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

[PJD]

In my case with an older board, once all the LED's are flashing, the charging current continues to decline, and the master LED only gradually changes from red to orange to yellow to greenish yellow. I am assuming that although all the shunts are on, they are still on a low-percent duty cycle, and the the cells are still taking current, so I have been allowing about 45 minutes after all LEDs are on before I consider the pack fully charged. If the charger is left on for about 1.5 hours or more, the LEDs will stay lit, then fade off over several seconds, after the charger is shut off. I interpret this to mean the charge sufficient so the cells are passing practically no current, and consequently, their voltage is constant, and apparently slightly higher than the shunt voltage, so the shunts must "bleed" the voltage down when the charger is shut off

So with your current version, I am concerned that the charging will get cutoff before the cells have seen a full completion of the CV charge cycle. Is this a concern?

 

[fechter]

I never understood the purpose of the connections on the board for the charger positive and pack positive leads. Only the negative lead needs to go through the board and then to the negative pack terminal. The charging positive lead can simply be run directly to the positive terminal of the pack.

Am I missing something?

No. You can leave those connectons empty. They're just there for wiring convenience if you want to use them.

As far as the automatic cutoff goes, as soon as one cell get up high enough to start throttling the charge current, then the current in all the cells will be around 500ma. By the time the last shunt comes up enough to trigger the opto, the current going through the last shunt will be nearly 500ma also, meaning that the cell itself is not taking any more current because it is fully charged.
I can't think of any conditions where it would cut off before all the cells were fully charged.

 

[reagle]

Guys, I know there's been talk about respinning the board as an SMT version for easier manufacturing/smaller size/lower cost - is that still the plan?
I happen to work at a company that does contract design and also manufacturing, so if needs be we could look at making these on an automated pick and place line.(Just need the layout to be done in a tool that generates Gerbers for better choice of board vendors- Express PCB is good, but not when you need to shop for pricing. You also need fiducials for pick and place machine to see them, and finally need a solder paste stencil. I could probably respin the board in Pads if needed)
Think a panel of say 20 individual board, with parts getting placed by a robot in less than a minute, then going through reflow oven, and you have all done at once. A few issues I see are:
Currently, the end user buys bare boards and parts and then populates just as many sections as needed. When building on a line, you probably would have to build all, and then break them according to scoring? 2. Somebody would have to buy all the parts on reels. and 3. Financing it all get tricky, as you d' have to hold stock.
Gary, PM me if you need help with this- I may be able to get my boss interested in building these.

 

[fechter]

Yes reagle, you are correct.
Neither of us has software that can do Gerbers right now.

We need to find someone who has it and knows how to use it. Next step is to take an ExpressPCB version of it and translate to the new program manually. I don't know of any other way for now. Can we bribe you?

Anyone else have design software that can generate Gerber files? I know Gary is working on this too and I don't know where he is on it. Could be already in the works for all I know.

Yes, in a factory built version, the entire board would be populated, but we could do a master and add-ons that can be ganged together for more cells, rather than the one big board that we cut off.

We don't have any money either, which is also a problem.

 

[ldoolit]

I have experience with PCB http://pcb.sourceforge.net/, and have used it for much more complex surface mount boards than what you have in mind. Yes, it puts out Gerbers and x-y placement files. Although my time is ... overcommitted, I would be happy to give this a shot. At a minimum I would need a PDF (or high-res bitmap) schematic, and a sketch of parts placement. It would be nice to also have a text netlist, so I could put it through DRC (design rule check). I would post design files so others could build on my work.

BTW, PCB is http://en.wikipedia.org/wiki/Free_software, so you could use it too. Be warned that its learning curve is steeper than a lot of people are comfortable with.

