Used 18650s for ebike, have I made reasonable assumptions?

[GalFisk]

Hi everyone! Electronics hobbyist here, but new to 18650s (and Li-ion in general).

I'm planning to build an 18650 ebike battery to replace the crappy 3s 12V SLAs I'm currently using (used bike, bought cheaply). I have measured the current draw from the current batteries to 13A at stall (brushed motor, so I'm guessing the controller is unsophisticated and stall current is max current).
I have 120 salvaged 18650s with reasonable capacity, no overheating, and low leakage. They consist of 6 different types. I want to build a 10s12p pack out of these, because I have enough space in the casing and because it will still be lighter than the SLAs.

The capacities range from ~1450 to 2500 mAh. The reason I want to include the low mAh cells is that they are from a power tool battery (original capacity 1500 mAh/cell) and reasonably new.

In order to get a reasonable balance in internal resistances, I arranged all cell groups so that they got an equal number of each type of cell, as closely as possible. Every group got three red cells, three green, two blue, one power tool, one light blue, and two mixed (I've checked that battery type and color corresponds).

In order to get every group as close as possible in capacity, I made a spreadsheet of every cells capacity, that listed cell types vertically and cell groups horizontally. I then calculated the average capacity of all groups, and made a matrix showing which cells of the same type to swap between high and low groups in order to get them as close to the average as possible. Using this method, I got the groups within 12 mAh of one another, which is less than a tenth of a percent from the average of 24.4 Ah.

I have not made any absolute measurements of the internal resistances of the batteries, but I've ran them all through a Liitokala lii-500 for capacity testing. The resistance testing on this charger is not reliable, but it gave ballpark figures, and no cell that worked fine showed unusual values when it came to internal resistance.

So my question is, how do you think this battery will fare?
I made an assumption that cells of the same type would have their internal resistance rise proportional to the wear, and also have their capacity drop in the same proportional way, so that in assuring that all groups have about the same numbers of each type, and the same total capacity, they will have reasonably matched inner resistances. Is that a safe assumption?
I also assumed that the voltage drop due to internal resistance will be low to negligible when loading the cells well below 1C. Is that a safe assumption to make, without having measured the actual internal resistance? Or to reformulate, is it safe to assume that it is unproblematic to load a used, tested 18650 below 1C?
Can I further assume that cells that differ in both capacity and internal resistance will work well in parallel at this load level? I'm thinking that the power tool cells may be drained a little quicker, since they have both lower internal resistance and lower capacity, but unless I plan on draining the pack very low, I think that won't be a big issue.

I plan on using a BMS and/or other battery supervision circuit, in order to keep the groups in balance and see if any of them start to develop issues. What electronics do you recommend? I read about a bluetooth BMS somewhere here, where you could get lots of info to your phone; is that thing accurate, useful, cost effective and shippable to the EU? If not, what would you recommend?

 

[titusmc]

Did you monitor temp as they were discharging (when you tested capacity)? Even a back of the hand test should have let you know whether you feel comfortable using them. Back when I used recycled cells, I always considered cells that were excessively hot to the touch as not usable for ebike application.

If you did and they all ran relatively cool (warm but not hot), then you should be okay with your pack size using a BMS. I used to run a hub motor with a 20A controller using a 13S7P recycled pack - as long as you limit your riding behavior to keep the current draw less than 1A per cell (except for short peaks), it will be fine.

Big question now is how do you plan to build your pack? I would highly recommend an Arduino spot welder kit and a car battery as opposed to soldering. I began with the solder wick method (or tinned copper braid) and switched to spot welding about 6 months ago. Spot welding produces much cleaner, consistent results and in my experience the packs are more reliable (I have had a few soldered packs fail - 0V cell group - even with brand new cells).

Good luck. I'm sure I've missed things and others will chime in.

 

[GalFisk]

Thanks for your reply.

I charged all cells at 1000 mA and discharged them at 500 mA using the nor test (charge-discharge-charge). I monitored the temperature by hand during the first charging, and all cells with heat issues got unpleasantly hot during CV if not before, while the ones that worked fine only got lukewarm during the entire cycle.

