"Zephyr" - Finally, the "v4" Fechter/Goodrum/Hecker BMS...

[fechter]

To make DIY boards like that, it would probably be much easier to use surface mount parts as much as possible to minimize the number of holes needed. Drilling tiny holes by hand would be pretty challenging.

[thedarlington]

well drillings nothing compared to soldering well all those tiny bits, what if we messaged some of those open source nano pc(smoke pack size) guys who does the boards for them? they already sell a $25 700mhz mini pc -raspberri and theres plans for a $15 1ghz one soon -rhombus, thats with CPU, RAM, and plug sockets usb, hdmi phones etc, i bet they wouldn't mind helping us out atleast with tips or maybe take us under the wing to get some prices from whoever makes their stuff, after nano scale mobos im sure these boards are cake

by the way, are you guys still integrating off the shelf stuff like Cell Logs? i found a nicer $10 copy that comes with a large external alarm module probably makes for a nicer trigger,
also an already boosted Battery Medic (like garys old one?) $12 50w & $17 150w, using a light bulb & circuit to bleed-balance, these 2 are pretty much 80% of the the brains & triggers aka the complex bug prone stuff, maybe they could even make for a lite BMS? good enough for now with a more plug and play approach vs extra soldering

http://www.ebay.com/itm/3-in-1-li-polym ... 836wt_1198
http://www.ebay.com/itm/BVM-8s-1-8-cell ... 1e6ed2992f

[andrew.box]

Found this thread and project way later than I should have, looks awesome. I could use one of these, the main drawback of DIY lipo packs is still battery management. I understand the predicament you're in - to finish development to the point of a polished product that can be sold at a profit, or even break-even, at a reasonable number of units, is tough if not impossible given the niche market of DIY electric bikes and small electric vehicles.

In the meantime, are there any plans to make up and sell some bare boards with a parts list and basic instructions for assembly/testing that people could populate and assemble at their own pace? Seems like the margin per unit could be MUCH higher, since labor would be significantly reduced. Maybe a good stop-gap way to recover some costs before proceeding with more development?

I agree that a plug-and-play solution in a nice, pretty case would be sweet, but I for one wouldn't' mind taking care of the connectors, case and details on my own if I had a good quality PCB generated from a well thought out circuit like this.

I'll be subscribing to the thread, this is neat stuff.

[Tommy L]

Nice work here!

Subscribed!

Tommy L sends.... \\m//

[mj2412]

Hello, i'm still trying to solder this BMS. Beause of bad experience i have little concerns about connecting the cell cables one by one to the battery. Is it possible to add a small fuse into the cell cables? I'm thinking about a small fuse to solder into the cables.

What kind of fuse could be used for this?

Thanks, Martin

[fechter]

Yes, I've used small 2A pico fuses that have legs on them. Another approach is to attach one wire at a time through a 6v or 12v light bulb. If the bulb does not glow, then connect directly and go to the next cell until they're all hooked up. If the bulb lights up, something is wrong.

[alternety]

I am just curious. I have followed this topic through I think two different threads, and found it very interesting and potentially useful. I have a battery powered tool using a 36V NiCad pack. I was looking at making a Lithium replacement in the long term using this device. As it turned out, the battery pack failed after something less that 10 charge/recharge cycles. So I am back at looking at this solution.

My question (as in just curious) is has anyone looked at doing this with microprocessors rather than assorted analog solutions. Microcode is much easier to update than an analog implementation. I would think a digital solution would have less probability of circuit layout changes and could reduce power dissipation in some areas by using switching power regulation. Just a thought. It could possibly reduce component count and make it easier to tune performance. A simple microprocessor may cost less that a power resistor. It would also allow a diagnostic code display vs lots of LEDs. With an LCD display, much more operational information could be available.

[heathyoung]

Yeah, there are microcontroller solutions out there, have a look for peter perkins BMS. The biggest problem with micros is noise, lockups and power consumption.

The Cell-Log was used as a BMS for a while by GGoodrum/Fletcher IIRC (I have the board images here somewhere).

Its not a bad idea, the problem is though the unbalanced consumption between the pins. This is solvable, I've done it with external modification, rather than modifying the internals of the unit (yuk). Everyone complains about the cell-log unbalancing their cells if used permanently - the answer - rail to rail opamps set up as voltage followers (ie unity gain). But I digress.

