The five ohm shunt resistor is a little high, only providing about 0.8 amps. Usually a 1-2 ohm shunt at 10 watts are used with the larger AH cells.

User adjustable pots are questionable. Can the user connect his pot across one or two cells for a reference voltage and make the pot adjustments? Many users can't. SMT eliminates easy bias resistor changes by the user or you to recalibrate users BMS units.

I have found in testing that each NPN opto output has a drop of around .15 volts. Fifty cells would work fine from the 12 volt accessory battery, but 100 cells won't.

Good work Dimitri, keep it up. You're a real go-getter.

Russ from Russco Charger. Currently testing 600 amp IGBT modular controller priced well below $1K.

 

[Dave Koller]

I have done considerable work on a lithium LV HV shunting BMS.

I feel using 1960 comparators with a series zener reference can be accomplished much more precisely with the common 431 precision IC. Two 431's would provide LV and HV monitoring. If the HV monitor is set higher than the shunting voltage, such at 4.2 volts and 3.70 volts, a third 431 would be required for the shunting monitor. The five ohm shunt resistor is a little high, only providing about 0.8 amps. Usually a 1-2 ohm shunt at 10 watts are used with the larger AH cells.

User adjustable pots are questionable. Can the user connect his pot across one or two cells for a reference voltage and make the pot adjustments? Many users can't. SMT eliminates easy bias resistor changes by the user or you to recalibrate users BMS units.

I have found in testing that each NPN opto output has a drop of around .15 volts. Fifty cells would work fine from the 12 volt accessory battery, but 100 cells won't.

Good work Dimitri, keep it up. You're a real go-getter.

See my above schematic (uses 431's) - Dimitri plans on setting up the pots for all the various conditions - I think a person needs a reference supply to set them so that is something to think about on a DIY board. but a simple multi-meter may work.. He will be using FET optos - no drop to speak of.
BUT I will let him comment on that..

 

[Tesseract]

I feel using 1960 comparators with a series zener reference can be accomplished much more precisely with the common 431 precision IC.

Soooo.... you see some merit in using the 1970's era TL431 over the 1960's era '339, huh?

Two 431's would provide LV and HV monitoring. If the HV monitor is set higher than the shunting voltage, such at 4.2 volts and 3.70 volts, a third 431 would be required for the shunting monitor.

I really can't see how using 3 separate TL431's is any better than a single bandgap reference feeding voltage dividers/comparators. Especially since it's easy to give comparators hysteresis so the switching action can be snappy, rather than fuzzy (the TL431? not so much).

The five ohm shunt resistor is a little high, only providing about 0.8 amps. Usually a 1-2 ohm shunt at 10 watts are used with the larger AH cells.

I definitely agree with this, though. If you are going to top balance 160Ah cells then you need to be able to shunt up to 2A. Whether you SHOULD top balance seems to be a matter of considerable debate, but that's for another thread...

User adjustable pots are questionable.

Yep... I hate trim pots, but it still might be good/necessary to have a limited adjustment range... Dunno. I'm also concerned about the temperature coefficient of the cell voltage.

SMT eliminates easy bias resistor changes by the user or you to recalibrate users BMS units.

Maybe, but I can tell you for certain that unless these boards are totally SMT technology and machine-assembled it simply won't be worth anyone's while to sell them at the price DIY'ers, etc., are willing to pay. I mean, you can maybe charge $20 per cell of which you'll have a good $7-$9 in parts and pc board cost. If it takes more than 10 minutes to assemble, test and calibrate each one then it's simply not profitable. If your labor cost is $10 per hour or less, and still fast and accurate, then sticking with PTH technology might still be okay, but even then why bother?

Really - this product begs to be machine-assembled: it is simple, but must be totally reliable. All of the problems people had with the Volt Blochers appears to be from poor soldering or mechanical injury (ie - the heatsink getting bumped so it shorts out an adjacent resistor). SMT vastly reduces the chances of such happening and with clever design you can make the pc board the heatsink (at least for a couple of watts).

Russ from Russco Charger. Currently testing 600 amp IGBT modular controller priced well below $1K.

Heh. You know how to make a small fortune in the controller business? Start with a large fortune and spend it developing and making something that undersells Kelly, who will then come back and undersell you because their parts costs, labor costs and even the Yuan/Dollar exchange rate are all in their overwhelming favor. But hey, good luck anyway!

