Upload Your Circuit Schematics & Requests Here...

[GGoodrum]

I'm doing a 10-cell version of the circuit Bob posted here, in this thread. I've already got enough parts for about 50 of these and I should have the boards by Wednesday. Here's what the board layout looks like:

Two standard RC-style balance plugs are connected to "A" and "B", and a two-conductor cable that runs up to the controller connects at the top of the board.

-- Gary

 

[bobmcree]

nice job, gary.

folks should note that they can use the 2.1v devices and optionally tack on a pair of resistors to move the trigger point upward or use the 2.7v parts for the higher cutoff voltage. unfortunately there is not a part in between in this series. they are laser trimmed at the factory and if somebody wanted to produce a bms they could get any voltage they like, and might want to use the surface mount versions of the parts to make the units even smaller.

the leaded parts i suggested can also be handwired to each cell, with just a 2 conductor cable connecting the output of the optos together. using the standard bms connector arrangement gary is implementing should let the a123 dewalt pack plug right in to the bms board with the right cable, and that is certainly simpler when using 10 cell subsystems.

 

[bobmcree]

One of the unique characteristics of a123 cells, over other LiFe and Li-Ion chemistries is that they can be charged at very high rates. I have personally charged single-p packs at 20A rates and they barely get warm. What we really need is an intelligent charger/balancer design that can individually charge a block of cells at a high rate, say at least 20A. You'd be able to do a 4p pack in about a half an hour.

-- Gary

i've considered this and wondered how well balanced a parallel set of cells would stay when charging at very high current. it seems like if i have 4 cells in parallel and i pump 20a through them one cell could easily hog current and overcharge before the others had a chance to balance it out naturally.

have you charged your parallel packs at higher currents without any problems? this issue is the main reason i am arranging my packs in series strings first then combining them at the ends so i can run balanced current through each string of cells during charge and monitor each cell for low voltage independently in my 120 cell pack.

 

[Jozzer]

Before I built my pack I did a quick test along these lines, with a 1s3p pack. THe voltages of all three cells stayed the same throughout charging (at 18A). I took that to mean that all 3 cells were getting thir share of current. After charge I snipped the balance leads and discharged each cell seperatly, with all cells managing within 20mah of each other. (the same criteria I had used to choose which cells to use in the test).
I didn't try charging any faster than that, since I knew I would never have a charger powerfull enough to charge over 3c.

 

[GGoodrum]

If the cells in the packs are connected in parallel first, and then the paralleled blocks in series, all the cells in each block are always at the same voltage, even during charging.

I have a Zivan NG1 48V SLA charger that has been "tweaked" (via a could of trim pots inside...) to max out at 18A. I have charged my 16s2p and 16s4p packs with this charger a lot, and the packs are fine. It doesn't just dump 18A into the pack until the cutoff, though. It actually has a two-stage constant current (CC) mode, before switching to the constant voltage (CV) mode to finish up. If the pack voltage is down below about what works out to 70-75% capacity, it starts charging at the max 18A rate. Once the pack voltage rises above that 70-75% point, it drops the charge current to 6A, and stays there until the cutoff voltage is hit, which I have adjusted with the trim pots to be 58.4V, which is exactly what a 16s a123/LiFe pack needs. It then switches to the CV mode, and the charge rate slowly drops off. When it gets to about .25V, the charge is basically complete, and the packs are basically as full as they are going to get. The charger, though, then goes a pulse mode, where it will hit the pack with about .25A for a few seconds, and then it waits a few, and repeats this process for awhile. After about 10 minutes of this, the green light finally comes on and a buzzer goes off. This is very similar to the 3rd phase in the AstroFlight RC chargers, which also uses a PWM-type method, instead of the more "normal" CC/CV modes that most chargers use. In any case, I think this pulse mode at the end is used by the NG1 to balance SLA cells. If I connect balancers to the packs I'm charging, it also works quite well with a123 setups.