I did like the original design, use it for LVC, delta V and HVC (with a simple shunt reg) - it appeals to the geek factor, plus you can have an indicator of your cells health all the time.

[megacycle]

Unlike an ebike, were generally guys want to maximize charge on larger packs, i was lead to believe power tool lithium packs often dont have much in the bms, basic, so you can fast bulk charge and terminate below the max recommended voltages, so practically they are slightly undercharged.
Please correct me if Im wrong.

[circuit]

if you don't mind large connector on the battery pack, it can contain only an LVC circuit (for each cell). The balancing and OVC should be in the charger then.
If large connector with many contacts is not a solution, then a proper BMS must be integrated in the battery pack, doing LVC, OVC and balancing.

Anyway, some sort of 'advanced' electronics is needed in the pack.

[fechter]

I've still been simmering the design while not active in the project.
The Seiko chips are the closest thing to what I would call ideal, but they are hard to source.

Another one I like is the Maxim MAX11081 chip. One chip does 12 cells and the voltage set points are programmable by jumpers in 100mv increments over a small range. so can be used with any lithium chemistry. This chip has super low standby drain, is highly fault tolerant, and doesn't need optocouplers. They can be ganged together for up to something like 320 cells. And you can actually buy them. It doesn't have shunts or balancing, though, so shunts would need to be added for balancing. I'm working on a digital shunt that has super low standby drain. The 431 linear regulators in the old design draw more current than I like (up to 400uA when close to the set point).

[heathyoung]

Interesting - TSSOP may discourage some though... unless you go to breakout boards (ie TSSOP to DIL) or pre-soldering.

[fechter]

Interesting - TSSOP may discourage some though... unless you go to breakout boards (ie TSSOP to DIL) or pre-soldering.

That's pretty much why I didn't try it already. I don't think it would be too bad though, as the chips are at least big enough to see with the naked eye. Since there's only one chip per 12 cells, the extra time it takes to deal with the package will be more than made up in overall assembly time. For mass production, the package would be fine. I was thinking toaster oven or hot plate method.

[heathyoung]

Yep, especially when dealing with large cell counts (like my 44 series pack - sigh).

Farnell don't list them, I take it mouser or digikey?

[Alan B]

Try octopart.

[alternety]

I am just goggling for some devices. You probably know about the ones I find. But I will toss them in here for perusal. It will probably be a number of posts as I find something.

First one: http://americansolarchallenge.org/wordp ... _count.pdf. A presentation about a TI chip. I believe it is one of the Benq chips.

[fechter]

Yes, I've looked at the TI PowerPump. Certainly has possibilities, but I didn't like the high parts count needed for implementation.

Yes the MAX11081 is stocked at Mouser: http://www.mouser.com/Search/Refine.aspx?Keyword=max11081 1: $8.67ea; 25: $6.00ea

The sister MAX11080 is the same but with much more hysteresis in the alarm trigger. To me, the lower hysteresis model would be better as it would allow cycling near end of charge to complete balancing.

[alternety]

Fechter, I was just doing this survey for my own knowledge. Please don't infer I felt this information was not known. But I figured, maybe one of these was newer or overlooked. A number of the devices I have found are oriented oriented toward automobile battery stacks. They may have some interesting capabilities (like reporting status/conditions) but they use CANbus and at least an additional level of microprocessor. Way to complicated and expensive. But nifty.

Amtel has one (http://www.atmel.com/search.aspx?q=+ATA ... collection) with a fairly user hostile web site. It does about what you are doing but everything is in the digital domain so is less prone to noise in various various places.

I completely agree on the parts count on the TI Power Pump. I did find a couple of things interesting. When targeting small electric vehicles (e.g., bikes) I believe they they care about efficiently recovering power from dynamic braking. Using the inductive redistribution would seem to help in this objective. Would it be noticeable; no idea. Personally I just need to recharge a 36V battery from mains power. Stupid string trimmer. Works much like a gas powered unit. Although the string is too thin. But it used NiCd batteries. Charged for 24 hours, ran for 15 minutes if you were lucky and had real short weeds. After about 6 cycles with the battery - no -good, refuses to charge. No replacements are available.

I am curious about the chemistries your device can deal with. Some newer formulations (e.g., LiNiMnCo) seem to have a good price performance. I was going for A123 cells, but those look interesting. Two less cells for the pack.

Do you actually have PC boards available at this time?