 

[lawsonuw]

Huh, figured you guys were using more sophisticated software. Personally I use TinyCAD to generate schematics, and a net-list for the layout program. To layout a board I then use FreePCB to manually route the traces. TinyCAD is a bit quirky, but does the job, and FreePCB is an EXCELLENT manual layout program. (both programs have downloadable windows binaries)

FreePCB can generate pictures of each layer when it makes gerbers, but ViewMate is another good option. http://www.freedfm.com is a free service from Advanced Circuits that will check your gerbers for obvious errors. Advanced has decent prices for production, but rather steep for just one. (though the "bare bones" spec ain't too bad) If I just need one and can wait for the board BatchPCB is my favorite. I'm also trying to get some experience with surface mount parts. Routing everything on one layer and a full ground plane solves a lot of noise problems. Using surface mount parts with this type of routing keeps all the critical tolerances on the one top layer. So, it shouldn't be too hard to homebrew a PCB via toner transfer.

Marty

P.S. who writes the manual for a piece of graphics software and doesn't use ONE picture? reason for retorical question =>um... why no PICTURES!!!

 

[ejonesss]

radio shack used to sell and i think they still sell the etching supplies for diy and prototyping.

also you can get the prototyping sockets
http://images.google.com/images?hl=en&q=breadboard&um=1&ie=UTF-8&sa=N&tab=wi

it shouldn't be too hard to homebrew a PCB via toner transfer.

Huh, figured you guys were using more sophisticated software. Personally I use TinyCAD to generate schematics, and a net-list for the layout program. To layout a board I then use FreePCB to manually route the traces. TinyCAD is a bit quirky, but does the job, and FreePCB is an EXCELLENT manual layout program. (both programs have downloadable windows binaries)

FreePCB can generate pictures of each layer when it makes gerbers, but ViewMate is another good option. http://www.freedfm.com is a free service from Advanced Circuits that will check your gerbers for obvious errors. Advanced has decent prices for production, but rather steep for just one. (though the "bare bones" spec ain't too bad) If I just need one and can wait for the board BatchPCB is my favorite. I'm also trying to get some experience with surface mount parts. Routing everything on one layer and a full ground plane solves a lot of noise problems. Using surface mount parts with this type of routing keeps all the critical tolerances on the one top layer. So, it shouldn't be too hard to homebrew a PCB via toner transfer.

Marty

P.S. who writes the manual for a piece of graphics software and doesn't use ONE picture? reason for retorical question =>um... why no PICTURES!!!

 

[ejonesss]

isn't there any mac versions?

Huh, figured you guys were using more sophisticated software. Personally I use TinyCAD to generate schematics, and a net-list for the layout program. To layout a board I then use FreePCB to manually route the traces. TinyCAD is a bit quirky, but does the job, and FreePCB is an EXCELLENT manual layout program. (both programs have downloadable windows binaries)

FreePCB can generate pictures of each layer when it makes gerbers, but ViewMate is another good option. http://www.freedfm.com is a free service from Advanced Circuits that will check your gerbers for obvious errors. Advanced has decent prices for production, but rather steep for just one. (though the "bare bones" spec ain't too bad) If I just need one and can wait for the board BatchPCB is my favorite. I'm also trying to get some experience with surface mount parts. Routing everything on one layer and a full ground plane solves a lot of noise problems. Using surface mount parts with this type of routing keeps all the critical tolerances on the one top layer. So, it shouldn't be too hard to homebrew a PCB via toner transfer.

Marty

P.S. who writes the manual for a piece of graphics software and doesn't use ONE picture? reason for retorical question =>um... why no PICTURES!!!

 

[michaelplogue]

Sorry to keep asking these silly questions - sooner or later I'll eventually run out....

I'll be running six packs with these BMS in parallel. If one of them trips the LVC, will I be able to tell which one caused the cutoff by looking at the master LED's and/or individual cell LED's? That way I would be able disconnect that pack and continue on with the others.

And once an LVC does trip the cutoff at the controller (brake or throttle) , would disconnecting the 'bad' pack allow the controller to work again, or would I need to disconnect all of them in order to "reboot" the system?

 

[ejonesss]

not to worry

each question answered creates a new question.

