[HN Gopher] Fire hazard of WHY2025 badge due to 18650 Li-Ion cells
___________________________________________________________________
Fire hazard of WHY2025 badge due to 18650 Li-Ion cells
Author : fjfaase
Score : 63 points
Date : 2025-08-06 06:29 UTC (2 days ago)
(HTM) web link (wiki.why2025.org)
(TXT) w3m dump (wiki.why2025.org)
| bravetraveler wrote:
| That's a lot of juice for a trinket. Shorted a 'mitigated'
| battery of these in a charger once, was enough to reconsider
| vaping entirely
| pjc50 wrote:
| Given how much torch you can power off one of these, I'm not
| clear why you'd need _two_ of them for a badge, since any
| conceivable use for that much current is going to start heating
| up the badge PCB fairly quickly. They 're usually sized for >1A
| each, and you can get >10A off them for short periods, which is
| very lively for a badge pinned to flammable clothing.
| arghwhat wrote:
| Assuming decent cells with low ESR (say, <30mOhm), one such
| cell will deliver _hundreds_ of amps when shorted, making
| things a little bit more lively than your estimation. :)
|
| (A few hundred amps isn't a lot for a shorted battery, but
| these are tiny cells so that's what you'll get.)
|
| Two cells was probably selected for one of: Voltage to avoid
| boost converters, capacity to avoid having to do extensive
| power optimization to make it run the whole event, balance to
| make it hang even off your neck.
| piva00 wrote:
| Quite oversized amounts of power for 2 ESP32s running, last
| I measured a ESP32 board I have running it would consume
| some 120-180mA.
| arghwhat wrote:
| That is _highly_ dependent on the exact ESP32 model and
| current power state. They can use in excess of 500mA with
| peaks of 800mA, and can easily use >250mA constant at
| 3.3V. Some draw _much_ more than that.
|
| A board with a few bits and bobs on and a single 18650
| cell might only last, say, 8 hours on a charge.
|
| Now, a well optimized board with a low-power ESP32 and
| proper use of sleep states would make that number go from
| 8 hours to over a week, but that does take a lot of extra
| effort and _may_ not be worth it over just slapping on
| another cell.
| phire wrote:
| 2 ESPs and a 4" color LCD screen (and a keyboard).
|
| Depending on how bright the backlight was, that could eat
| through battery. And if they were using the wifi for any
| active communication, that increases power too.
|
| I suspect they wanted it to last the entire weekend with
| the display always on. The original design probably only
| had one cell (maybe even smaller battery with built-in
| protection) and they hurriedly switched to two 18650
| cells at the last moment.
|
| They probably went with parallel because that seemed
| easier, no need to switch to another voltage regulator
| and charge controller.
| farhaven wrote:
| > Two cells was probably selected for one of: Voltage to
| avoid boost converters, capacity to avoid having to do
| extensive power optimization to make it run the whole
| event, balance to make it hang even off your neck.
|
| It's likely not voltage because they're connected in
| parallel.
| bsder wrote:
| > The WHY2025 badge was designed to be powered by 2 Li-Ion 18650
| battery cells connected in parallel.
|
| Wait, what?
|
| I was under the impression that Lithium batteries were really
| difficult to put in parallel without a _LOT_ of engineering work.
|
| The discharge curve for Lithium batteries is super flat. If you
| put them in parallel, even a small differential between the two
| means that one battery will completely discharge simply trying to
| bring the voltage of the other up to match. This is very
| different from the discharge curve from alkaline which has a nice
| slope and the batteries can equalize without burning up very much
| of their capacity.
|
| These don't look like they're matched in any way. The connection
| between them doesn't like very big--I suspect a non-trivial
| voltage drop if one battery tries to empty into the other.
|
| If you need the power, it's much better to put them in series and
| use a buck converter to bring the final value where you want it.
|
| This seems more like a fundamental engineering flaw rather than a
| fault in the boards (although, to be fair, the creepage and
| clearance don't look great).
| exmadscientist wrote:
| Li-Ion chemistry is pretty happy with 1S2P configurations.
| (That doesn't necessarily mean you should do it.) 2S1P is where
| the fun starts.
