Carried over from: "Bought these (retro) hardware today"

yawetaG wrote on 2019-09-11, 11:11:

GigAHerZ wrote:

I just found that there are ML2032 coin cell batteries, that are physcally identical to CR2032, but are rechargable. Quick search to aliexpress lead to offers at around 3$ a piece. 😉

EDIT: And LIR2032, which is 3.6V rechargable in the physical shape of CR2032!

And if there's no circuitry present in the mother board to charge those batteries correctly, you've just obtained yourself a nice way to have a surprise fire via overcharging...

I was looking for suitable battery replacement options myself, and aside from 1:1 NiMH (or even NiCd) batteries or rigging up non-rechargeable lithium cells to the external battery connectors (or disconnecting the resistor from the charging circuit, or adding a diode), ML2032 cells look like a reasonable, safe, long-term replacement for rechargeable CMOS batteries.

It's the LIR ones that are problematic with those simple NiMH/NiCd charging circuits and they can be even more dangerous by appearing safe or harmless at first, but being problematic (or explosion/fire prone) under specific conditions. (overheating during short-circuit or allowed to drain excessively between charges) The ~4~4.5 ish max voltage on those motherboards tends to be OK with fully charged LIR cells (higher than the 4.2V recommended, but not crazy), but discharged cells might have dropped low enough for the charging circuits to overload, depending on the particular circuit. (LIR/LiPO cells are supposed to have stepped voltage for proper, safe charging, but the NiCD type circuits are constant-current ones that might or might not exceed safe limits ... it's possible that a constant current circuit with that 4~4.5V cap works out OK and has a voltage curve suitable for the load of the LIR cell, but that'd need specific testing to be sure).

ML2032 cells don't have those problems, though they have a nominal voltage of 3.0V (similar to CR2032 cells), that's sufficient for most/all SRAM data retention and RTC circuits. They also take a charge similar to or slightly greater than the traditional NiCD or NiMH 3-cell barrel batteries (65 vs 60 mAh) and tolerate high charge voltages, and should work fine with a typical constant current trickle charging circuit using a resistor (as on old motherboards and a few game consoles and arcade boards).

See this article for a quick reference to different Lithium cell types.
http://www.cr2032.co/chemistry-article.html

LIR type batteries also might seem cheaper than ML batteries in ebay lots, but there's a lot of shady/crappy/recycled LIR batteries out there on ebay (and some rebranded/rerated and supposedly refurbished ones), and the not-so-cheap priced listings are around the same as the ML ones anyway, if not more expensive. (and if packaged poorly and allowed to short-circuit in shipping, ML cells won't tend to fail catastrophically like LIR ones will, plus will more likely still be useable if heavily discharged)

Plus ML2032s tend to be in the same price range (or cheaper than) actual 3-cell NiMH barrel battery replacements. (and being a single cell, don't have the same range of failure modes that those 3-cell batteries do, including cell reversal or internal shorts when one of the 3 cells goes bad, leading to overcharging/swelling/leaking, among other things: some specific to NiMH and some to NiCd)

OTOH, it's way easy get plain non-rechargeable CR2032s cheap and new in decent sized lots, battery holders can be had really cheap, and usually isn't that hard to adapt one of those boards to use them. (soldering on a coil cell holder and loading a CR2032 into it without disabling the charging circuit could be problematic, though, but I've seen some people do that too: and given the 4-4.5 max charging voltage, it might not be that much of a problem, just not ideal or recommended) Tracking down the right resistor to disconnect might be troublesome in some cases.

It's also probably worth looking into what electrolyte these different cells use (ie which is a worst-case if it does leak), though I've seen boards damaged by both NiCd and NiMH leaks, and both seem capable of wicking through solder joints and into PCB traces. (some might use some ammonium salt, given how sensitive copper is to ammonia/ammonium ions or amines, and the dark blue corrosion crust of copper-ammonium complexes ... alkaline battery leakage doesn't usually cause the same sort of copper corrosion, but it's horrible on aluminum, though something like acetic acid might do that too, and explain the faint vinegar-like odor on some corrosion; ammonium acetate would also do that)