It would be nice to be able to disassemble the pack and swap out the cells, both because some cells are closer to their end of life than others, and because I acquire new, potentially better cells whenever a laptop battery dies at work. The Vruzend kit looks interesting, and okay for my limited current demands, but I'd also enjoy designing and 3D-printing something of my own design using springs or compressed foam to keep contact.
If I do something spring-loaded, the springs will not carry the current. I've read the thread about the foam pack and know about selecting the right material. In those cases I'll probably use copper or brass braid as a current carrier. An advantage with braids is that they will make good contact even if I don't get rid of the spot welding burrs on the cells completely. A a potential worry is that the burrs, or springs, wear down the braid so bits get loose and potentially make shorts.

If I don't build a disassemblable pack, I will probably build a spot welder or resistance soldering setup. I have several MOTs in my junk bin that I could rewire. I'm somewhat partial to soldering because it's easier to disassemble, but I don't want to use a regular iron and spend a lot of time (and heat) on each cell. I did see a video where a guy using a soldering iron had to dunk a cell in ice water because solder had wicked down along the button and made contact with the can.

I don't have an Arduino, but this may be a good excuse for buying one and learning to code for it. Or I could build a 555 timer and dial in the weld or solder time on a trimpot, using junk cells as victims.

Edit: I see that the milliseconds are important when spot welding, so if I go that route the timer will most definitely be digital.

Another question: would anyone be interested in my battery sorting/capacity balancing spreadsheet, or maybe a software program built on the same principle? I won't put a lot of effort into it just for my own use, but it could be fun to flesh it out for others.

 

[titusmc]

I currently use this, combined with an Optima Yellow Top: http://www.instructables.com/id/DIY-Arduino-Battery-Spot-Welder/

 

[GalFisk]

Update: I have now assembled the battery. I ended up using the Vruzend kit, as I got three largely unused kits for the price of one from a seller at Electric Skateboard Builders. It is 10s8p and fits very snugly inside the old SLA enclosure. I could possibly have made it 10s9p, but then I would've needed to redo the build after I'd already started. Also, I possibly would've hade too little space for the BMS, which I haven't received yet.

I had to redo the cell sorting in order to go from 10s to 8s, and also to add some better cells I'd acquired, and remove the weakest ones. Now all cells range in capacity from 2197 to 2452 mAh, with the majority being 23xx. The entire battery is 18515 mAh, and the highest and lowest groups differ by 8 mAh. I feel comfortable with using this without a BMS for a few days. All groups are sitting at 4.15-4.18V. Max current draw will be 13A.

The battery case sits vertically in the bicycle, so I'll pad the bottom where the weight of the battery rests, with some closed cell foam. Maybe I'll also wrap some light foam around the entire pack, if there's space.

A photo of the pack itself: https://imgur.com/a/W3Noi

My experiences with the Vruzend kit:
The easiest way of assembling the kit, is to attach one of the bottom caps to its neighbors, then pushing the cell down into it, and finally attaching the top cap and pushing it down so it's flush with the neigbours. This way you only have to deal with two pairs of tabs and slots at a time, and all the pushing can be done by hand.
I measured the voltage between the screws on each cell before attaching the bus bars, in order to check that the contact and polarity was okay.
I had issues with four of the caps, where the screw inside started rotating as I screwed on the nut. This was very anoying as it would only be discovered after it was far too late to remove only this one cell. I managed to get the screws on reasonably tight anyway, and left it at that.
Being sloppy and having one of the tab sides not mate with the slots was annoying, as it was usually not discovered until it was time to build the next row, and I had to push on the entire pack to get them to mate. Fortunately it's possible to push them in from the side.
Screwing more than three bus bars onto a single post is difficult, and sometimes impossible depending on the exact thread lengh. Thus the depicted layout. It always has at least six bus bars connecting one group to the next.

My suggestions for the Vruzend kit:
Taper the tabs and slots slightly near the end, this will make mating much easier. Make it backwards compatible with the existing design.
Do something to make it impossible for the screws to rotate. Maybe squeeze the metal together around it hard enough to flatten the screw?
An optional spring fastened inside the sheet metal contact would enable greater contact pressure for a higher cost. Most of the current would not pass through the spring.
Make a hexagonal verison for packing the cells even tighter. If you put tabs on three adjacent sides and slot on the three others, assembly will be nearly as simple as it is now.

Edit: I've mounted the battery (sans BMS, for now) in the ebike, and it works great.
Edit again: my pack has 666 Wh