| rollulus wrote:
| It's what I thought as well, but I'm not too much into
| electronics to hold an opinion. It looks like there's a
| balancing resistor between them:
| https://gitlab.com/why2025/team-badge/Hardware/-/blob/main/C...
| bsder wrote:
| With _floating grounds_ due to those MOSFETS adding 50
| milliohms or so (on the order of the internal resistance of
| the batteries!)!
|
| YIKES!
| eqvinox wrote:
| Putting the protection circuit* on the battery's negative
| pole is standard best practice (due to NMOS efficiency, and
| it not being a problem in the slightest), and the 50mO
| actually improves balancing. Please avoid making comments
| like this based on half knowledge.
|
| [*] I do wish it were an actual full protection circuit. It
| isn't. Then again a run of the mill protection circuit
| commonly doesn't cover reversed polarity [between protector
| and cell], which is rather important for this specific
| appliation.
| bsder wrote:
| > Putting the protection circuit* on the battery's
| negative pole is standard best practice
|
| Pointer? Especially since LiPol paralleling seems to want
| to use bus bars to minimize wiring resistance.
|
| Admittedly my experience is all about _avoiding_ parallel
| LiPol batteries ...
| eqvinox wrote:
| https://www.ti.com/lit/gpn/bq77908a
|
| https://www.diodes.com/assets/Datasheets/products_inactiv
| e_d...
|
| Look at the reference circuits, it's a pair of antiserial
| NMOS on the negative pole.
|
| (Those 2 protection circuits are at the opposite ends of
| complexity & features)
|
| To be clear, using 2 PMOS on the positive pole is also
| quite common, my choice of words with "standard best
| practice" might be a bit misleading.
|
| > use bus bars to minimize wiring resistance.
|
| Those come _after_ the protection circuit, there should
| always be 2 MOSFETs in series with the _individual_ Li-
| Ion cell in a design like this (specifically: user
| swappable cell).
|
| (Protecting paralleled cells _together_ is kinda
| nonsensical because you also want to protect them _from
| each other_ , I don't think I've ever seen a 2P combined
| protection circuit.)
| bsder wrote:
| Those datasheets show creating a series pack/cell. They
| don't show the circuitry to then parallel the packs
| together.
|
| I guess I need to do more research on this.
| eqvinox wrote:
| > Those datasheets show creating a series pack/cell.
|
| You seem to only have looked at the TI one, the Diodes
| one is for a single cell.
|
| & if the cells are "permanently" connected in a pack, you
| wouldn't have individual cell protection and just have
| them properly balanced before connecting them in factory.
|
| > parallel the packs together
|
| You parallel cells, not packs.
| pjc50 wrote:
| Is it just me or is that schematic hard to read due to bits
| of text being on top of each other? Also "LED will burn when
| battery wrong way round" .. how about fixing this problem
| which you have acknowledged? What happens to your balancing
| resistor when you put one battery one way round?
| ajb wrote:
| "LED will burn when battery wrong way round" .
|
| I don't know about the rest of it, but I think this is just
| an idiosyncratic translation of "LED will light when
| battery wrong way round" - IE it's a warning LED.
| eqvinox wrote:
| It doesn't matter what happens in the 10 to 90% range, if one
| discharges before the other, it's perfectly fine. It's not like
| this is an application that needs the combined current
| capability of both cells. What does matter is that neither cell
| is overcharged nor deep discharged, and the [dis]charge curve
| is absolutely not flat in those areas.
| stavros wrote:
| The GP is talking about the case where you plug in two
| batteries of varying charge levels or health, which I agree
| is not an amazing thing to do.
| eqvinox wrote:
| That's not obvious from the comment and logically
| inconsistent with parts of it; if the trace between the
| cells is small it would act as an auxiliary fuse, and there
| is that balancing resistor (whose value I can't read
| because whoever drew that schematic didn't bother
| repositioning overlapping labels.) I'm also a bit confused
| about the 2 polyfuses.