Sorry to keep asking these silly questions - sooner or later I'll eventually run out

Sorry to keep asking these silly questions - sooner or later I'll eventually run out....

I'll be running six packs with these BMS in parallel. If one of them trips the LVC, will I be able to tell which one caused the cutoff by looking at the master LED's and/or individual cell LED's? That way I would be able disconnect that pack and continue on with the others.

And once an LVC does trip the cutoff at the controller (brake or throttle) , would disconnecting the 'bad' pack allow the controller to work again, or would I need to disconnect all of them in order to "reboot" the system?

 

[ldoolit]

P.S. who writes the manual for a piece of graphics software and doesn't use ONE picture? reason for retorical question =>um... why no PICTURES!!!

With a little more digging, you can find http://www.delorie.com/pcb/docs/gs/... plenty of pictures there.

Sorry to stray so far off topic. If people want to talk free CAD tools, and the relative importance of freedom vs. no charge, we should start another thread.

 

[fechter]

Sorry to keep asking these silly questions - sooner or later I'll eventually run out....

I'll be running six packs with these BMS in parallel. If one of them trips the LVC, will I be able to tell which one caused the cutoff by looking at the master LED's and/or individual cell LED's? That way I would be able disconnect that pack and continue on with the others.

And once an LVC does trip the cutoff at the controller (brake or throttle) , would disconnecting the 'bad' pack allow the controller to work again, or would I need to disconnect all of them in order to "reboot" the system?

No, you won't be able to tell which one tripped. The LEDs only work for the charging circuit. We didn't put any indicators on the LVC part since this would drain an already low pack further and possilby cause a cell to go too low.

It wouldn't be too hard to add a LED to indicate a LVC trip, but it would only indicate when the LVC is active. Normally, the LVC will trip and cut power to the controller, but will recover almost immediately when the load is removed. I suppose you could do some kind of latching indicator, but that would add quite a bit of complexity.

If the offending pack is disconnected, the others will work normally. I'd guess that normally by the time one trips, the others won't be far behind, so disconnecting may not get you much.

For parallel packs, I would normally suggest connecting them at each cell so that a single BMS can work for all of them.

 

[reagle]

I think PCB 123 will give you Gerbers after you order at least once from them, but the software is very buggy.
We use Mentor Graphics Pads at work, and at $20k a seat it's not something we should be picking for an open source project that needs to be maintainable by multiple people. I.e. I can pretty easily recapture schematic in that tool, then do layout from scratch, using industry standard libraries for footprints and generate something that looks like Gerbers (I am a bit fuzzy on that last step), but then the source file is useless to the rest of the people here if they wanted to tweak things. We could probably figure how to make Fiducials in Express PCB, but I am at loss as to how to make a solder stencil out of the DXFs the tool generates, or how to properly panelize them in the absense of scoring. I'll ask E-PCB just in case they got something in works.
Also, looks like Advanced Circuits have a free tool http://www.4pcb.com/index.php?load=content&page_id=306, but I have no experience with it.
I guess I'll try playing with a few other toolchains to see if they are any good. My own needs are dictating a board respin anyways. I am using a bunch of prismatic cells with screw terminals, and will probably do either a board per cell or one board that picks up voltages from the screw terminals, and does LVC, OV protection and also drives a couple of really big Fets to actually disconnect load and or charger.

As far as building them- what volume are we talking about near term?

Yes reagle, you are correct.
Neither of us has software that can do Gerbers right now.

We need to find someone who has it and knows how to use it. Next step is to take an ExpressPCB version of it and translate to the new program manually. I don't know of any other way for now. Can we bribe you?

Anyone else have design software that can generate Gerber files? I know Gary is working on this too and I don't know where he is on it. Could be already in the works for all I know.

Yes, in a factory built version, the entire board would be populated, but we could do a master and add-ons that can be ganged together for more cells, rather than the one big board that we cut off.

