BMS package for A123 cells

Jeff

1 W
Joined
Aug 24, 2007
Messages
61
Hi,
I see there's been a lot of discussions about monitoring and controlling charge on the newer phosphate chemistry cells.

I've been working with the A123 cells for awhile now, mostly for integration into large scale packs for EV's. I'm also a TidalForce bike owner, and have another bike with a C'lyte 408 motor and 72v controller.

To help our fellow TF owners, I assembled a relatively simple system to clamp the cell voltage during charge, and reduce the charge current when the preset voltage is achieved on any cell.

An MCU monitors each of the vclamp modules, linearly decreasing the charge current until charge is eventually terminated when the last vclamp activates after the charge current is reduced to I minimum. I minumum is 50ma per cell in any parallel configuration, or max time achieved, or temp max from an embedded sensor in the pack(s).

The charger is a 1kw lab power supply that is being driven by the PWM output feature of a Uchip MCU. I've ordered a couple of low cost 500w SMPS's from a HK distributor that will be used in the final design.

Image of the individual vclamp board.

An image of the vclamps installed on 6p A123 clusters.


The schematic of the vclamp in pdf.

I'll post the schematic of the MCU board when the power supplies arrive. It'll probably require a board revision to accommodate the changes needed for these cheap power supplies.

I did not include a low voltage cutoff (LVC) in the Tidalforce version, since the controller can do this task.
This is recommended if your controller does not include an LVC circuit and you're not familiar with the actual performance of the cells you have used to build the pack.
An LVC can be added to the clamp circuit with another small darlington, a few passive parts, and another optoisolator. Or a low voltage reset IC, and a couple passives, and the optoisolator.

Regards, Jeff
 
Yes, very nice. Did you make custom boards for those?
Any idea what the price would be?
 
fechter said:
Yes, very nice. Did you make custom boards for those?
Any idea what the price would be?

Yes, though they are very cheap at about $2ea in 100 qty with short leadtime.
I goofed on the first fab at component Q1 (note the funny rotation). I corrected the part configuration and generated new gerber and drill files for rev B. I've not run another batch yet.
If I increased the quantity to 250, with a 3week lead, that might cut the price to under $1. I'll look into it.

Knowing most everyone here *really needs* a Low Battery Flag, I might make a quick circuit change to add components for a LVC opto output to flag the controller or the rider via some separate visual indicator.
Though a drainbrain or other method might avoid battery damage, the safest way is to watch for it at the cell level.

Just about everyone makes a LV reset IC in SOT-23 (SC-70) packaging. I need to identify something sink-only, capable of driving an opto without a buffer, and stocked at a common trigger voltage for these cells, ie: 2.6-2.7v.

Regards, Jeff
http://www.evbones.com
 
Oh yes!
Somebody who knows how to get boards fabbed!

If the circuit could do both low and high voltage along with the cool shunt feature it already has, that would be sweet. It would be easy enough to interface to most controllers for LVC and most chargers for charge termination.

For added 'coolness' an LED indicator that shows when the cell is at the high limit or something might be a nice feature.

The price definitely sounds right.
 
If you want to presell some to increase your order size I'm in for for 100!
PM me if your interested Jeff..
 
I presume that's the price for the unpopulated board?

How much for a populated "turn-key" solution? (I can't solder surface mount components for sh!t).

Thanks.
 
DeeL2003 said:
jdh2550 said:
I presume that's the price for the unpopulated board?

How much for a populated "turn-key" solution? (I can't solder surface mount components for sh!t).

Thanks.

Count me in on the "can't solder worth ish list" too.
Actually that one is easy aside for the fuse and the misplace SOT23, and the SOT23 is mostly just the confusion factor.

I can ask but it would be for 100-1000 boards. Problem is it would cost us 500$ just for the artwork.... I do keep trying to get a milling system in here but as yet have gotten nowhere. (a miller would let is spin the boards as protos but would still be pricey)

Dan
 
Very cool indeed. I'm not sure I yet understand completely what this does, but thanks to Bob Mcree's LVC design, which I do understand, I can see how that function could be added.
 
cadstarsucks said:
Actually that one is easy aside for the fuse and the misplace SOT23, and the SOT23 is mostly just the confusion factor

Err, :? , your post mostly confuses me. Are you saying it's easy to solder? That might be the case for you, but not I.

Problem is it would cost us 500$ just for the artwork
Second up, your talk about $500 for artwork doesn't seem to jive with the fact that Jeff can get boards for $2 a pop now. Isn't the artwork required for producing the board (and the milling machine for drilling the holes).

I'm trying to understand what it would cost to receive a board populated with all the components which I would then hook up to the a123 cells. I'm not sure your post covers any of that -

or am I confused still :?
 
fechter said:
For added 'coolness' an LED indicator that shows when the cell is at the high limit or something might be a nice feature.