|
| That said you're right and I was focusing a bit too much on
| my reading/interpretation of the GGP post. I'm not sure
| I've ever seen a 1S2P LiIon configuration _with
| individually user swappable cells_. In the 2-cell design I
| did, I specifically decided to go for 2S1P and have the
| balancing circuit, to avoid this exact issue. It does have
| the downside that you need both cells, the WHY design works
| with only one populated... (which is what I 'd recommend
| doing in any case.)
|
| [ed.: the balancing resistor seems to be 200O. The
| polyfuses are 15mO. So I guess it's designed to trip one or
| both polyfuses if the cells are imbalanced. That's an...
| "odd"... design.]
| stavros wrote:
| > That's not obvious from the comment and logically
| inconsistent with parts of it
|
| I agree, but personally I decided to go with the most
| charitable interpretation, and that's the one that made
| the most sense to me.
| jacquesm wrote:
| Paralleling 18650's is relatively easy. You need to match
| voltage to within a few mV and make sure the connection is
| _really_ solid (welded) to ensure they stay paired perfectly.
| Flaky connections, putting cells in series, impact damage, bad
| chargers etc are the risky bits, a solidly connected pair of
| 18650 's is to a close approximation just as safe as a single
| cell, but it _does_ have twice the short circuit current so you
| are going to have to be more careful around them. But at least
| the casings will be at the same potential.
|
| I've built a 17P10S pack which was a pretty interesting (and
| scary) effort but it has been working flawlessly for years now
| with just one inspection of the guts after two years to make
| sure that nothing was coming loose (it's on an s-pedelec
| e-bike). In a big pack like that it's the spaces between the
| alternating blocks of cells and on top where the interconnects
| are that the real risk lies, besides the fact that the short
| circuit current of that pack is just shy of a kilo ampere so
| you _really_ don 't want to drop a tool or a piece of
| interconnect strip on that.
| ajsnigrutin wrote:
| Many devices have parallel lipo cells, from powerbanks to
| electric cars, nothing special here.
|
| If one cell is weaker, the other provides more current, there
| is no "one discharging/emptying into the other" during normal
| work (read below). No real need for any proper matching either,
| if you only care about capacity (if you care about current, you
| don't want to get into a situation where any of the cells has
| to provide more current than designed for and safe.
|
| The only "problematic" part of parallel batteries is making the
| first connection, where one might be at a much higher voltage
| than the other. Usually this is mitigated by equalizing
| voltages (either dis/charging to a fixed voltage, or do a
| parallel connection through a proper resistor), and after
| they're safely connected in parallel, it doesn't matter.
|
| On the other hand, two cells, user removable and replacable can
| cause exactly this issue, where the user removes one, recharges
| it in an external charger and replaces it (while the other,
| empty one, still stays inside)... but maybe there's a diode
| somewhere that prevents reverse currents.
| exmadscientist wrote:
| > The WHY2025 badge was designed to be powered by 2 Li-Ion 18650
| battery cells connected in parallel. The cells provided to
| visitors are of the "unprotected" kind
|
| ...I am going to put on my "client-facing consultant" hat for a
| moment, which means skipping the expletives, and just say that
| not only is this a Very Bad Design, it is such a Very Bad Design
| that someone should really have noticed this and not let it
| happen.
|
| Because this really is a Startlingly Bad Idea.
| fjfaase wrote:
| An earlier design by Badge.Team [1] did not use these cells and
| Badge.Team (no longer associated with WHY2025) strongly advised
| against the use of these cells.
|
| The earlier design has been matured into Konsool [2] and is
| available as Tanmatsu [3].
|
| [1] https://badge.team/
|
| [2] https://badge.team/docs/badges/konsool/
|
| [3] https://nicolaielectronics.nl/tanmatsu/
| stavros wrote:
| Why was that advice ignored?
| eqvinox wrote:
| WHY2025 orga has been a shitshow on the social &
| interpersonal level. The badge team was one of the
| casualties. (And triggers, from what I've heard they
| weren't paragons either, but that's even more hearsay.)
|
| (Source: I'm in c3noc/Internetmanufaktur, though not
| attending WHY. TBH I saw the shitshow coming and decided I
| don't need it in my life.)
| fjfaase wrote:
| Often it is the case that safety concerns are overruled by
| less technical savvy managers for some lesser critical
| reason.