And despite the above article's mention of tolerating (or withstanding) 'high charge voltages' actual documentation for ML type battery charging circuits tend to caution against going over 3.3V and avoiding trickle charging circuits, noting potential damage above 4.0 volts and especially at 5V or higher. (though the failure mode seems to be lost capacity and potential electrolyte leakage, so again not explosion/fire as with LIR cells)

https://industrial.panasonic.com/cdbs/www-dat … AAF4000COL8.pdf

https://www.sii.co.jp/en/me/battery/support/c … rging-circuit1/

Though from what I recall, a simple resistor-regulated charging circuit is used on those old motherboards and game/arcade boards, similar to the one in that second link above. (but that still assumes a charging voltage of 3.3 or lower)

I also found this thread, which does have anecdotes of using LIR2032 cells, but not enough info to really say how safe that is.
A very special question about old 486 CMOS batteries

LIR2032s might actually be within the size/capacity range to be relatively difficult to catastrophically overheat under normal ambient conditions, even if they're the same basic type of Lithium-ion cell that can explode/burn when abused, but I'd probably avoid that without some torture testing to prove it. (ie testing worst cases like over-discharged cells in the <1V range getting charged by the simple resistor current-limited circuit)

LIR2032s seem to be available in the 40 capacity range, which is rather small, so the actual energy content is relatively low and surface area + thermal conductivity relatively high, so charging load/resistance might stay relatively low even in adverse charging conditions (and even short-circuits might be less catastrophic, potentially less so than a new CR2032 getting shorted given those are around 220 mAh)

Testing tolerances of LIR2032s might be easier to test than ML2032 failures from long-term over-voltage that could lead to leaking and corrosion similar to the original batteries. (they tend to use an organic electrolyte of some sort, so I assume either an organic acid or amine of some sort, neither of which are good around lead/tin solder and copper, or it's using an organic solvent, which is a different matter altogether)
Except if the closed-circuit power-on voltage is around 4V or lower, in which case I'd probably not worry too much about it. (something easy enough to just check with a multimeter, though re-checking after a PSU swap might be wise)

And come to think of it, acetic acid (and probably ammonium acetate) would be up there for being able to eat through high lead content solders more easily than a lot of other electrolytes. (which seems to be the main way electrolyte/corrosion seems to be able to wick into the PCB itself without other scratches/cracks)

Though potassium or sodium hydroxide (typical alkaline electrolytes) could also dissolve lead/lead oxide as plumbate salts, and those salts themselves are good oxidizers, so that'd lead to further corrosion. And looking further, both NiMH and NiCd cells typically use Potassium Hydroxide electrolyte (as do non-rechargeable alkaline batteries), so that's almost definitely the case for all the corrosion I've seen from battery leaks on old boards.

Using cylindrical Lithium Iron Phosphate batteries might also work, but I'm not sure if they fare any better charging-tolerance wise. (though a single AA type holder could be used in this case for a 3.2V cell, but also 2 in series could be used and just consistently undercharged at 4~5V)

And a side-note about non-lithium batteries: NiCd (especially new production ones) tolerate full discharge well and are stable in storage at/near full discharge, but can develop problems from repeated cycles of partial discharge and charge (namely lost capacity, but in those 3-cell configurations, repeated drop to around 50% nominal cell voltage ~1.6V could increase the risk for a shorted or reversed cell in the circuit, and lead to venting/leaking of the remaining 1 or 2 cells when they're charged) but usually perform fine in full/near-full discharge scenarios if they're not allowed to short circuit or suffer similarly high drain (enough to overheat and/or reverse polarity internally and generate hydrogen and/or steam sufficient to start venting).

NiMH cells OTOH do fine if partially discharged and charged or just constantly topped up and don't like to be fully discharged (more like lead-acid cells).

So in theory, NiCd cells should do better in long-term storage, discharging very slowly to a low positive voltage, but plenty of real-world scenarios obviously contradict that. (potentially getting too damp/humid in storage and/or too hot, or were already ruptured during powered use before being stored, or old/cheap cell casings just degraded/corroded internally over time and ruptured on their own that way) Plus a lot of them were probably exposed to poorly regulated 5V charging circuits at some point in time (or relying on the PSU's voltage regulation alone) which would jump-start degradation.