The price definitely sounds right.


There is a green LED at D2 to indicate the vclamp is activated.

Regards, jeff
 
Jozzer said:
If you want to presell some to increase your order size I'm in for for 100!
PM me if your interested Jeff..

I can have the boards fabbed in volume from one of the many offshore suppliers.

I can also post the complete zipped fabrication data, and one of the members here can also do it.
A bill of materials will be included, and the components can be purchased from either Mouser or DigiKey.

I *really* do not want to assemble these boards in volume. The images and schematic were put here to help folks do it themselves, or through assistance by other members.

Placing these tiny surface mount parts are recommended only for the experienced, using tweezers and a temp controlled station.

Regards, jeff
 
jdh2550 said:
I presume that's the price for the unpopulated board?

How much for a populated "turn-key" solution? (I can't solder surface mount components for sh!t).

Thanks.

Yes, the price I estimated was only for the fabricated bare board in volume.
In qty 100, the component materials cost (with an LVC output) would be $7-$8 per board.

Hopefully someone here will opt in for assembling these in volume for other members. I just don't have the time to assemble them.

Regards, jeff
 
GGoodrum said:
Very cool indeed. I'm not sure I yet understand completely what this does, but thanks to Bob Mcree's LVC design, which I do understand, I can see how that function could be added.

Thanks for the heads up. I found Bob's postings about the LVC circuit in the ebike technical. He's a smart guy. The LVC can certainly be done nicely with a low voltage reset IC. It would only add another $1 in optional materials to the vclamp board.

Regards, Jeff
 
Cool!

I've been looking at designing something just like this, to get away from having to use 10 cells with the DeWalt charger, with slow charging, on a 10s4p pack. This is a great first building block in the construction of larger automotive packs.

A few questions:
1) It looks like you've improved upon the MetricMind clamper design (http://www.metricmind.com/ac_honda/bms.htm), by making precise voltage cutoff adjustment easier, using surface mount parts, and designing a board with a footprint less than a single A123 cell. Did your circuit result from experiments with the MetricMind design?

2) Those packs are beautiful. How did you get the spot welding done? Did you have access to a spot welder, or build your own? Also, what is the material used for the straps?

3) I'd love to see your BMS design - are you going to open it up? Using an AVR, PIC, or other?

My suggestions for the board are to add a through-hole connection at the top, so a larger LED could stick out of the heatshrink, and of course the Bob McRee LVC. Then I'd say sign me up for 15.

Also, a few comments to forum members:

- For small board fabrication at low cost, two nice options are batchpcb.com ($10 setup, $2.50 per sq in, min qty 1) and the bare-board service from pcbfabexpress.com ($40 setup, $.60 per sq in, min qty 4, no silkscreen or soldermask).

- For larger fabrication, I've had good results with pcbfabexpress.com and 4pcb.com.

- Surface mount soldering quickly becomes faster than through-hole soldering, but only if you have the right tools. You'll need the following:
-- temp-controlled soldering iron with fine tip (I use a Weller WESD51)
-- SMD tweezers
-- flux pen - always apply to pads before soldering
-- fine solder (I use .015in -.020in diameter flux-core solder)
-- good lighting and ventilation really helps

- If you're going to do small production runs, I've gotten professional results from untrained students using the solder paste, laser-cut stencil, and toaster method. Sparkfun has tutorials on this, and one example board can been seen here, with around 1000 pads per board (http://photos1.blogger.com/blogger/6304/1572/1600/IMG_2095.jpg), mainly tiny TSOPs. The method works well for medium-size boards as well.

Thanks in advance for the question answers, Jeff!
 
Well, if we're going to redesign the board, why not add two pairs of holes for the outputs of the optocouplers on each board so the outputs could be easily daisy-chained with wires.
 
Below is a copy, with permission, of questions asked to Jeff via PM. He is busy, but is working on an amazing A123-powered EV Attack, with details at http://www.attackforums.com/showthread.php?t=2419

[First Email]---------------------------------------------------------

Hi Jeff,
I had a few questions about your BMS that I posted to the forum, but I have to assume you didn't get them:

A few questions:
1) It looks like you've improved upon the MetricMind clamper design (http://www.metricmind.com/ac_honda/bms.htm), by making precise voltage cutoff adjustment easier, using surface mount parts, and designing a board with a footprint less than a single A123 cell. Did your circuit result from experiments with the MetricMind design?

Hello Brandon,
sorry, I've been too busy at work lately, and have not had any time for reading the forums. Spare time has been spotty lately.

On the Vclamp; I started off with the basic tl431 circuit published by electronic design back in 2000.
http://electronicdesign.com/Files/29/1155/Figure_01.gif
This circuit did not deliver a sharp enough "knee" for the clamp voltage using the bipolar transistors listed, and I continued looking until I came across an improved circuit by Victor using darlingtons. It was flexible, and required only small component changes to adapt to the A123 cell cutoff voltage. The high gain darlington transistors were the answer.