| ginko wrote:
| But surely two 18650s are just overkill. I might
| understand switching from the recommended LiPo cell to a
| single 18650 for cost cutting reasons (even though that's
| still probably a bad idea safety-wise), but why two?!
| slacktivism123 wrote:
| What's with the mock-security-advisory with logo for the
| 'vulnerability' (Heartbleed, anyone?)
|
| Why is the important safety advice buried in a bunch of
| interpersonal drama and administrivia?
| Juerd wrote:
| Author here. I didn't add the logo, but it's a wiki so others
| can theoretically change things.
|
| I think the logo is cute though, so let's keep it. I think it
| was made with the WHY2025 logo generator at
| https://design.why2025.org/
| eqvinox wrote:
| I'm confused that noone is pointing out most protected 18650
| cells won't even fit in those holders, since protected cells are
| generally in the 18690...18700 pseudo size range. That's too long
| to get into those holders.
|
| Source: the holders are likely Keystone 1042
| [https://www.keyelco.com/product.cfm/product_id/918], which I've
| worked with before. For a protected cell, cf. for example
| https://imrbatteries.com/products/panasonic-ncr18650b-3350ma... -
| note 69.41mm length.
|
| [ed.: it's the China equivalent of a Keystone 1042,
| https://www.lcsc.com/product-detail/C2988620.html - I can't
| confirm but am 95% confident a protected cell won't fit; if it
| would, the hold on an unprotected 18650 cell would be quite
| loose.]
| supakeen wrote:
| Sure, that means the idea of using unprotected cells was
| already there when the holders were selected :)
| eqvinox wrote:
| Yeah, I'm just saying, you can't even buy regular protected
| cells and put them in, because they won't fucking fit. I do
| _think_ "actual" 18650 protected cells exist, but they would
| be rare and expensive because you can't build them out of
| mass manufactured bare 18650 cells (for obvious reasons of
| where do you put the damn protection circuit.)
| supakeen wrote:
| Ah yea then I misunderstood. That's right you can't easily
| switch out the cells for protected cells yourself :(
| unwind wrote:
| That is mentioned in the article, although could perhaps be
| more emphasized since it does mean the "obvious" fix is not
| possible:
|
| _Commonly available protected 18650 cells don 't fit in the
| badge's cell holders because they are slightly longer._
| eqvinox wrote:
| I did indeed totally miss that, thanks for pointing it out!
| krekr wrote:
| I've worked with the Keystone holders and can confirm that
| those will hold protected cells, at least some I got from
| nkon.nl
| leptons wrote:
| The Keystone holders are nice but expensive, but they do not
| fit most protected 18650 cells, and I don't like the PCB
| mounting options.
|
| I designed my own 3D printed 18650 holder for my project,
| including a positive battery tab cut-out to prevent reverse
| battery insertion. I get to decide how big the battery can be,
| and protected cells are 100% the way to go.
|
| I've never had a problem with a short with the protected cells,
| and my circuit also cuts off power to the load using a mosfet,
| if a short ever occurs. It's been working great for years.
| giantg2 wrote:
| Funny, I'm also going with a 3D printed 18650 battery
| module... for a lepton (FLIR) project.
| ryao wrote:
| At work, whenever we design hardware that uses Li-Ion cells, I
| always discuss safety before we even have done the design work so
| that we build something that is safe. Why did these guys not do
| that? Did they only learn about videos of li-ion explosions/fires
| after design this?
|
| By the way, they probably should have used a LiFePo4 chemistry
| instead. It would not have the same runtime, but it would be much
| safer in worst case scenarios.
| burnt-resistor wrote:
| Why the fuck didn't they just use say 4 safe(r) CR2032s or simply
| have only +5V USB-C power in?
|
| They could've eliminated most of the risk by simply ripping the
| 18650 holders off the badges and rely on USB power.
| andrewstuart wrote:
| Avoid battery fires: design your boards for alkaline batteries.
|
| They're safer and in many cases just fine for the job - for
| example a conference badge needs nothing more than alkaline
| batts.
