N o a h ?

A framework for building Li⁺ batteries

With my new­found2024 3D print­ing capa­bil­ity, I decided to try to build an RC trac­tor. Of course, I wasn’t about to get a quick dopamine hit: it was going to take like a week just to print the plas­tic parts, nev­er­mind putting every­thing else together. So I went down a few small rab­bit holes. One of them is bat­tery tech­nol­ogy.

Types of Li⁺ batteries

These days, RC mod­els are mainly pow­ered by lithium-ion bat­ter­ies, which I have broadly referred to here as Li⁺ bat­ter­ies for a rea­son we will under­stand clearly in a moment. Most of your portable con­sumer elec­tron­ics, elec­tric cars, stor­age bat­ter­ies for solar arrays, and lots and lots of other things are also wired up to Li⁺ bat­ter­ies.

There are many kinds of Li⁺ bat­ter­ies, each with dif­fer­ent trade­offs. Some never became com­mer­cially viable; oth­ers have mostly fringe indus­trial appli­ca­tions. But a few are prac­ti­cal for hob­by­ist or house­hold use cases:

Before we con­tinue, notice one type of bat­tery tech­nol­ogy here is usu­ally referred to as lithium-ion or Li-ion. This seems to be the source of an enor­mous amount of con­fu­sion for many peo­ple. To try to limit the blast radius, I will refer to this tech­nol­ogy exclu­sively as Li-ion to dif­fer­en­ti­ate it from Li⁺ bat­ter­ies con­cep­tu­ally.

Com­mon chem­i­cals for a lay­ered oxide cath­ode are lithium cobalt oxide (LCO), lithium nickel man­ganese cobalt oxide (NMC) and lithium nickel cobalt alu­minum oxide (NCA). The spinel oxide cath­ode is usu­ally lithium man­ganese oxide (LMO). Some­times bat­ter­ies are referred to by cath­ode mate­r­ial alone; in this case, the bat­tery is implied to be an Li-ion chem­istry. You can be more spe­cific, too. For exam­ple, you can make an NMC/LTO or LMO/LTO bat­tery. Some­times mul­ti­ple com­pounds are used in a cath­ode in a fixed ratio, for exam­ple, LCO/NMC 2:1.

The cath­ode mate­r­ial, anode mate­r­ial, elec­trolyte, sep­a­ra­tor, and other (often pro­pri­etary) addi­tives make up the phys­i­cal struc­ture of a cell. We can see how the struc­ture varies for these bat­tery types, often in only one dimen­sion. Vary­ing the mate­ri­als changes the char­ac­ter­is­tics of the cells. I’ve iden­ti­fied a few prop­er­ties I care about (there are many oth­ers you may care to research your­self):

• Energy den­sity is a mea­sure of the amount of energy that can be stored per unit vol­ume (vol­u­met­ric) or per unit mass (gravi­met­ric). This is impor­tant to deter­mine the over­all space required to hold your bat­tery in your device and how heavy your bat­tery is going to be, respec­tively. Gravi­met­ric energy den­sity is also called spe­cific energy.

• Power den­sity is a mea­sure of how quickly a bat­tery can be charged or dis­charged, that is, power per unit vol­ume or mass. RC is obsessed with it, and RC bat­ter­ies are often specced with out­lan­dishly imprac­ti­cal dis­charge rat­ings like 100 C. Cell man­u­fac­tur­ers do pro­vide this infor­ma­tion in their datasheets, usu­ally in terms of max­i­mum con­tin­u­ous dis­charge cur­rent (e.g., 60 A) and a time-lim­ited peak dis­charge cur­rent (e.g., 120 A for 10 s).

  A C rat­ing is a ratio that scales the over­all capac­ity of the bat­tery to charge or dis­charge over a period of 1 hour. To con­tinue the exam­ple above, for a bat­tery with a capac­ity of 5 A·h, it would sup­port con­tin­u­ous dis­charge at $$\frac{60\mathrm{A}}{5\mathrm{A}\mathrm{\cdot }\mathrm{h}}=12\mathrm{C}\mathrm{\text{.}}$$ You could fully dis­charge this bat­tery in 1⁄12th of an hour. The C rat­ing is con­ve­nient to work with because it remains the same when you add cells in par­al­lel.

  C rat­ings have noth­ing to do with the coulomb, which, like the amp hour, is a unit of elec­tric charge.

• The safety pro­file is sub­jec­tive.

