Quick sort LiFePo4

[Taiden]

Hi all,

Is there an effective way to quickly sort almost two hundred LiFePo4 18650 cells into sub-groups for a 16s10p battery pack?

I suspect that the parallel groups will self balance... but I'd like to sort out the worst of the bunch and keep them as spares.

 

[Matt Gruber]

i would put them, 1 at a time, on a load of 1/2 MAXIMUM rated C. for whatever precise time you are willing to spend x 200. i'm thinking 1 minute each.
on each cell write the V at end. this would be for new cells. looking for excessive IR, cells that get too warm or have lowest V get eliminated. Since i've never tested LiFePo4, guys will chime in and tell us why this won't help eliminate to weak ones, or will suggest something better.

 

[Taiden]

Thanks matt! Just to be clear, this is for the used cells from this ad: http://endless-sphere.com/forums/viewtopic.php?t=59646

Datasheet: http://www.valence.com/sites/default/files/18650_cell_datasheet_may_2013.pdf
"174 Valence LiFePo4 cells (around 700 watt-hours)
Last member had them for about 200 charge cycles."

 

[dogman dan]

I agree, sort by internal resistance, how to make it fast beats me.

 

[Taiden]

I had an idea, maybe it's crazy but hear me out.

What if you put all cells in a single parallel configuration, and then charged them, and let them sit and balance a little (not sure about this, time, feasibility, etc) before recording voltage of each cell.
Then discharge the entire pack and check voltage for each cell immediately afterwards. (Maybe they would self balance fast enough during discharge to not make this a good idea?)
Use delta V to sort.

 

[Matt Gruber]

could do 30-36 in series. line up against wall, box in with 1x2. hook to 120v 100w light, then go from cell to cell and X ones that get low. write time on each cell. keep best, do next group. stop test before any get too low, unless u will throw away anyway.

 

[Taiden]

I like that. I assume they'd have to be charged and balanced before you begin.

 

[Taiden]

Newbie question, electrics are not my forte. (Mechanical engineering and manufacturing background, but never actually spent time with real-life electrics. Sometimes you have to run before you walk because walking is just too damn slow.)

I assume I can use a DC power supply with adjustable voltage and current to do all my charging (matching datasheet specs for voltage and current) until I can get an appropriate charger. Truth?

 

[John in CR]

A major, if not the primary, sort criteria of A123 is based on self-discharge. Charge them in parallel to an identical volage and let them sit for a week or two. Those with the lowest voltage are the weakest. If you have the ability and time to sort them by capacity, then do that too, and use the best of the bunch.

I'm not sure what you meant by "self balancing" in your initial post. Parallel cells will have identical voltages, which may vary slightly while under load due to differences in IR, but at rest for any reasonable time their voltage will be the same. For balance purposes simply treat paralleled cells as a single large cell, so you will have 16 cells to compare for balance and to check for LVC with your BMS.

 

[Taiden]

I just meant that in parallel any voltage differences in each 18650 cell would cause current to pass from higher voltage cells to lower voltage cells thus self balancing the parallel group.

So lets say I get a nice gradient of sort from these 18650 cells. I figure I'd exclude the lower ones, and then group the others by discharge. So the top 95th percentile would be one parallel group, the remaining 90th percentile would be another parallel group and so on and so forth.

The other option would be to mix them so each parallel group had on average the same characteristic with which we sorted, but I suspect that consistency between cells is more important than averaged consistency between parallel groups.

 

[Matt Gruber]

your QUICK SORT time is OVER.
now that you see how hopeless this is, you understand why i bought 2 new dewalt packs, and was riding my dirt bike 1 hour later.

 

[Tench]

Build yourself a little wooden cell holder board that can hold 10 cells wired to an 10cell icharger, fill it with cells and read the ir, sort them into groups, any with extremely low voltage put in sub groups to receive further attention to establish if they are any good. Then see if you have enough within a close tolerance to build a pack.
There is no quick simple or definitive way to establish the condition of the cells, time and testing will be needed. And before you can assemble a pack they will all need to be charged to the same voltage.

 

[Taiden]

The whole exercise seems almost pointless in all honesty. Someone who knows more than me about this stuff said I should sort it. But is that like saying you should change your oil at 3000 miles when in reality 4000 miles isn't going to really make a lick of difference?

 

[TheBeastie]

One dodgey and somewhat impractical way to do it is to use a BC168 / Una9 balance charger and hook up each cell and watch them charge.
I have noticed that the dodgiest cells always report a higher voltage during charge, its kind of the opposite to when you discharge them as the weakest cells always report the lowest voltage under load.
http://www.myrcmart.com/unrc-una9-plus-battery-lipo-balance-charger-parallel-charging-p-5502.html
So to rephrase you could put them on charge and take note of the highest voltages during charge and mark them as crappy, all cells would probably need to be balanced first.