In any case, what would be good is charger that did 20A indivdually, but with either a CC/CV profile/function, or by doing the same sort of pulsed mode the AF chargers use to "top off" the cells. Yesterday I ordered 10 of these: https://s.p10.hostingprod.com/@www.voltphreaks.com/ssl/catalog/product_info.php?products_id=28. They are single cell LiFe chargers, good for 2A each. I am going to wire these into a set of connectors that will plug into the balancer plugs on my 20s4p packs. Using a couple of powerstrips, this will be mainly as a convenient "travel" charger. I don't know how well this would scale up to 20A, but that would be the idea.

-- Gary

 

[bobmcree]

it's great to have the benefit of your experience, guys. what you are saying that things work fine with parallel groups of cells as long as all the cells are ok. but what happens when a cell fails?

if you have 4 cells in parallel the voltage across them will of course be the same even if the internal impedance of the cells is different and they are charging at different currents. you could not see this without disconnecting the parallel connection, or connecting them together through resistors.

my major concern is a shorted cell, the secondary concern is the difficulty in isolating a cell that is not providing rated output. i fear that if a cell shorts it then will draw down its parallel partners and damage them, if nothing worse. i also want to be able to tell which cell fails when the pack drops out 20% below rated output because one cell in a 5p cluster is not holding a charge.

i think what you guys are seeing is that for hobbyist users like us these kinds of failures occur seldom enough that they are manageable, but i am not sure that translates into turning the bike with parallel connected cells over to someone else to use and keep running who is not a techie.

what kind of hours/cycles do the guys running multiple parallel cell a123 packs have without any problems? this does not include using the 10s packs in parallel, just people using clusters of single cells that are then connected in series?

 

[rf]

if you have 4 cells in parallel the voltage across them will of course be the same even if the internal impedance of the cells is different and they are charging at different currents. you could not see this without disconnecting the parallel connection, or connecting them together through resistors.

my major concern is a shorted cell, the secondary concern is the difficulty in isolating a cell that is not providing rated output. i fear that if a cell shorts it then will draw down its parallel partners and damage them, if nothing worse. i also want to be able to tell which cell fails when the pack drops out 20% below rated output because one cell in a 5p cluster is not holding a charge. ...

After listening to folks here. on other forums, and the A123 guys, It seems there are many advantages to building packs from `super cells' -- two, three or four cells directly paralleled together to become one. Lots of advantages:

• * Fewer balancing and cutoff-monitoring lines and associated electronics.
  * Protects weak cells.
  * Increases output for packs of unmatched cells.

Protection against undervoltage and reversal is the biggest item -- since it's the Achilles heel of A123 cells. Fewer cells in series means less chance of undervoltage/reversal. My electronically challenged brain says a single pack with `Super Cells' is more robust than a multi-pack array achieving the same voltage/current. Just as R/C modellers seldom seem to need LVC monitoring for their smaller series packs (just balance-charge and go.)

My guess is that a 10s pack made of 4x SuperCells will be much more likely to survive without LVC than the equivalent four packs with single cells. Even if one of the embedded cells is damaged/half-capacity. (I've done more guestimating than actual numbers so far. And I'm not advocating leaving off the LVC.)

One cell in a quartet dying and taking the others with it seems extremely unlikely. A weak cell being nicely propped up and equalized by the others seems much more likely.

Richard

 

[rf]

what kind of hours/cycles do the guys running multiple parallel cell a123 packs have without any problems? this does not include using the 10s packs in parallel, just people using clusters of single cells that are then connected in series?

I've seen several folks online paralleling Dewalt packs to use a single charger for two or more packs and keeping them together for discharge (one BMS per array.) One guy is selling connectors to join packs via their balancing connectors (which seems slightly funky for discharge, considering the light wires involved.) So far no complaints of bad cells and no LVC.

http://readytoride.biz/?p=105#comment-5332

I have a better link somewhere showing construction.

(Very cool on those boards, Gary!)