  Many Li⁺ chemistries require care­fully man­ag­ing bat­tery volt­age and tem­per­a­ture to pre­vent ther­mal run­away. Exceed­ing either the spec­i­fied min­i­mum or max­i­mum cell volt­ages (when dis­charg­ing or charg­ing respec­tively) can irre­versably dam­age cells, includ­ing inter­nal short cir­cuit (ISCr). Addi­tion­ally, some Li⁺ bat­ter­ies can enter ther­mal run­away at tem­per­a­tures as low as 130 °C and can reach tem­per­a­tures over 800 °C.

  Li⁺ bat­ter­ies with mul­ti­ple cells in series require bal­anc­ing because you must dis­con­nect the entire pack when any one of its cells reaches its min­i­mum or max­i­mum state of charge (SoC), i.e., the volt­age lim­its. The sim­plest design is pas­sive bal­anc­ing using bleeder resis­tors, which sim­ply dis­si­pate excess charge as heat. This is inher­ently waste­ful, so many appli­ca­tions use at least switch­ing pas­sive bal­anc­ing or even micro­con­troller-dri­ven active bal­anc­ing instead for effi­ciency. Espe­cially with active bal­ancers, I’m think­ing about com­po­nents fail­ing in a way that allows the bat­tery to over­charge or overdis­charge.

  My cri­te­ria for safety are:

  • Cell sta­bil­ity when over­charged, overdis­charged, or phys­i­cally dam­aged
  • Ther­mal run­away onset tem­per­a­ture
  • Max­i­mum tem­per­a­ture dur­ing ther­mal run­away
  • Mate­r­ial tox­i­c­ity
• Cycle life refers to the num­ber of times a par­tic­u­lar cell can be charged and dis­charged before it no longer holds an accept­able capac­ity. I hap­pen to care about this a lot because I sim­ply dis­like replac­ing bat­ter­ies, both phys­i­cally and emo­tion­ally. Most peo­ple just put up with short cycle lives, I think.

RC appli­ca­tions have largely set­tled2025 on LiPo bat­ter­ies for their excel­lent power den­sity and rel­a­tively good energy den­sity. With other safety mech­a­nisms in place, this could be a rea­son­able trade­off. Unfor­tu­nately, there’s more to dis­cuss.

I don’t under­stand chem­istry or mate­ri­als sci­ence, so I’m not try­ing to sound smart by talk­ing about these. I’m just try­ing to put a face to a name. But it would be inter­est­ing to get a bet­ter under­stand­ing of why each one behaves the way it does.

Pack theory

Series and par­al­lel con­nec­tions in bat­ter­ies are expressed in the form $N$s$M$p, where $N$ indi­cates the num­ber of series con­nec­tions and $M$ indi­cates the num­ber of cells in par­al­lel per series. The total num­ber of cells in a $N$s$M$p bat­tery is thus $N\times M$. When $M$ is 1, you can omit that part of the syn­tax, and just write $N$s, which is com­mon in RC.

If we assume a con­stant inter­nal resis­tance (fine for our use case), Ohm’s law dic­tates why you’d con­nect cells in series or par­al­lel. Adding in series increases the volt­age. Adding in par­al­lel increases the capac­ity and max­i­mum dis­charge cur­rent. For exam­ple, a 4s1p 5 A·h LFP bat­tery has a nom­i­nal volt­age of $$4\times 3.2\mathrm{V}=12.8\mathrm{V}$$ and a 5 A·h capac­ity. Dou­bling the num­ber of cells in par­al­lel to con­struct a 4s2p pack would give us a 12.8 V bat­tery with a capac­ity of $$2\times 5\mathrm{A}\mathrm{\cdot }\mathrm{h}=10\mathrm{A}\mathrm{\cdot }\mathrm{h}\mathrm{\text{.}}$$

Battery management systems

To make charg­ing and dis­charg­ing Li⁺ bat­ter­ies safer and more con­sumer-friendly, many (but not all) sit behind a bat­tery man­age­ment sys­tem (BMS). A decent BMS pro­tects the SoC range, mon­i­tors tem­per­a­ture, and pro­vides at least rudi­men­tary bal­anc­ing func­tion­al­ity.

To work cor­rectly, a BMS needs to be con­nected to each series in the pack. That is, in addi­tion to the usual pos­i­tive and neg­a­tive ter­mi­nals of the bat­tery itself, you must wire in a (much smaller) ter­mi­nal between each series for the BMS to mon­i­tor SoC and per­form bal­anc­ing. The BMS then sim­ply exposes a sin­gle lead that can be con­nected to any dis­charg­ing cir­cuit or con­stant cur­rent/con­stant volt­age charg­ing cir­cuit.