3x NiMH cells would pretty well tolerate 5V line voltage and can often charge to an open circuit voltage of 4.2~4.5V, plus scenarios with 1 dead/shorted cell would still be at 2.8~3V, which wouldn't be nearly as far off as 2 NiCd cells (closer to 2.2~2.4V), and in better-designed circuits actually capping the voltage to 3.6~4.0V, a 2-good-cell NiMH battery with non-leaking dead cell would probably hold up fine. (so similar cases where ML and possibly LIR batteries would also be safe)

for the life of me I cannot understand why anyone would use ANYTHING but a lithium primary connected to the external battery header

they last for 5-10 years, and no stupid recharging BS and leaking filth to deal with

Why not? If it works (the trick with the ML2032) you can have the cell on board with its cell clip, and avoid to have a big dumb battery holder attached to the board. I mean, is the same as with the Dallas Realtime epoxed chips... With the incompatible versions, iIf there would be a way to extract the chip inside these blocks to so them could be soldered to an small board with the cell clip and the crystal, I would like to try the method, instead having a dremel-ed ugly block attached to the board..

the big dumb battery holders last ~10 years, the coin cells do not

Question of tastes. Durability matters for you. Not having extra things attached to the board matters for others.

I've also noticed some boards allow the onboard battery to be configured as rechargeable or primary/non-rechargeable by placing/removing a jumper on the external battery header.

I'm not sure if any other manufacturers did it, but PC Chips did for some (or all?) of their 286 boards. I'd noticed the middle 2 pins on my M205's external battery header were jumpered (and had been since I got it ~8 years ago), and looked around, finding this:
http://www.vcfed.org/forum/archive/index.php/t-55934.html

About the battery: I placed a CR2032 holder in the same original place the barrel battery (that is to say, the internal battery) […]
Show full quote

About the battery:
I placed a CR2032 holder in the same original place the barrel battery (that is to say, the internal battery) was, using the same contacts.

The 4-pin external battery header (CN10) plays an important role here. Through tests and trial and error I found out the following:

Pin 1 is the external battery +
Pin 2 carries 4,6v.
Pin 3 is connected to the internal battery positive (+)
Pin 4 is internal battery/external battery negative (-)

When I got the motherboard pins 2 and 3 were jumped, so that the internal battery would receive 4,6v and so would charge.

I removed this jumper and jumped pins 1 and 3, so now the motherboard receives 3V from the CR2032 button battery and saves the BIOS configuration successfully.

So the M205 and M209 should work that way at least (and I think M216 and 219 as well), and maybe some other boards (PC Chips 386/SX maybe?). The set-up also seems to support external rechargeable batteries as well.

My case is also weird since the CMOS RAM seems to hold data with no battery connected at all, and I'm not sure why. (the real-time clock doesn't work without a battery, but BIOS settings stay saved for indefinitely long periods) It uses what appears to be a normal Hitachi HM6118A SRAM+RTC chip, so I don't know what's going on unless there's enough board capacitance to keep the SRAM powered. (particularly if the SRAM and RTC logic have separate power pins, since the SRAM should be much lower drain) I suppose it could be an unusual case of embedded FERAM being used, or this board isn't using the CMOS RAM inside the RTC chip at all and has an SRAM or FERAM chip elsewhere on the board, but it's strange in any case. (I could try pulling the RTC chip and seeing if that clears the CMOS data)

My particular M205 might be an early model or just an oddball, unbranded example if that has anything to do with it.
It looks identical to this one: Re: 80286 BIOS image collection

Except there's no M205 printed on the board and the chipset isn't rebranded to PCChips-1, but has a CityGate CG logo on it with the same 4L50F2052 part number as the PCChips stamped version on most M205s and 209s (apparently a CG chipset with native UMA+EMS support unless the BIOS configuration is lying) and a pair of UMC UM8237AE-5 DIP40 DMA chips instead of the Samsung KS82C37A-80 chips. Hmm, oh duh, and mine has the second oscillator spot (for an asynch NPU clock?) populated with a 32 MHz crystal and the main one 40 MHz along with a 20 MHz Harris CPU. (I swear Apolloboy found a download of the manual somewhere, but I have no idea what sort of search lead to that given the lack of identifying features)