Richard

 

[GGoodrum]

I can definitely attest to the fact that packs are even more robust when cells are paralleld before they are connected in series. I have one setup that has a bunch of sub-packs, the biggest of which are two 6s5p and one 4s5p. These are connected in one string to make a 16s5p configuration. I had a 2s5p pack that I connected in series, to give a final configuration of 18s5p. I had some sort of short in the wiring harness which caused a dead short across the 16s5p string, causing it to melt plastic to the point that some of it caught fire. It took me over 10 seconds to get the fire blown out and the harness pulled out to break the connection. My first thought was that the sub-packs were probably toast, but to my amazement, the packs/cells are all just fine. They didn't even get knocked out of balance.

-- Gary

 

[Jozzer]

it's great to have the benefit of your experience, guys. what you are saying that things work fine with parallel groups of cells as long as all the cells are ok. but what happens when a cell fails?

if you have 4 cells in parallel the voltage across them will of course be the same even if the internal impedance of the cells is different and they are charging at different currents. you could not see this without disconnecting the parallel connection, or connecting them together through resistors.

my major concern is a shorted cell, the secondary concern is the difficulty in isolating a cell that is not providing rated output. i fear that if a cell shorts it then will draw down its parallel partners and damage them, if nothing worse. i also want to be able to tell which cell fails when the pack drops out 20% below rated output because one cell in a 5p cluster is not holding a charge.

i think what you guys are seeing is that for hobbyist users like us these kinds of failures occur seldom enough that they are manageable, but i am not sure that translates into turning the bike with parallel connected cells over to someone else to use and keep running who is not a techie.

what kind of hours/cycles do the guys running multiple parallel cell a123 packs have without any problems? this does not include using the 10s packs in parallel, just people using clusters of single cells that are then connected in series?

Only up to 80 odd cycles so far. I've had one "incident" that nearly cost me some cells, I made the mistake of powering my series/parr relay from a 12v tap on the pack. I thought it wouldn't be a problem, as the relay is only activated during charging, and the power rating of the coil is so low, but after some 30 cycles I noticed the pack voltage take a dive when the pack should have been only half discharged. I discovered that the 4 cells feeding the relay were all around 2v, whilst all the others were still evenly balanced at 3.3v. I split the pack and recharged and tested the 16 affected cells (4p) individually (took all night!) to find that all had survived and still held full capacity. Lesson learnt. A LVC would have been VERY usefull in this scenario. Had I not been paying attention I might not have noticed the voltage drop, and had I pushed the cells for another few mins they might have been damaged.
If a cell does fail or lose capacity, then the paralell string with the drasticly lowered voltage would be easy to find, and with your LVC implimented properly then the rest of the cells should be protected. To find the particular cell would require testing each individually. Not a job for the end-user, but I believe rarely costing more than a bit of labour testing and an occasional cell to be replaced.
I guess in the event of a shorted cell the entire string IS at risk, but presumeably only large numbers of diodes or carefully rated fuses between each cell could save this from happening?

BTW Bob, remember I have only just about enough knowledge to make myself dangerous when you read my words

 

[bobmcree]

a shorted cell is my main concern. to protect against that causing damage it takes only one diode at the positive end of each serial string that is combined together, for a loss there of .5v. if i have 120 cells in 8 parallel strings combined with 8 diodes and a cell shorts, that string will just stop contributing and my charger will notice it on the next cycle if i do not notice that i lost 1/8 of my capacity. i can charge each of the 8 serial strings separately so i can realistically charge the whole pack in under an hour.

if i have clusters of 8 cells and one shorts its 7 partners dump all their power into it and likely damage occurs. clearly it is safer to connect them in series first, but then a low voltage detector is needed on each cell to do it right, and that does make things more complicated.

it will just take some time to learn if using the cells in parallel clusters provides the same service life as connected serial strings.

 

[GGoodrum]

Here's what my new 10s4p a123 packs look like:

Basically two 10-cell packs, pretty much just like they come out of the DeWalt cases, are connected in parallel and then a balancing harness connects the individual cells in a 10s2p parallel configuration. A second set of 10-cell packs are connected in a similar fashion but the harness has a set of the matching connectors. What this allows me to do is to independently charge/balance each 10s2p block, if in the unlikely event that they become unbalanced. Normally, I will connect the two pairs together in parallel, in a 10s4p configuration, by having a connector harness that wires the two main outputs togethre, and by pluging one set of balancer plugs into the matching plugs on the other two. That way the one set of balancing connectors are then connected to all four packs, in a true 10s4p configuration. It is these connectors that will be used by the LVC circuit, and will also be used for balancing. Also shown in the last picture is a TP-210V 10-cell auto balancer.