Bewil­der­ingly, BMSes aren’t used for pro­duc­tion RC mod­els today2025, either due to igno­rance, lack of con­cern, or, per­haps most likely, because they weigh a few grams. RC bat­ter­ies pro­vide a set of bal­ance leads because RC charg­ers do han­dle over­volt­age pro­tec­tion and bal­anc­ing. But once the bat­tery is in the model, it’s directly con­nected to an elec­tronic speed con­trol (ESC) that usu­ally only has an under­volt­age cut­off for pro­tec­tion. This can be prob­lem­atic, though, because it may not detect cells that are wildly out of bal­ance! For exam­ple, an ESC with a 4s bat­tery and a cut­off volt­age of 3.2 V would not be effec­tive with a dan­ger­ously low 2.5 V LiPo cell in series with three healthy cells at 3.5 V because $$4\times 3.2\mathrm{V}<2.5\mathrm{V}+3\times 3.5\mathrm{V}\mathrm{\text{.}}$$ A BMS would instantly detect this con­di­tion and dis­con­nect the bat­tery.

Battery-building communities

There are very few peo­ple in RC build­ing their own bat­ter­ies. I sup­pose the off-the-shelf LiPos work fine for vir­tu­ally every­one, but they’re just not to my taste. Rad­i­cal RC will build you a pack from very nice Lithium Werks cells, but they still have exposed bal­ance leads instead of an inte­grated BMS.

On the other hand, I’ve found these com­mu­ni­ties to have enthu­si­as­tic bat­tery builders, maybe too enthu­si­as­tic even:

• Elec­tric skate­board­ers, who have a stun­ningly high bar for safety and many good ideas about how to build packs. Their bat­ter­ies are put through a ton of abuse, for bet­ter or worse.
• Elec­tric bicy­clists, also with a lot of good ideas. More care­free than elec­tric skate­board­ers, but still smart enough to know their bat­ter­ies need to sur­vive a lot of jostling.
• Sta­tion­ary stor­age and solar folks, an inter­est­ing blend of envi­ron­men­tal­ists, prep­pers, and recy­clers try­ing to make bizarre franken­bat­ter­ies do their bid­ding. The safety pro­file is dif­fer­ent: they like LFP because they have a huge amount of energy stored in their house, but they also aren’t as con­cerned with the qual­ity of indi­vid­ual cell con­nec­tions because their bat­ter­ies rarely move.
• Flash­light enthu­si­asts. I haven’t spent too much time here, but for smaller packs, they could have some valu­able insight.

• Peo­ple who vape. Not my cup of tea. They’ve also some­how caused so many prob­lems that OEMs had to put explicit warn­ings on cell cases.

Practical cell choice considerations

LTO and LFP are both com­pelling tech­nolo­gies. Between them, LTO has bet­ter power den­sity and cycle life, while LFP has bet­ter energy den­sity and is slightly safer. LTO cells also cost more than LFP cells right now2025. I’m already giv­ing up a lot of energy den­sity by mov­ing away from LiPo, so LFP seems like a more log­i­cal choice in this sit­u­a­tion.

You can find Li⁺ cells in three basic pack­ages: pouch, pris­matic, and cylin­dri­cal.

Pouch cells have tabbed ter­mi­nals (often right next to each other!) attached to a flex­i­ble hous­ing made of lay­ers of alu­minum and plas­tic. These cells require care­ful ther­mal man­age­ment and need to be packed into a more rigid holder to work prop­erly; I’m not super inter­ested in them and won’t be explor­ing them fur­ther. They’re most com­monly used for LiPo any­way.

Pris­matic cells have large, rigid cases for energy stor­age appli­ca­tions. These fel­las are like the blade servers of the bat­tery world. They’re easy to wire up and have good ther­mal man­age­ment out of the box. Unfor­tu­nately, they’re too big for RC and they usu­ally don’t have great power den­sity (because it’s not required for large-scale stor­age). Oth­er­wise, they would be ideal.

Cylin­dri­cal cells look like what you imag­ine when you think about bat­ter­ies, e.g., sim­i­lar to AA or C alka­line cells you could buy at any gro­cery store. Man­u­fac­tur­ers spec­ify these cells by their dimen­sions. A 26650 cell, for exam­ple, has a 26 mm diam­e­ter and is 65 mm long. As long as you don’t acci­den­tally let the cath­ode and anode touch (they come quite close to each other, but are insu­lated, so it would require a han­dling error), these cells are rel­a­tively safe, and they’re what I’m going to use for my first hand­built bat­tery.