We don't have any money either, which is also a problem

 

[reagle]

Hmm, what do you know- I just talked to Express PCB and it turns out you can actually buy Gerbers from them for any previously ordered board. It costs $60, and is done via email

"You can order gerber files by email for PCBs that we have previously manufactured. The cost of these files is $60. This fee will be billed to the credit card number used in the original order for the boards. You will receive these gerber files in RS-274X format:
+ Top silkscreen layer
+ Top soldermask layer
+ Top copper layer
+ Bottom copper layer
+ Bottom soldermask layer
+ Drill file
+ 4 layer designs also include the 2 inner layers "

 

[cycle9]

Thanks again for the great board and all the good information posted here. I've read through it all, and don't think my questions have been answered (I apologize if they have).

I built a board over the weekend (and another partial one). I believe I fully understand the theory of operation, but I don't understand what I'm seeing in my testing.

The problem seems to be at the optos. I can connect my power supply and ramp up the voltage until all orange led's are on (and the main is red). However, even when all the orange ones are on, I have to increase the supply voltage to equal about 6V/cell before the red starts turning green, and the FET starts significant current limiting. Also, before that point, the current is around 0.8 A, not 0.5A as it should be. So the signal controlling the output LED and FET is wrong.

So, next I measured the voltage at the "any shunt active" and "all shunts active" to see what is happening. It is getting pulled down, but not by very much. Again, by the time I have my supply ramped up to 6V/cell, the opto-outputs are pulled down to only around 4-5v (seems like it should be zero at that point). When the orange LEDs are first lit, there appears to be no pull-down at the opto output. I checked the pull-down resistor at the opto input, and it is the correct value (820 ohm). I checked the voltage drop across the opto input, and it is around 1V when it should be active.

Does any of this make sense? I'm confused as to why these seem to sort-of be working, but not correctly.

TIA,
Morgan

 

[cycle9]

To follow up on my last post, I did some further diagnosis... the opto couplers are working.

When the voltage at individual cells > 3.7V, then "Any shunt active" does get pulled down to about 2V (it is 12V before). I don't know what my problem was last time I measured this, but this part of the circuit is working now.

Unfortunately, the gate driver does not seem to be responding - neither the out A(not), pin 7, nor out B, pin 5, are responding to an input change at pins 2 and 4, from +12V down to +2V. It seems like it should be responding, so maybe I fried it somehow. Unless I misunderstanding how it is supposed to be working.

Thoughts?

Thanks,
Morgan

 

[fechter]

Hmm.... 2v is close to where the gate driver should toggle. There is a pull-up resistor on the gate driver inputs. Make sure you have the right value there. I think it should go down to closer to 1.5v when an opto is on. There could be a bridged trace on the any shunt active line pulling it up.

You can power up the board so the status LED is red, then try shorting the input of the gate driver to ground. This should make the LED go green. If it does not, then the driver is bad.

You could also try shorting the all shunts active line to ground and see if the SCR triggers and holds the LED green.

Hold the board up to a light and look for bridged traces. When backlit, any solder bridges will show up as dark.

 

[cycle9]

Thanks a bunch for the quick response.

Here are the diagnostics:
- shorting gate driver to ground: gate driver works, LED goes green
- pull-up resistor at gate input 4.7k ohm, as specified
- check for bridged traces - check, none found
- shorting all shunts active to ground - LED goes and stays green

so, gate driver is working. But, voltages coming out of the optos are strange:

When I first hit ~4.7V/cell and the orange LEDs come on, the opto outputs read around 8v, then drift down to ~5v over about a minute or so
Input to the optos is about 1V

Trigger point for LED's is almost exactly correct at 4.7V per cell. Pull-down resistors at + input of opto-couplers looks correct at 100 ohm, and 820 ohm at the cathode.

Very strange - it seems very close to working, but not quite right. Any more thought appreciated.

Thanks again,
Morgan

 

[fechter]

Yes, most of it sounds right.

I'd suspect a short between the 12v output of the LM7812 and the any shunt active line. There's a 12v bus that runs along the optos. You could try to locate the two buses and measure resistance between them with the power off. Use diode check if you have it.