This pack weighs 7 pounds 9 ounces and is 2.5" x 6" x 11".

-- Gary

 

[bobmcree]

nice job gary, and i like the idea of using the serial strings that are then connected in parallel; it is certainly more complicated, but if the pack can be expected to last 2000+ cycles then it is worthwhile to invest a bit more in the bms to make it easier to pick out a bad cell a couple of years from now.

your pcb will make this a nice setup. i think the major difference between our systems is you use connectors to separate the strings and i combine them with diodes; a minor difference that is just mainly a matter of choice. the diodes do cost a bit of power, but i use them on a couple of my bikes so that i can just plug in any battery in parallel with any other and the higher voltage one will run the bike.

i run 15 cell serial packs as that is the voltage i need to cruise comfortably at the legal limit and then some. the 15s a123 pack produces about 49.5v and supports my 48v NiMH pack through a pair of diodes very well. the a123 runs the bike until it drops down below the 40 cell NiMH pack voltage and then the higher voltage battery just takes over the load. The packs share the load well so that the a123 pack still provides some high current drain voltage support even after it stops powering the bike at normal half throttle.

 

[rf]

a shorted cell is my main concern. to protect against that causing damage it takes only one diode at the positive end of each serial string that is combined together, for a loss there of .5v. ...

Can you suggest a good diode, Bob? Mine aren't nearly that good.


Richard

 

[Jozzer]

Doesn't connecting each cell in paralel first help protect against 1 cell taking the brunt of an imbalance though? As Gary said earlier, all his cell failures have been on single p strings..
I didn't fancy using the dewalt packaging, and would prefer to build the lvc into the pack directly. (my pack wouldn't have fit on my bike for instance with the extra packaging).

 

[GGoodrum]

Jozzer is right, connecting the packs together with the balancing connectors does parallel all the cells together. In the 10s4p pack shown above, with the 2nd set of balancer plugs connected to the 1st set, you end up with 10 blocks of 4 paralleled cells.

He is also right that that the only problem I've had killing cells has been with single string packs. There are only two ways I've found that you can kill these things, one is applying too much heat to them with a soldering iron, or a heat gun (applying shrink wrap...), and the other is to over-discharge them. I am convinced the way to get long life out of these cells in ebike applications is to first parallel as many cells as possible, and then have some sort of LVC function that will prevent you from over-discharging the packs. You really don't need anything more than that.

In my setup, I plan to normally use two of these 10s4p packs in series, each one with one of the LVC boards. To charge each pack, I will unplug the LVC boards and use the balancing plugs with the balancer and charger.

-- Gary

 

[Jozzer]

I am convinced the way to get long life out of these cells in ebike applications is to first parallel as many cells as possible, and then have some sort of LVC function that will prevent you from over-discharging the packs. You really don't need anything more than that.

I agree, as long as balancing is taken care of (either each charge, periodically, or just when the pack capacity drops due to a low bank.)
So far, I can only balance manually, if I am to sell packs like this on e-bikes I need an easy balance solution too (hopefully part of the LVC circuit and so scalable through the many applications I have planned!)
I dont feel I can achieve this with RC balancers TBH..

 

[bobmcree]

a shorted cell is my main concern. to protect against that causing damage it takes only one diode at the positive end of each serial string that is combined together, for a loss there of .5v. ...

Can you suggest a good diode, Bob? Mine aren't nearly that good.


Richard

to combine 2 batteries i use a 40CPQ060 which is a common cathode dual schottky 40A 60V part with a forward voltage of .53V. you can use the 40CPQ100 for 100V with a bit higher forward voltage. they are $4-$5 in small qty.

dual schottky diodes seem to be generally a better deal and more available. mounting them to some kind of heat sink is a good idea if the current will be over 20A for long periods.