A123 Sys­tems was an Amer­i­can man­u­fac­turer that used a unique LFP tech­nol­ogy from MIT called Nanophos­phate. Once a bea­con of renewed inter­est in domes­tic pro­duc­tion, the com­pany quickly suc­cumbed to a com­bi­na­tion of macro­eco­nomic head­winds and poor man­age­ment. Through a series of acqui­si­tions, their cells are now made in China and sold under the name Lithium Werks. I find these cells appeal­ing because they have a some­what leg­endary rep­u­ta­tion for their lifes­pan and they have very good power den­sity to boot.

I’d like to build a 4s bat­tery from their 26650 cells. And I’d also like that bat­tery to have a BMS. Accord­ing to their spec­i­fi­ca­tions, I should end up with a 13.2 V 2.6 A·h bat­tery capa­ble of pro­vid­ing up to 52 A con­tin­u­ously.

Cell-to-cell connections

A few fac­tors that make work­ing with Li⁺ cells more chal­leng­ing than other com­po­nents:

• Since the whole point is to store energy, a short cir­cuit won’t just pro­duce some “magic smoke.” It could cause all the energy to be released at once, which could seri­ously hurt some­one.
• Heat­ing the cells beyond their spec­i­fied oper­at­ing tem­per­a­ture will dam­age them, at a min­i­mum reduc­ing their lifes­pan. We want our cells to last as long as pos­si­ble. It’s one of the rea­sons we picked LFP. The accept­able tem­per­a­ture range for many Li⁺ cells is sur­pris­ingly small; for exam­ple, the ANR26650M1B should dis­charge between 0 °C and 60 °C.
• Con­nec­tions need to be resilient enough to han­dle what­ever cur­rent we want to pro­vide. If we do want to go up to about 50 A, we’ll need to think more seri­ously about what mate­ri­als we use.

As con­ve­nient as it would have been, plas­tic spring-loaded bat­tery hold­ers can’t han­dle more than 3 A, so they’re out. When you buy a fac­tory-made Li⁺ bat­tery, the cells are usu­ally con­nected using thin strips of metal. But you can’t sol­der the strips directly to the cells: it would heat the cells up too much. Instead, they’re spot welded.

Spot weld­ing uses a pulse of elec­tric­ity to heat two met­als in phys­i­cal con­tact, caus­ing them to become joined. The phys­i­cal mech­a­nism is known as Ohmic or resis­tive heat­ing. The short dura­tion of the process—only a few mil­lisec­onds—and the small sur­face area involved make the approach ideal for attach­ing a metal strip to a cell ter­mi­nal with­out affect­ing the cell’s chem­istry.

Until a few years ago, it was vir­tu­ally impos­si­ble to get access to a decent spot welder on a hob­by­ist’s bud­get. To get your foot in the door might run you upward of $2,000; more sophis­ti­cated mod­els (ones that could weld a metal strip thick enough to carry 50 A) likely cost quite a bit more. On the less-than-decent side, peo­ple dis­cov­ered that you could make a spot welder with some minor mod­i­fi­ca­tions to a microwave oven trans­former.

Luck­ily, we now have a few DIY options, plus an assort­ment of ques­tion­ably good pre­built mod­els from China. None of them will set us back more than a few hun­dred dol­lars. The most expen­sive (and capa­ble) is the kWeld. The kWeld boasts a very nice design that inte­grates power deliv­ered with respect to time to reach a total spec­i­fied amount of energy to apply to the weld. (Com­pare to other mod­els at its price point that run on a timer.) Another good option is the Malectrics spot welder, which is Arduino-based and has open-source firmware. The options on AliEx­press are dizzy­ing, but I’ve seen good reviews of Glit­ter-branded units.

The effec­tive­ness of a spot welder is defined by Joule’s law of heat­ing, $$Q=I^{2}Rt\mathrm{\text{,}}$$ where $Q$ is the heat gen­er­ated, $I$ is the cur­rent through a con­duc­tor, $R$ is the resis­tance of the cir­cuit, and $t$ is time. We have no con­trol over $R$ (except for select­ing the con­duc­tor) and we want to min­i­mize the amount of time we spend cre­at­ing the weld, so we need to max­i­mize $I$.

We usu­ally pick mate­ri­als with high resis­tiv­ity to reduce the cur­rent demand on the spot welder. But this is at odds with our other require­ment: to carry a lot of cur­rent between cells once the weld is in place. In my head, a nice option would have been some­thing with high elec­tri­cal con­duc­tiv­ity but low ther­mal con­duc­tiv­ity so that the heat of the weld would remain as close to the elec­trodes as pos­si­ble. Then I learned about the Wiede­mann-Franz law and decided to pur­sue some­thing more real­is­tic.