 

[tomofreno]

Dimitri plans on setting up the pots for all the various conditions - I think a person needs a reference supply to set them...

The pot for adjustment of HVC on VBs is set using a simple resistive voltage divider, battery, and dvm. Same should work here.

Dimitri, I believe you said you want to preset everything to eliminate user adjustments/errors. If so, is it more difficult to just give several fixed resistive divider options in place of the adjusting the pots?

Please mount the quad comparator chip in a dip socket rather than soldering the legs directly on the board as the LTV826 chip on VBs.

Regarding mounting on the cell, the VB approach of one rigid contact bolted to the board and one wire seems to work well. The fixed contact supports/holds the board in place, and the wire gives flexibility in routing around any rod for cell hold-downs. A fully rigid mount between terminals has to clear the vent in between them.

I like it! Simple, low current draw, resistor limiting of current in event of a component failure. I agree that hysteresis is desirable. I think it should be smt, ready-to-install, nothing for a user to screw up. The more user intervention, the more headaches for you.

 

[dimitri]

Everyone, thanks for great comments!
Obviously there are many ways to skin a cat and I tested variety of designs. Russ, I tried your design and it works the same way, but has more components. Main reason for this design is minimum number of discrete components on PCB, which is my main goal, I am minimalist

I don't see a need for more precision, I have 0.01V precision with this circuit, does anyone need more? I observe very little hysteresis, I tried adding feedback and IMHO improvement wasn't worth adding more resistors ( did I mention that I am minimalist ? ).

Dave, your circuit is cool, but I feel that any time / pulse manipulations belong on the head end, cells should report every pulse and be simple. You have way too many components to be on each board, its not practical for mass production. Did I mention that I am minimalist?

I tried to stay away from trimpots, but there are just too many variations of HLVC levels that different people might need, after much consideration I reluctantly came to conclusion to use trimpots to make a product that fits everyone. Calibration can be done by the vendor when customer picks specific values, or by customer, using a simple calibration board which can be sold along with BMS and cost perhaps $10-$15.

As for trimpot adjustment range, LVC trimpot has a good range since reference voltage is right in the middle of typical 2.5V LVC, and HVC and shunt trimpots have their range improved by R3 shifting voltage divison down, I tested it this way and its quite manageable with single turn trimpot, so I don't want to add anymore resistors. Values of R3 vs. trimpots can be selected carefully to further improve the adjustment range, but I feel that 50/50 split should be good enough.

More details in next post, need to step away.....

 

[Dave Koller]

Dave, your circuit is cool, but I feel that any time / pulse manipulations belong on the head end, cells should report every pulse and be simple. You have way too many components to be on each board, its not practical for mass production. Did I mention that I am minimalist?

.

KISS principal you are right.. Keep going!

 

[Dave Koller]

One thought: did you find a need forDecoupling capacitors, Bypass ?
I know it is a few more parts - but maybe a noise suppression help..

 

[dimitri]

I've assembled over a hundred modules with 2 amp shunting, which requires 10W resistor and a heat sink on transistor. From my observations, if we shut off charger at HVC, we don't need to shunt so much current, especially since top balancing is becoming so controversial lately. I want to avoid heat sink and use 5 Watt ceramic resistor to make it compact and low profile, which means less than 1 Amp shunting. I tested with 4.7 Ohm and I had 0.75 Amp going thru and transistor was barely warm, so I think this is optimal based on my requirements. BMS should not be used alone for initial balancing, but for ongoing daily balancing less than 1 Amp is quite sufficient even for 200Ah cells.

Heatsink on BMS modules is one of my biggest headaches, so I will not make same mistake again

What does the 1960 era have to do with the choice of comparators? It costs 22 cents and Digikey has gazillion in stock, is there anything wrong with it? Do you have specific alternatives which would be better?

Tom, why would you want IC in the socket??? I don't get it...

 

[dimitri]

* restrict the range over which the LVC and HVC can be adjustable with additional resistors. We know about where each should be so make the adjustment window maybe 0.4V or so.

I thought about it, but I cringe at adding more parts, while I was able to dial voltages with single turn trimpot just fine, so even though this would make it easier to calibrate, I just don't find it critical. I used R3 to make HVC range more useful, I think this should be sufficient.