 

[rf]

Aloha,

I've been brainstorming about balancers and searching for parts.
I've ordered another microcontroller that looks good: ATmega168,
$4 quantity one. It has 23 I/O lines and an 8-channel 10-bit A/D.
It's reasonably fast and appears to have enough space to hold a fairly
large program. Still looking at multiplexers and power MOSFETs.

The plan so far is to charge or discharge one cell at a time. More if
that seems too slow. Balancing by adjustment in just one direction
doesn't seem sufficient. Providing for both doesn't seem that hard.

With 8-channel analog multiplexers available for 50 cents a piece it
should be possible to monitor up to 64 cells with the 8-channel A/D on
the microcontroller. Looking at 12 or 24 cells initially.

A power MOSFET per cell seems workable to provide charge voltage or
discharge load. (~$1 each)

Is there a better way to do this? I'm a programmer not so much
a hardware hacker (yet.) The controller is easy. All that funky
hardware is another matter. Though I've built lots of kits and stuff.
Even programmed some microcontrollers for Reagan's SDI many years ago.
They've gotten much easier to use and cheaper since then.

Any ideas or observations?


Richard

 

[fechter]

A power MOSFET per cell seems workable to provide charge voltage or
discharge load. (~$1 each)

Most balancers I've seen have one fairly small FET and a big resistor across each cell that will turn on when you reach the upper voltage limit on that cell. The larger the current capacity of these parts, the higher the charge rate you could run and still maintain balance.

During discharge, there's one big FET that passes all the current and will open when any cell reaches the minimum voltage limit. There may also be a current limiter circuit that will open the same FET if the current tries to exceed the maximum allowable.

I sort of don't like the idea of having anything other than a fuse between the batteries and the controller. If a separate LVC / overcurrent protect FET is used, it just needs to be a really big one (several in parallel).

You could avoid needing the series FET if the controller was tied in to kill the throttle when either the low voltage or high current limit was exceeded. Most controllers already have a current limiter, so this seems redundant. If there's a fault, the fuse should handle things.

 

[GGoodrum]

Most of the RC standalone balancers work exactly how Richard descibes, using a resistive load to dissapate 150ma per cell fo each of the cells that are above the lowest one. This is how the TP-210V and Astro "Blinky" work. The more expensive Schulze LiPoProfiBalancer series (http://www.schulze-elektronik-gmbh.de/index_uk.htm reroutes current from the higher voltage cells back into the ones with lower voltages.

With a123-based packs, current limiting isn't required. You can't pull enough current out of these to hurt the cells unless you have a dead short. Even then, as I found out, the wires/harness will melt before you will hurt the cells. All that is needed is a big fuse. I also agree that the LVC implementation that trips the brake inhibit line will work fine for setups that use Clyte controllers. For other setups, adding a big FET to provide absolute power cutoff would be desirable, I think.

What I really want to do is to eliminate having to bring out the balancing wires/plugs. To do that I need to not only move the LVC circuit to inside the pack, but I also need some sort of balancing circuit. Basically, all that is needed is to have some way to "hold" the voltage at the 3.65V level for each cell. Then, an external CC/CV charger can handle the the gradual reduction in the charge current (i.e. -- CV mode...). Adding the big FET to provide a positive LVC cutoff would make this a complete a123-tailored BMS solution.

-- Gary

 

[GGoodrum]

Here's another version of the 10-cell a123 LVC board I'm doing:

This one eliminates the need for the cross-connected balancer harness that I had to make in order to properly connect the two DeWalt end cap conectors with the RC-style balancer plugs. That was a major pain. Now I just cut the plugs off the DeWalt BMS and solder them to this board. Here's what the current setup looks like, with the harnesses:

Two of these boards will be used, but only one set will be populated with the LVC parts. The balancer leads from the 2nd board will connect into the 1st board. The end result is a 10s4p pack, with the embedded LVC functionality.

-- Gary

 

[Doctorbass]

Great work!

 

[Jozzer]

Indeed Gary, Great work.

Whats the actual size of the board to be?

 

[GGoodrum]

Thanks.

The board is about 2-1/2" wide and 7" long.

-- Gary