Basi­cally, we have two choices for con­duc­tors: nickel or cop­per. Nickel is about four times less con­duc­tive than cop­per. In other words, it’s eas­ier to weld, but to carry the cur­rent we want, we’ll need a lot more of it between each cell ter­mi­nal. Large-scale man­u­fac­tur­ers use nickel strips, vary­ing the thick­ness to reach a desired ampac­ity.

I spent an unfor­tu­nately long time try­ing to val­i­date the com­mu­nity-dri­ven ampac­ity guide­lines for nickel strips with lit­tle suc­cess. At least the­o­ret­i­cally, there are sim­ply too many vari­ables and I made too many assump­tions to come up with a “smart” esti­mate. The best I can fig­ure at this point is to be ruth­lessly con­ser­v­a­tive for my first few bat­tery packs, do some empir­i­cal analy­sis on them, and build a model around that.

Finally, we need to con­sider the power source for the spot welder itself. Remem­ber, we need to max­i­mize $I$. A few com­mon options:

• A car bat­tery. Lead-acid starter bat­ter­ies are usu­ally rated for 400 CCA or more, and can han­dle about twice that when being briefly short cir­cuited for spot weld­ing pur­poses. A bat­tery rated for 800 CCA could prob­a­bly deliver the amper­age needed to weld fairly thick nickel if the welder cir­cuit could han­dle it.
• A LiPo bat­tery. Appar­ently works pretty well, but for me it defeats the whole pur­pose of not bring­ing unman­aged LiPo bat­ter­ies into my damn house. (You can’t use a BMS because, again, you’re effec­tively short cir­cuit­ing the bat­tery.)
• A mains-volt­age trans­for­mer with a rec­ti­fier bridge. Basi­cally a more expen­sive ver­sion of the microwave oven trans­for­mer cir­cuit since we need DC out­put.
• Super­ca­pac­i­tors. Works rea­son­ably well, but can take time to charge. They need active bal­anc­ing in series, which com­pli­cates the cir­cuit. More expen­sive per unit of energy than a car bat­tery, but less annoy­ing to carry around with you.

Based on my rough analy­sis, I’m going to der­ate my first bat­tery to a 25 A con­tin­u­ous draw. That will allow me to use a 0.2 mm by 30 mm nickel strip for the cell con­nec­tions. Many of the cheaper Aliex­press welders seem to have trou­ble with nickel more than 0.15 mm thick, and the more capa­ble ones cost about as much as the DIY options, so I’m going to look at either the kWeld or the Malectrics.

The Malectrics is rated for 800 A. You can power it from LiPos or car bat­ter­ies out of the box; to go the super­ca­pac­i­tor route, you’d have to build a mod­ule for it your­self. They unof­fi­cially endorsed one such design, but I’m not sure I want to peel another layer off this prover­bial onion. A full kit (with foot switch and case) for the Malectrics costs about $130.

The kWeld sup­ports up to 2000 A, and, in addi­tion to sup­port­ing LiPos and car bat­ter­ies, has a super­ca­pac­i­tor acces­sory avail­able to pur­chase sep­a­rately. The kWeld kit itself costs about $250 and the kCap is another $200.

If I had a spare car bat­tery, I’d prob­a­bly give it a try first. Unfor­tu­nately, I don’t, and in the cheaper range of car bat­ter­ies (less than, say, $100, from auto parts stores) you won’t really find much rated above 600 CCA. To take advan­tage of the extra cur­rent the kWeld can push, we’d prob­a­bly want to spend about $150 on a bat­tery any­way. At that point, you may as well opt for the con­ve­nience and longevity of the kCap.

By com­par­i­son, the kWeld can weld thicker mate­r­ial, has a bet­ter algo­rithm for deter­min­ing weld tim­ing, and can use a more con­ve­nient super­ca­pac­i­tor-based power source, but the design and firmware aren’t open source and it costs about twice as much as the Malectrics. After (some­what) care­ful con­sid­er­a­tion, I opted for the kWeld.

Conclusion

Now that I’ve decided on a chem­istry, pack­age, cell, and spot welder, I can start gath­er­ing the mate­ri­als to put a bat­tery together. I am sur­prised by how many deci­sions I needed to make and the trade­offs involved! There’s no one-size-fits-all or even best-but-most-expen­sive solu­tion, which I sup­pose explains the pro­lif­er­a­tion of tech­nolo­gies.

• Curi­ous how I arrived at my 25 A rat­ing? I try to explain it.
• I doc­u­ment the build process for the kWeld.
• Skip ahead to the assem­bly of my first bat­tery.