N o a h ?

Assembling a 4s LFP battery

With a func­tion­ing spot welder in hand, I set out to build a bat­tery with Lithium Werks ANR26650M1B cells. These are LFP cells nom­i­nally at 3.3 V with a capac­ity of 2500 A·h.

Design

Since this bat­tery is going into a 1:6 scale RC trac­tor, I have some oper­a­tional con­straints. First, I’m going to try to fool a Hob­by­wing QuicRun 1080 G2 ESC into believ­ing it’s run­ning with power from a LiPo. This ESC sup­ports 2s and 3s LiPo bat­ter­ies, which means it should oper­ate between 6.4 V and 12.6 V. A 3s LFP con­fig­u­ra­tion will give us a usable range of 7.5 V to 11 V; for 4s, 10 V to 14.8 V.

In gen­eral, a 4s LFP bat­tery is going to be more use­ful than a 3s one, sim­ply because 12 V nom­i­nal cir­cuits are more com­mon. I sus­pect I can get away with a 4s pack here as the elec­tron­ics in the ESC should be rel­a­tively sim­ple. I hope the worst case is that I might fry the BEC com­po­nents, which would force me to use an exter­nal BEC instead.

The other con­straint is size. This bat­tery has to fit under a hood that is only 77 mm wide on a plat­form that lim­its its height to about 45 mm, so there isn’t a ton of room for wires or even the BMS. Ini­tially, I wanted to have two cells on top of two cells (so the axial ori­en­ta­tion would look like a square), but since I had more length than height to work with, I opted to make it inline:

Safety

Since we’re using a BMS, our first step is to syn­chro­nize the cell volt­ages. A BMS won’t work at all if the volt­ages of indi­vid­ual cells are too far apart: it assumes some of the cells have been dam­aged. Since cells ship at a low SoC, I gen­er­ally just pick a volt­age above their cur­rent charge level and let them all charge up to that point. Then I dou­ble check every­thing with a volt­meter.

I won’t do a full charge until they’re in the bat­tery as I’d rather spot weld mostly dis­charged cells.

Two mate­ri­als are usu­ally used to limit the risk of short cir­cuits: adhe­sive-backed fish­pa­per and Kap­ton tape. We use the thicker fish­pa­per for sep­a­rat­ing elec­trodes and Kap­ton tape to hold the fish­pa­per (which seems to have a rather weak adhe­sive) and other elec­tri­cally iso­lated com­po­nents in place.

For most cylin­dri­cal cells, the entire hous­ing is the anode, and the cath­ode pro­trudes on one side, sep­a­rated by a thin layer of plas­tic insu­la­tion. (For the ANR26650M1B, the hous­ing is the cath­ode, but it is oth­er­wise struc­turally sim­i­lar.) To pro­tect the plas­tic insu­la­tion, we use a fish­pa­per gas­ket matched to the cell diam­e­ter.

Like­wise, we want to pro­tect each par­al­lel group from each other. To do that, we wrap the hous­ings of the entire group in fish­pa­per after assem­bling them, leav­ing only the elec­trodes exposed. Once the groups are assem­bled in series, we can addi­tion­ally wrap them, leav­ing only a small area exposed for attach­ing the bal­ance wires. With the 1p bat­tery I’m build­ing, I’m going to put a layer of fish­pa­per over each cell.

We’ll also use fish­pa­per any­where wires cross each other, around the area the BMS over­lays the cells, and when­ever nickel strips or wire are near cell hous­ings.

To aid in heat dis­si­pa­tion and reduce the risk of a sin­gle mal­func­tion prop­a­gat­ing to adja­cent cells, it’s usu­ally a good idea to phys­i­cally sep­a­rate cells from each other unless size con­straints pre­vent it. You can buy gen­eral-pur­pose plas­tic spac­ers, but I decided to 3D print a model spe­cific to this bat­tery. It has raised areas between each series con­nec­tion. These serve both as addi­tional short-cir­cuit pro­tec­tions and as build guides. The over­all width of my spac­ers is just over 30 mm, which means I can lay my nickel strips directly on top of the cells with­out wor­ry­ing about wrap­ping them around the cell hous­ings.

Finally, many peo­ple like to round off the (poten­tially quite sharp) edges of the nickel strips before weld­ing them to pre­vent them from poten­tially dig­ging into the fish­pa­per and even­tu­ally short­ing the cell hous­ing. This isn’t nec­es­sary with my cell spacer design.