It's actually the same board mentioned here:
Building a 286 system
(god I forgot it was $5 at WeirdStuff ... that's lower than most as-is boards I remember asking about, but maybe they discounted it for the battery corrosion)

Backup SRAM doesn't take much power at all and I know plain old CR2032 cells can keep 2 or 8kB SRAMs (if not a fair bit larger) powered for decades (plenty of late 1980s and early 90s vintage NES cartridges with still-working backup batteries), but the realtime clock circuits are usually a much bigger burden on those cells. (the few game carts with RTC chips onboard tend to die within a few years, like with the Pokemon Gold and Silver carts; some other games might have used battery-backed clocks with separate/redundant batteries or EEPROM/flash and avoided that issue of dead save data) The 32kB BB-SRAM cartridges used for N64 save PAK (controller plug on modules) also seem to hold up pretty well. (in addition to games that used internal SRAM, though some used EEPROM/Flash by that point)

And oddly or ironically, Sega's Sonic 3 carts used FERAM (8kB Fujitsu chips iirc), but the particular chips have a higher failure rate than plain old BB-SRAM implementations of the same era (and are more of a pain to replace than a dead battery); I assume Sega got a good deal on some supply of FeRAM for whatever reason around 1993.

AFIK, no motherboards bothered to use separate RTC and CMOS RAM backup batteries, though that would've been really nice. (if those DALLAS style RTC+battery chips had done that internally, many of them would still probably save CMOS data even though the clock battery would be dead)
Same deal with some game consoles that doubled up RTC and BB-SRAM on a single shared battery (Sega Saturn did that, maybe some others) or the weirder case of using a rechargeable cell for battery backup when no RTC is used. (the Sega CD did that and those rechargeable cells seem to wear out/die faster than a normal CR2032 would doing the same job and without the same risk of leaking)

Some CR2032s seem to last well over 10 years without problem in certain 90s or 2000s era boards (seemed to be the case with most of my family's S370 and 462 boards) and some made it close to 20 years before finally giving errors on boot, so there's probably tons of variables depending on the board, use conditions, storage conditions, etc. (also whether a board tolerates old/weak batteries dropping to around 2V or even lower, but still retaining data; most seem to work fine on mostly-dead batteries that still put out 2.8V open-circuit) I'm not sure if newer boards fare similarly well or got worse again; there may have been a sweet spot around 1999-2004 for low-power RTC.

And as to using rechargeable batteries, I'd think the main practical appeal would be avoiding anything dangling off the external header while also not having to modify the original onboard battery circuit (or risk using a CR2032 in the existing circuit, potentially causing bulging/leaking).

I can think of lots of problems beyond plain annoyance external battery holders could cause, from thin, dangling wires getting easily broken or caught, or a typical coin cell holder being used and leaving the cell's surface exposed to potentially short on the board if the holder came loose. Not all headers may use a compatible pinout, and those that do aren't keyed, so it's still easily possible to plug the battery in backwards (which could be worse than that depending what actual pinout is used) Plus if you use some sort of tape to secure a battery pack/dongle on the board, you might end up with a degraded sticky and/or brittle mess down the road. (or the pack falling free at some point and causing problems)

Different problems depending if the board's going to be installed in a system most of the time or swapped out periodically, plus the dangling/falling off issue would tend to be a bigger problem for towers rather than horizontal floor/desktop type cases.

I'd much rather just hard solder on a 2032 sized battery holder to the board itself, especially if the original solder holes already match those or are really close (otherwise a bit of spacer material, maybe hot glue like a lot of the barrel batteries seem to have used, helps avoid strain on more tweaked/extended connections for a still mostly-flush mount). Then it's just a matter of deciding whether to install a rechargeable 2032 sized cell or standard CR2032 and whether you'll need to desolder something (or cut a leg, or trace) to disable the charging circuit or not even bother with that. (or the convenient jumper configuration option)
You could also add some extra hot glue (or probably better, some sort of flexible calk or sealant, though silicone can be tricky if it's the acid or ammonia based stuff) on the battery holder before installing it to help isolate possible leak/seep/wicking points.