One other possibility is one or more optos is installed backwards. It is hard to see the notch on these. Look at the closeup pics earlier in the thread.

Also, making voltage measurements with no cells connected will tend to be inaccurate when the control circuit is throttling. If you slowly increase the voltage until the orange LEDs are barely lit, then you should be able to get an accurate reading.

 

[Tiberius]

Richard, Gary,

I have a question/observation for you about the circuit, if I may. Have you run this at 24 cells, and what charger system did you use?

If I understand it correctly, when the first cell reaches full V, Q2 will turn off. Because Q2 disconnects the pack negative rail, the LM7812 regulator U1 will then see the charger open circuit output voltage, less the drop through the top cells. U1 probably has a 35 V upper limit.

With a 16 cell pack, you could have a charger open circuit voltage of 70 V, with a drop of 11 cells at 3.3 V each, leaving 33 V into U1. Higher cell counts will push that peak input voltage higher. What do you reckon?

Nick

 

[reagle]

Another question- when using your BMS with Thundersky batteries, am I correct to assume that one would need to tweak the voltage limits on both charge and discharge?
Their curves seem to show 2.5v or so ideal LV cutoff, and a charge up to 4.2V , which is kinda strange for LiFePO4. I am contemplating getting a charger/cells package from Elitepowersolutions, but the difference in voltages for their cells vs the PSI/BI/Headway round ones worries me a bit

 

[ejonesss]

with those voltages can a normal lithium ion charger be used? meaning can you be able to wire up a cell phone to use the cell phone as a charger?

Another question- when using your BMS with Thundersky batteries, am I correct to assume that one would need to tweak the voltage limits on both charge and discharge?
Their curves seem to show 2.5v or so ideal LV cutoff, and a charge up to 4.2V , which is kinda strange for LiFePO4. I am contemplating getting a charger/cells package from Elitepowersolutions, but the difference in voltages for their cells vs the PSI/BI/Headway round ones worries me a bit

 

[reagle]

I am guessing we could- I am just questioning if the TS's data is accurate, as every other LiFePO4 cell seems to be charging at 3.6-3.7V or so.

 

[fechter]

Hmm, what do you know- I just talked to Express PCB and it turns out you can actually buy Gerbers from them for any previously ordered board. It costs $60, and is done via email

"You can order gerber files by email for PCBs that we have previously manufactured. The cost of these files is $60. This fee will be billed to the credit card number used in the original order for the boards. You will receive these gerber files in RS-274X format:
+ Top silkscreen layer
+ Top soldermask layer
+ Top copper layer
+ Bottom copper layer
+ Bottom soldermask layer
+ Drill file
+ 4 layer designs also include the 2 inner layers "

Excellent. Thanks for that.

 

[fechter]

Richard, Gary,

I have a question/observation for you about the circuit, if I may. Have you run this at 24 cells, and what charger system did you use?

If I understand it correctly, when the first cell reaches full V, Q2 will turn off. Because Q2 disconnects the pack negative rail, the LM7812 regulator U1 will then see the charger open circuit output voltage, less the drop through the top cells. U1 probably has a 35 V upper limit.

With a 16 cell pack, you could have a charger open circuit voltage of 70 V, with a drop of 11 cells at 3.3 V each, leaving 33 V into U1. Higher cell counts will push that peak input voltage higher. What do you reckon?

Nick

Yes, you are correct. 24 cells will be pushing the limit for the voltage regulator.
One thing that helps is Q2 will be on when you first connect the pack, and the cells have to charge up before Q2 opens for the first time. By then, the voltage difference will be much less and within the limitations of the 7812.

For higher cell counts, it is recommended to feed the charge control circuit from a separate small 12v power supply (only needs less than 50ma). This eliminates the possibility of over volting the regulator. It would also be possible to use a zener/transistor regulator in place of the 7812 that has a higher voltage rating.