Preparing the nickel strips and lead wires

I had a really, really hard time fig­ur­ing out how to attach the bat­tery leads, which are 10 AWG wires, to the nickel strips. Even though I der­ated my over­all design to 25 A, I have a 60 A BMS: bet­ter safe than sorry for every­thing in front of it, I guess. Any­way, every­one just says “sol­der them”—well, eas­ier said than done!

I’m using a Hakko FX-888D, which has plenty of power (you’ll need it), a 3.2 mm chisel tip, high qual­ity eutec­tic sol­der, and RMA flux paste. I wouldn’t skimp on any of the mate­ri­als, per­son­ally. It’s hard to get enough heat into the chonky nickel strip.

I started by tin­ning the wire and the area of a “pad” I wanted to cre­ate on the nickel strip. The sol­der did not stick well to the nickel, even when cleaned with IPA. I had to scratch up the sur­face using some medium-grit sand­pa­per and apply a lib­eral amount of flux to the area. (I also used a lot of flux on the wire itself.)

To hold every­thing in place, I put the strip in my vise and aligned the wire using a help­ing hand. Even so, I found it very dif­fi­cult to get the sol­der to flow prop­erly. I prob­a­bly tried 5–10 dif­fer­ent approaches (vary­ing tem­per­a­ture, amount of flux, etc.) in this vein before I went back to the draw­ing board (that is, search­ing the inter­net for clues).

The most use­ful tip I found was to fold and crimp the nickel over the wire before sol­der­ing. In prac­tice, this made it dif­fi­cult to tell when the sol­der had filled the gaps inside the fold, and I also had trou­ble ori­ent­ing the iron to get con­sis­tent heat­ing, so I ended up with a slightly mod­i­fied pat­tern: mak­ing a semi­cir­cu­lar cup around the wire and fill­ing it in, sim­i­lar to how you’d attach a wire to any other sol­dered con­nec­tor. I lightly pressed the nickel into the wire with pli­ers. Then I was able to wedge the iron into the cup on one side and feed sol­der from the other.

My final rev­e­la­tion was to sig­nif­i­cantly increase the tem­per­a­ture for this step. I ini­tially tinned the wire and metal strip at around 375 °C, but I bumped my iron up to 415 °C to make the con­nec­tion. (And it still took an awk­wardly long time.)

I also tinned the loca­tions on the nickel strip where I plan to attach the BMS bal­ance wires. These wires are very small and shouldn’t take much time to sol­der, but I fig­ured I’d get every­thing ready as long as I had my iron out.

Most of the time, a BMS is inte­grated into either a bat­tery itself (“pro­tected”) or part of the con­trol cir­cuitry for the prod­uct that uses the bat­tery. I bought a DALY BMS to attach to the bat­tery, but I don’t really want it wrapped into the bat­tery itself or sol­dered directly to the bat­tery ter­mi­nals. In my mind, the fail­ure modes and lifes­pan for the cells and the BMS seem dis­tinct, so they should be sep­a­rate parts.

To con­nect the BMS to the neg­a­tive bat­tery lead, I’m using Amass XT150s, which are sin­gle-pole con­nec­tors that can han­dle up to 100 A con­tin­u­ously over 8 AWG wire. The DALY BMS also has ter­mi­nal hous­ings for its bal­ance and ther­mis­tor wires, so the whole unit should be swap­pable with only minor incon­ve­nience.

To con­nect to the load, I also opted to use XT150s, mostly for their sin­gle-pole design, which allows me to con­tinue to keep the BMS and cells phys­i­cally sep­a­rated. Also, with an XT150, I should be able to split the load side out into four 14 AWG wires (equiv­a­lent to about 8 AWG, using the total coss-sec­tional area of the wires) to con­nect to the ESC, charger, or any­thing else I want to power.

Power sources are always attached to the female con­nec­tor for safety rea­sons (it’s harder to acci­den­tally short a female con­nec­tor, which is usu­ally fully obscured by its hous­ing), so I went ahead and sol­dered them onto my pos­i­tive lead wire, neg­a­tive lead wire, and the load side of the BMS. I also sol­dered a male con­nec­tor to the bat­tery side of the BMS.

XT150 con­nec­tors can be a bit dif­fi­cult to manip­u­late. Here’s what I learned:

• The female con­nec­tor goes in the smaller hous­ing.
• It’s eas­i­est to push the hous­ing over the con­nec­tor while it’s still hot, within a minute or so after sol­der­ing.
• If the hous­ing gets stuck and won’t budge, I actu­ally had suc­cess spray­ing into it with a bit of IPA to tem­porar­ily lubri­cate it, as unin­tu­itive as that sounds.
• The hous­ing needs to click over the con­nec­tor. Until it clicks, it can still be pulled back with sig­nif­i­cant enough force.