* Bypass the pot wipers to both + and - so they are always at AC ground - important to prevent switching ripple from the controller tripping the circuit.

Sorry, I don't get it. You mean bypass with capacitors on both sides? That would be 6 more parts.... . Maybe I just miss your point? I don't always get your pro slang, sorry about that

My current modules don't have any caps and I don't see any false tripping, so perhaps this falls into "nice to have" category? Once I drive around with prototype modules in my car, I will pay attention to this.

* Add hysteresis (small amount of positive feedback) to the comparators so they don't oscillate. Oscillating comparators are really annoying.

I thought about this a lot and even tried on the test board, don't see a need for it in real life because HVC will just trip the charger and latch on, but LVC will oscillate anyway since it will cut controller current and driver will let go of the pedal, so there is no point reducing these oscillations. Again, I tested this with my current modules, didn't see any need to reduce hysterisis.

* it might be more costly in parts, but much cheaper in labor, to use multiple SMT transistors in parallel with thermal vias to a copper heatsink/ground plane. Try to avoid going 4-layer if at all possible.

I will definitely stay with 2 layer PCB, but I again missed your other point. You mean transistors INSTEAD of comparators? How is that cheaper in labor than one 14 pin SOP?

 

[dimitri]

Here is the parts breakdown, comes to $4.69 in parts per cell. My guesstimate for PCB cost is $1.50 per cell, so $6.19 in hardware per cell.
Anyone has any ideas to reduce cost? Maybe there are cheaper alternatives on Digikey?

Looks like $10 per cell assembled is not going to happen, perhaps $15 per cell? Any ideas?



Can't paste from Excel and keep formatting....

 

[Tesseract]

I thought about it, but I cringe at adding more parts, while I was able to dial voltages with single turn trimpot just fine, so even though this would make it easier to calibrate, I just don't find it critical. I used R3 to make HVC range more useful, I think this should be sufficient.

Okay. After messing around the TS160 cells in the New Beetle Rebirth is converting it seems that you could simply set LVC to 2.65V and HVC to 3.85V and start shunting at around 3.65V. Why make them adjustable at all? You may cringed at adding a few more resistors that cost less than $0.02 each but I cringe at putting a trimpot on a board that has to survive moderate to high vibration without failure.

Sorry, I don't get it. You mean bypass with capacitors on both sides? That would be 6 more parts.
...
My current modules don't have any caps and I don't see any false tripping, so perhaps this falls into "nice to have" category?

I bet Ken Hall said the same thing when he designed the PakTrakR...

Bypassing is just good engineering practice; ignore at your risk

I thought about this a lot and even tried on the test board, don't see a need for it in real life because HVC will just trip the charger and latch on, but LVC will oscillate anyway since it will cut controller current and driver will let go of the pedal, so there is no point reducing these oscillations.

Yes there is - an oscillating comparator will not necessarily trip at the correct voltage anymore. Once again, it only costs 1 resistor to add hysteresis to a comparator circuit while the end user may or may not notice a functional difference, it makes for a more reliable circuit. Anyway, go for the minimalist approach and see what happens (easy to suggest because that's what you're probably going to do anyway )

I will definitely stay with 2 layer PCB, but I again missed your other point. You mean transistors INSTEAD of comparators? How is that cheaper in labor than one 14 pin SOP?

Nope. I mean using several SMT power transistors (DPAK or SOT-223 packages) in parallel along with individual shunt resistors in, perhaps, the 2512 package instead of a single power transistor and single resistor, both PTH (that's Plated Through Hole, aka, parts with leads on them). All with extra vias to a ground plane on the underside of the board to act as a noise shield and heatsink. Usually good for 1W per sq. in. of board area.

 

[dimitri]

Okay. After messing around the TS160 cells in the New Beetle Rebirth is converting it seems that you could simply set LVC to 2.65V and HVC to 3.85V and start shunting at around 3.65V. Why make them adjustable at all? You may cringed at adding a few more resistors that cost less than $0.02 each but I cringe at putting a trimpot on a board that has to survive moderate to high vibration without failure.