[fechter]

Sorry, no boards are available right now. At the rate things are going, I'm going to have to retire first before I have time to really get things rolling again. I'm still working on it in the background though. If I could hook up with somebody that is interested in producing boards, it could speed things up a lot.

I agree it's a good idea to keep checking on commercial chips. They come out with new ones every so often. A microcontroller based system still has potential to be cheap and effective.

My earlier versions are all capable of working with any lithium chemistry as the voltage set points can be changed depending on parts selected.

I've also looked into taking an existing chinese design and modding it to increase the power handling. None of them have a really great track record for reliability.

Methods' HVC/LVC boards do what they are designed to do and may be a solution, but a bit spendy for me (I'm cheap).

[methods]

Methods' HVC/LVC boards do what they are designed to do and may be a solution, but a bit spendy for me (I'm cheap).

I was searching the google with a Maxim part number and I found this thread -
I am glad to hear that my boards work (not being a smart ass - that is actually good news to hear )- and you are absolutely right, I do have them over-priced - hence why I am searching the internet for that maxim protection chip.

Anyway - I read something above about populating boards. Matt has a crew now that can populate. Have not seen this new BMS but if it is as parts heavy as the old one someone is going to have a sore wrist Let me see a finished picture and let me know what you are wanting to pay. Doing 1 is way harder than doing 10....

My vote is to go surface mount - big ones - then stencil on the paste and bake. -FTW- On our last run we got clever - we found a way to stencil on all the surface mount parts, bake them, then populate all the JST-XH headers, then dip into the solder pot. At just the right angle I could get 36 perfect solder joints in about 1.5 seconds. Only bridge a few of the tiny SOT-23-5 pins when I missed. Speaking of that.

SOLDER POT FOR THE WIN

Just push all those parts into the board, paint it with flux, dip it once, and trim the leads.
Done - and every joint will be picture perfect.
FTW - FTW - FTW

Make a Ghetto Solder Pot with a bunch of old rolls and a blowtorch! Just enough stove heat to almost melt it - hit it with the torch - dip the board. Just dont breath the smoke for hours like I did last week

-methods

[alternety]

Maybe Boeing would chip (sorry about that) in. I understand they may wind up looking for a good Lithium charge controller.

Or buy A123 cells.

[fechter]

Funny how ES seems to show up in a lot of my searches too. We're almost famous here.

My latest thought is to graft a MAX11081 to a new version of the shunt circuit.
The old 431 based shunts draw close to 400uA on standby when the cells are near full. This drops to about 200uA by the time they approach LVC, so average around 300uA.

The new circuit will drain more like 10uA on standby, and operates in switching mode so the transistor doesn't dissipate much heat. It's bad enough to dissipate all the heat from shunt resistors, so if the transistors don't get hot, that makes heat sinking much easier.

Since the MAX11081 already has the 'any' logic built in, we can use that and add a chain of opto couplers to detect the 'all' condition and terminate charge. Optos suck for LVC since they draw so much. MAX takes care of that. During charge, power consumption is not an issue, so cheap optos will work fine. 12s boards can be concentrated to one opto to get chained to other boards. This should be fine for up to about 240s cells. For less than 12, a jumper needs to be added.

The charge control/LVC interface should be separate. 12s cell boards with shunts could then be added.

There are still a few issues with the new shunt circuit, but initial testing is promising. I don't know what the parts count will look like yet, but for sure much less than previous designs. Much smaller too, depending on shunt current.

I can do a 8s version for Boeing.

[SlyCayer]

I can't wait for you guys to complete this, I am eager to build one.

[methods]

I have the opto's I am using working at less than 100uA. Those are about $0.33 each. There is another flavor that will work down near 10uA but they are more like a buck each.

Hey - you know that those detectors that you pointed me to for my design are now for sale @ sub 3,000pcs right? They can now be had in small batches depending on availability. My method is to truncate the part number and feed it into OctoPart. This brings up every flavor ordered from cheapest/most available. Then I take the part number and cross-reference the setting points - there are a few that might work for what you are doing.

Maxim chip is probably a better route.
If you need I can spin a tiny piggy-back board that converts it to DIP

-methods

[bigmoose]

I liked the MAX11081 too: wished the LVC range extended up a few tenths of volts more though.

LVCHVCbottom-1 (300 x 225).jpg (27.96 KiB) Viewed 180 times

LVCHVCbottom-2 (300 x 225).jpg (20.29 KiB) Viewed 180 times