Cell spacer improvements

With my cell spacer design, the 30 mm-wide nickel strips com­pletely obscure the cell con­tacts, so it’s annoy­ing (or dan­ger­ous, depend­ing on your per­spec­tive) to align the spot welder elec­trodes in the right place. I added small notches to help with visual ori­en­ta­tion. I also could have used a Sharpie, but I was going to reprint the spacer any­way, so I fig­ured I’d make it pre­cise.

A more sig­nif­i­cant change is adding some rout­ing guides for the BMS bal­ance wires. It’s impor­tant to pre­vent the bal­ance wires from cross­ing over each other or the lead wires because fric­tion dur­ing move­ment could wear the insu­la­tion, even­tu­ally lead­ing to a short cir­cuit. I think this is actu­ally eas­ier to man­age with larger packs because you sim­ply have more sur­face area at your dis­posal; either way, this par­tic­u­lar approach unfor­tu­nately doesn’t scale to a dif­fer­ent pack design, but I do like it for an inline con­fig­u­ra­tion.

I also decided to add a bracket for hold­ing the BMS. This neces­si­tated a few holes for heat set inserts in the spac­ers.

Finally, late in the process I dis­cov­ered the way the ANR26650M1B cells vent is by split­ting the can on the pos­i­tive elec­trode. Accord­ing to their design doc­u­ments, this requires 1 mm of space avail­able at the edge of the cell, which I hadn’t accounted for.

Spot welding woes

I was mak­ing the changes to the cell spacer design, but it hap­pened to be an unusu­ally nice March evening in Port­land, so I took my kWeld out­side and started play­ing with one of the cells instead. My goal was to dial in the energy required to spot weld the pack, which I expected to be around 30 J for the neg­a­tive elec­trode and 40 J for the pos­i­tive elec­trode.

Frus­trat­ingly, the pos­i­tive elec­trode sim­ply would not make a reli­able weld nugget. Why? Why?! Well, the ANR26650M1B can is actu­ally made of alu­minum with a small 0.2 mm-thick pure nickel disc lam­i­nated to the end. It dis­si­pates heat extremely effi­ciently, which meant I needed to push more cur­rent than a sin­gle kCap unit could pro­vide. (I tried up to 150 J, but the time scaled with the cur­rent, so the instan­ta­neous heat trans­fer remained the same.)

I mulled over some options:

• Import another kCap mod­ule from Canada and run them in par­al­lel. This would get the kWeld up to about 1700 A, which would be pretty neat and I’m sure it would work, but it’s also a rather expen­sive option, con­sid­er­ing I still didn’t have even a sim­ple func­tional bat­tery at this point.
• Dis­as­sem­ble the kWeld, buy a lead-acid starter bat­tery (or LiPo), and try to use it that way. Also expen­sive and annoy­ing to boot.
• Try 0.15 mm-thick nickel strips instead. No guar­an­tee I’d fare any bet­ter (although I sus­pect I would). I’d have to increase the width to 40 mm to carry the same cur­rent.
• Try another cell model that doesn’t have the same can design. Assum­ing I can locate one with a more tra­di­tional nickel-plated or stain­less steel can, I should be able to suc­cess­fully weld to it.

I haven’t found a good source for 40 mm-wide nickel strip (should I be buy­ing sheets and cut­ting them myself?), so I started look­ing for alter­na­tive cells. The clos­est one may be the Vap­cell IFR26650. My under­stand­ing is Vap­cell rewraps large Chi­nese man­u­fac­tur­ers’ cells to sell to con­sumers, and there are no real datasheets or design guide­lines to go off of for this cell, so we’re more or less tak­ing a gam­ble they will be drop-in replace­ments for the ANR26650M1B. But they seem to per­form ade­quately, they have a rep­u­ta­tion for main­tain­ing their capac­ity for a lot of cycles, and they’re widely avail­able domes­ti­cally, so I’m going to give them a try.

Once I have a work­ing bat­tery and a bet­ter under­stand­ing of the entire process, I’ll prob­a­bly buy a sec­ond kCap to fin­ish mak­ing packs with my ANR26650M1B cells.