Sure, this would work for people with TS160 cells, in fact I have those same voltage levels on my car and I am quite happy, however, think about people with SE cells and people up north where LVC can be significantly off and people who climb hills at 4C, etc etc, there isn't even agreement on LVC between 2 people using same exact cells. Same goes for HVC, people may need to dial it up or down depending on their specific needs and/or charger specs, etc etc.

Trust me, I hate adding trimpots, but more I think about it more I get convinced that we can't get away without them if we were to satisfy many demands.

Also, have you looked at the datasheet for these trimpots? They appear to be made for volatile environments.

 

[dimitri]

Nope. I mean using several SMT power transistors (DPAK or SOT-223 packages) in parallel along with individual shunt resistors in, perhaps, the 2512 package instead of a single power transistor and single resistor, both PTH (that's Plated Through Hole, aka, parts with leads on them). All with extra vias to a ground plane on the underside of the board to act as a noise shield and heatsink. Usually good for 1W per sq. in. of board area.

Funny, I realized what you meant few moments after I posted my reply
There is no room for multiple parallel components. I plan to use DPAK package KSH127, so only resistor will be PTH (5W SMT resistors are too damn expensive), not too bad, just 2 leads. I also planned to use extra vias as a heatsink for DPAK module.

I really appreciate your feedback even if it appears that I am ignoring it , every piece of advise is registered and will be tested, but as you can see I am really trying to minimize cost and components.

Actually, let me ask this question, since dealing with noise is my biggest weakness . If you were to add 1 or 2 capacitors to this circuit and get biggest bang for the buck, where would you place them and what values?

Please note that during charging noise is minimal, so shielding HVC is not as important as shielding LVC.

Thanks

 

[moreati]

Trust me, I hate adding trimpots, but more I think about it more I get convinced that we can't get away without them if we were to satisfy many demands.

Just a thought from a naive interloper. Could you place pads on the pcb to give the option of either a trim pot or fixed resistors. Same board different configurations...

 

[Russco]

Soooo.... you see some merit in using the 1970's era TL431 over the 1960's era '339, huh?

I guess you're right. The quad 14 pin plastic case 339 is very futuristic next to the single 311 comparator in a metal TO5 case I used in 1972.

I really can't see how using 3 separate TL431's is any better than a single bandgap reference feeding voltage dividers/comparators. Especially since it's easy to give comparators hysteresis so the switching action can be snappy, rather than fuzzy (the TL431? not so much).

To each his own.

I definitely agree with this, though. If you are going to top balance 160Ah cells then you need to be able to shunt up to 2A. Whether you SHOULD top balance seems to be a matter of considerable debate, but that's for another thread...

Yep... I hate trim pots, but it still might be good/necessary to have a limited adjustment range... Dunno. I'm also concerned about the temperature coefficient of the cell voltage.

Six of one, half a ....

Maybe, but I can tell you for certain that unless these boards are totally SMT technology and machine-assembled it simply won't be worth anyone's while to sell them at the price DIY'ers, etc., are willing to pay. I mean, you can maybe charge $20 per cell of which you'll have a good $7-$9 in parts and pc board cost. If it takes more than 10 minutes to assemble, test and calibrate each one then it's simply not profitable. If your labor cost is $10 per hour or less, and still fast and accurate, then sticking with PTH technology might still be okay, but even then why bother?

You don't see me manufacturing BMS boards now, do you? Mine interfaced with the Russco Charger, providing a cell by cell voltage regulation loop, which PWM'd the charger to reduce the amps out.

Heh. You know how to make a small fortune in the controller business? Start with a large fortune and spend it developing and making something that undersells Kelly, who will then come back and undersell you because their parts costs, labor costs and even the Yuan/Dollar exchange rate are all in their overwhelming favor. But hey, good luck anyway!

That's OK. You need not worry, Tesseract. You make Cadillac controllers. I'll deal with the Pinto controllers. The Pinto controllers on the market today don't even have active precharge. Duh!

Russco went into the EV market 25 years ago manufacturing controllers and I'll probably go out manufacturing controllers. That's what I do!

 

[dimitri]

Guys, I really appreciate your feedback, but please, let's not turn this into "who's got bigger d...." discussion

I've analyzed and tested both "single shunt" and "multiple shunts" designs, they both work equally well, but I think single shunt can be done with fewer components and I personally like how comparators make it a clean, almost digital setup.

If you can please help me with minimum capacitors placement for noise reduction, I'd really appreciate it.