2) Those packs are beautiful. How did you get the spot welding done? Did you have access to a spot welder, or build your own? Also, what is the material used for the straps?

I use a table mounted 240ws parallel probe resistance welding station with a foot operated control. It's a Seiwa used for volume production. The A123 cells require large tabs, and a high power/short duration pulse welder to achieve a strong connection.
I could not find the exact model, though a similar Seiwa system with parallel probe head is shown below.
http://www.seiwamfg.co.jp/e/manu/top_spot_a.html

The strip material is .010" thick x .250" wide Nickel 201. Basically pure Nickel foil in ten mils thickness. I bought it last from Sunstone engineering, in roll form. I believe they slit cut it from a log to the customer specs. Terribly expensive.


3) I'd love to see your BMS design - are you going to open it up? Using an AVR, PIC, or other?

I hope to have the processor section posted shortly. I'm awaiting a 1kw power supply to arrive from overseas, and will work to adapt the control section to the power supply.

The processor (currently) is a PIC 16F88, with a relatively simple program for monitoring the opto outputs of each vclamp. The PWM output of the processor can vary the output voltage of the power supply, functioning as a cc/cv profile charger.
Bulk charge begins and continues until any one of the vclamps activate. Charge current is linearly reduced until the vclamp(s) releases, or current drops to 50ma per cell X n cells in parallel, and all vclamps are reporting.

I added a little bit of extra code to monitor a 16B20 temperature sensor, and a test routine to pulse the pack to measure a single bank of cells out of balance and lagging the others by more than a couple of minutes. A short duration high current charge is applied, and if the bank has not yet reported full, it can be used to estimate how long until it is full, based on how much current is applied to trigger the vclamp. The excess energy is discarded as waste heat from the other banks that are already reporting full.

I'm planning on posting it all as open source. I'm hoping this can be followed on and improved by others.

Regards, Jeff Thomas

[Second Email]-----------------------------------------------------

Thanks for the reply - can I put it on the forum, unedited of course?

Sure, I wish I had time to read what's going on there. I'm completely buried with work right now.

A few more questions:

-How did you get access to the welder? Aside from going to a battery research lab, I'm not sure how to find a spot welder. Are there are other places that typically have spot welder? Another option is to make one.

I purchased the Seiwa welder to attach the nickel tabs to cells that are being assembled in volume for full size EV packs.
One of the projects underway utilizes a few thousand A123 M1 cells, and was posted to the forum awhile back. You can read about it here:
http://www.attackforums.com/forumdisplay.php?f=11
A thread about the EV attack project details the cell modules and their construction.
I've not updated the forum in the last month, but hope to soon.

-Why spot welding vs soldering to cut tabs on the A123s?

The cells (many thousands) I purchased do not have tabs installed on them, unlike the hobby cells or DeWalt DC9360 packs.

I know A123 recommends against soldering, and I've heard it's the lowest-resistance connection - but I doubt that answer because the Killacycle pack has spot welds. Is it for easier pack manufacturability and to reduce the risk of cell damage?

Yes, do not solder to the cells directly. Though there is some heat stress applied to the cell by tab soldering, hobby users have no other means of interconnecting them.

-Why the .01" thick nickel? I've seen others use .003" thick for batteries, but am not sure how to choose the thickness.

The nickel was scaled to meet the needs of this particular EV project. The thinner materials could not pass adequate power without losing some as heat. Using heavier nickel requires a much more powerful welder to achieve an acceptable weld nugget bond.

-Did you make a jig for easier pack construction? Not sure if this is necesary.

Actually, these ebike packs were simply glued together with RTV, and allowed to fully cure (with a large rubber band around them for tensioning) before I began the welding sequences.
The full size EV packs do require a jig to secure the cells in individual clusters until the collector plates are attached.


Regards, Jeff
 
@Jeff's schematic: I think the schematic has a typo! Q1's symbol is reversed from what it should be. As it's wired now Q1 is reverse biased and is unlikely to ever conduct. Also the schematic doesn't show where the positive terminal of the monitored a123 cell is connected.

The link posted to the starting point for this circuit (http://electronicdesign.com/Files/29/1155/Figure_01.gif) Is a nice simple circuit. Did you ever try eliminating R2 and then testing how "sharp" it's voltage cut-off was? (R2 in this case reduces the gain of the above circuit) If eliminating R2 provides satisfactory performance, then Q1 should be able to be replaced with a darlington, R3 shorted, and Q2 removed. Adding a capacitor between the positive and reference pin on the adjustable Zener should suppress any oscillations.

That's all for now,
Marty
 
Back
Top