Spot welding Vapcell IFR26650 cells

The Vap­cell IFR26650 cells I received were mar­gin­ally larger than their Lithium Werks cousins, and while they fit in the spacer I designed, I felt they were now way too snug, so I decided to increase the diam­e­ter by about 0.5 mm. These cells seemed to have a more con­ven­tional mech­a­nism for vent­ing through their cath­odes so I wasn’t as con­cerned about that prob­lem; I just wanted a good fit that didn’t exert any force on the cell hous­ings.

I was able to suc­cess­fully spot weld 0.2 mm-thick nickel to the cath­odes of these cells at 60 J and the anodes at 70 J. Feel­ing out the qual­ity of the welds is a bit of an art. To cre­ate a good weld, you seem to need to apply a lit­tle bit of pres­sure, but not a lot, to the joint. The elec­trodes of the kWeld should be close to each other and posi­tioned at some­where between a 30 ° and 45 ° angle apart. Aside from pulling on the nickel strip, the most reli­able proxy for qual­ity I’ve found is the resis­tance value given by the kWeld in its post-weld read­out. For this cell, it seems like a value around 0.8 mΩ–0.9 mΩ is ideal. Higher val­ues indi­cate a poorer weld, and any­thing over about 1.2 mΩ prob­a­bly isn’t usable.

Making the series connections

After I printed my updated cell holder design, I made sure every­thing looked good before weld­ing the “pro­duc­tion” cells.

Weld­ing the cells was fairly unevent­ful. I went in order of the series (that is, I flipped the assem­bly over after I attached each nickel strip to a cell) and dou­ble-checked volt­ages and inter­nal resis­tance as I went. These cells had an inter­nal resis­tance around 4 mΩ–5 mΩ each and I charged them up to about 3.35 V.

I had a men­tal check­list for each con­nec­tion I made (which I have repro­duced here any­way, so I should have just writ­ten down in the first place, but I digress):

1. Set the kWeld to the proper energy, either 60 J or 70 J depend­ing on whether we’re weld­ing to the cath­ode or anode, respec­tively.
2. Posi­tion the elec­trodes using the guides in the cell holder. Make sure the angle is good. I put a lit­tle torque on them because I felt it made it eas­ier to apply con­sis­tent pres­sure as I clearly favor my right hand oth­er­wise.
3. Weld the cells. Leave the foot pedal engaged after each weld to make sure the resis­tance was less than about 1.2 mΩ. Aim for at least two and ide­ally three very high qual­ity welds per con­nec­tion.
4. Check the volt­age of the entire bat­tery so far and make sure it adds up cor­rectly.
5. Check the inter­nal resis­tance of the entire bat­tery using a microohm­me­ter and make sure it isn’t get­ting too out of whack.
6. Flip the bat­tery over and repeat.

Wrapping everything up

Before we start lay­er­ing tape over every­thing that shouldn’t be exposed, we need to make sure the ther­mis­tor is in a prac­ti­cally use­ful place for the BMS to be able to detect over­heat­ing. I decided to put it roughly in the enclosed area between two cells, although I also con­sid­ered tap­ing it directly to one of the ter­mi­nals. (The way I chose felt a lit­tle cleaner, but I think in the­ory it might take slightly longer to trip the BMS if some­thing goes wrong.)

To con­nect the bal­ance wires, I first routed them through their respec­tive guide holes. I was happy the other end was already assem­bled in a hous­ing because I didn’t have to worry about shorts. I made them fairly taut because I’m going to back them out a bit once I strip and sol­der them. With the tinned pads on the nickel strips, it took less than a sec­ond to sol­der each wire, so I have no con­cerns about excess heat affect­ing the cells.

At this point, I no longer had a coher­ent plan, and it shows, because things got ugly. Not dan­ger­ous, just aes­thet­i­cally dis­pleas­ing. I started indis­crim­i­nately wrap­ping every­thing that looked slightly metal­lic in lay­ers of fish­pa­per and Kap­ton tape.

In ret­ro­spect, I could have built out the cell holder a bit more to pro­vide addi­tional struc­ture for the fish­pa­per, or even to replace it entirely to get more of a lap­top bat­tery look. I’ll prob­a­bly take the whole thing apart at some point to do just that, but for now, I’m excited to have some­thing that works!

Parts list

Conclusion

What’s that famous adage? “Mea­sure twice, cut once.” Well, it doesn’t help when you delib­er­ately mod­ify the plan after you for­got what the con­straints were. Adding the bracket and sled for the BMS made my bat­tery too tall and it didn’t fit inside the trac­tor. So, for now, some vel­cro straps are doing the hard work of keep­ing every­thing roughly in place while I design a new hood.