|
| Also, alkaline batteries are not an expensive nightmare to ship.
| burnt-resistor wrote:
| Rechargeable safe(r) chemistries. Alkaline batteries are insta-
| e-waste. Like the common form-factors of AA/AAA Ni-MH that are
| acquirable locally.
| rgovostes wrote:
| Conference badges are basically insta-e-waste too.
| em3rgent0rdr wrote:
| AA/AAA won't fit in a flat conference badge. I've wanted to
| get coin cell format NiMH, but they only seem to be the more
| thicker button type (or I don't know the magical term to
| search for), and I can't seem to find cheap options either.
| The 40 maH ones seem to be 5mm tall and I can't seem to find
| many smaller capacity ones that are thinner.
| michaelt wrote:
| In the badge from the article they're using 18650s, so AA
| batteries would actually have slimmed it down by 3.5mm.
|
| Lower voltage, of course.
| charcircuit wrote:
| Hopefully, they get sued to clearly set the precedent that this
| is not acceptable. Just because it may be a devkit, targeted
| towards knowledgeable individuals, or amateur made that doesn't
| mean that they should be distributing unsafe electronics like
| this because it's cheaper than making a safe version.
| seszett wrote:
| This is the Netherlands, they obviously (and luckily IMO) won't
| get sued when there has not even been any incident and they
| have a clear advisory for the safety problem.
|
| I don't even know on what grounds could anyone sue them.
| krekr wrote:
| I've also written up some more findings. They can be found at
| https://www.krekr.nl/content/why2025-badge-fire-hazard-addit...
| 1oooqooq wrote:
| fireworks in the badge? take that defcon!
| 1970-01-01 wrote:
| It's all cute fun and nerding out until the room is filled with
| smoke. The art of one upping your hackercon badges is clearly
| getting out of control. Good job on calling it out now.
| lysace wrote:
| Repost from a previous comment of mine, that sort of applies
| here:
|
| > I get nervous when I see videos of people buying random Li-
| Ion/Po-battery powered crap from Teemu etc.
|
| > My personal policy for buying anything with such a battery: the
| seller must have a meaningful presence in my country, selling at
| least like $10M/year.
|
| I.e. they need to have a large enough exposure to handle a
| catastrophic house fire if something happens due to e.g. a bad
| design. I figure that at around $10M/year they start caring, even
| if they are psychopaths or incompetent.
| ShakataGaNai wrote:
| I love these concept of badges, but almost never are they well
| executed. Defcon has had TONS of problems with their badges of
| all types. OpenSauce has tried for the last two years with only
| middling results.
|
| Don't get me wrong, I'm not trying to say "don't do it" or "These
| people are stupid". It's just that people underestimate the time
| and effort required. It's basically bringing a product to market,
| for 20 to 50k people (depending on the event), in a few months
| time. But it also needs to be "cool" and "unique" and often
| "beginner friendly" and extremely cheap. Crazy crazy hard.
| Waterluvian wrote:
| Engineering is the first 80% of work. But productization is the
| second 80%. I find that the more nerdy a community or product
| audience, the more the latter suffers.
| knotimpressed wrote:
| Braving potentially getting spit roasted to ask:
|
| If I'm making my own 18650 USB C power banks, are there any easy
| to miss risks? I've got the cells in holders, not welded, but the
| holders are specc'd above the current I need. The cells are
| unprotected, but the Aliexpress listing for the power management
| board says specifically to use unprotected cells, as at 6A draw
| most protection boards don't do well (dubious). The cells are
| tested and mechanically protected by a thick enclosure. The only
| EE work I'm doing is soldering 2 high gauge wires from the holder
| to the board that's doing everything else. I know Aliexpress
| isn't a bastion of quality, but the seller has good feedback and
| I checked over the board to make sure there's at the very least a
| good counterfeit battery protection IC included.
|
| Currently, the concerns I have are: - the holder relies on good
| contact to deliver 6A without developing hotspots on the
| terminals - the board from Aliexpress perhaps should not be
| trusted
|
| If there's anything else anyone can think of, I'm happy to hear
| it.
___________________________________________________________________
(page generated 2025-08-08 23:00 UTC)