Some boards already used 2032 sized rechargeable cells in the first place (pancake type soldered-on cells) and/or had multiple holes for different sized batteries. A few boards also got battery mounting spots along with DALLAS or similar integrated battery backup RTC chips. (I assume those can be used along with the DALLAS chips, but were probably originally intended to allow the same boards to use normal RTC chips and an external battery)

I've gone through several different options after starting this thread and settled on these as the most viable options:

a) Just connect a 3xAA or 3xAAA external battery holder
b) Buy this and solder it on.

Is there a risk of overcharging the batteries when using 3xAAA battery holder and rechargeable batteries in external header?

Not at all, if the holder is connected where the barrel battery was. But NiCd and NiMH have high self-discharge rate, so if the system doesn't see at least weekly use I'd say you'll find it often losing cmos settings.

quicknick wrote on 2020-02-05, 08:37:

Not at all, if the holder is connected where the barrel battery was. But NiCd and NiMH have high self-discharge rate, so if the system doesn't see at least weekly use I'd say you'll find it often losing cmos settings.

I was talking about the external battery header. In my case the motherboard doesn't have any battery on board just the header. There is 5V between the header pins so the voltage is higher than the voltage of 3xAAA rechargeable batteries have. I used 3xduracell recharge ultra aaa batteries in a holder connected to the battery holder. If you believe the ads the batteries are suppose to be able to hold their charge at least 12 months when not in use.

Baoran wrote on 2020-02-05, 08:54:

quicknick wrote on 2020-02-05, 08:37:

Not at all, if the holder is connected where the barrel battery was. But NiCd and NiMH have high self-discharge rate, so if the system doesn't see at least weekly use I'd say you'll find it often losing cmos settings.

I was talking about the external battery header. In my case the motherboard doesn't have any battery on board just the header. There is 5V between the header pins so the voltage is higher than the voltage of 3xAAA rechargeable batteries have. I used 3xduracell recharge ultra aaa batteries in a holder connected to the battery holder. If you believe the ads the batteries are suppose to be able to hold their charge at least 12 months when not in use.

I have not seen any EXT_BAT headers that recharge, though I hear there are some in mostly XT and 286 boards..

The 5 volts you measure across that header are most likely because of the reverse leakage of the diode that should prevent the external battery being charged. Connect your multimeter in series with the battery pack and measure the charging current. If it is around 2-4mA your batteries are getting charged, if it's in the microamp range it's the reverse leakage current.

quicknick wrote on 2020-02-05, 10:18:

The 5 volts you measure across that header are most likely because of the reverse leakage of the diode that should prevent the external battery being charged. Connect your multimeter in series with the battery pack and measure the charging current. If it is around 2-4mA your batteries are getting charged, if it's in the microamp range it's the reverse leakage current.

It is bit difficult to connect multimeter series with the battery holder because of that 4 pin connector I soldered to it. I did connect the connector to the external battery header as if it was recharging. Negative side of the battery holder to ground and positive to the pin that had +5V in the header. It seems to be working fine, but I will see if I can figure some way to do the connection in series. It is the motherboard I mentioned earlier in this thread, so perhaps it recharges using the header because it doesn't have any on board battery at all.

Ok forget about "years" lets talk real numbers:
a CR2032 has a capacity of ~200mAh give or take at slow discharge rates
a lithium AA has a capacity of about 3000mAh at similiar discharge rates

So we are talking about a lifetime improvement of an order of magnitude. I can rest easy knowing that when I shove some lithium AAs into a PC I'm building that I will very likely never have to touch them again. Maintenance is a scourge

If you have to replace a CR2032 cell every 10 years (and that's assuming a pretty outrageously quick discharge compared to most of what I've seen), that's probably the least of your worries. Anybody here not open their retro PC case for 10 years to do something? Anybody? I barely go 3 months before I have to swap sound cards or install a different hard drive.

I absolutely hate doing the same thing twice, doesn't matter if I'm in there for something else I don't want to fuck with it

It's a hassle if you don't have the right batteries on hand (and not misplaced) when you need them. Honestly that's usually a bigger issue in my experience than the trouble of just opening the case and getting to the battery socket/holder.

Also, some CMOS RAM seems to tolerate very low voltage levels and holds data even with a weak (like 0.9V) charge. Though voltage that low on a NiMH cell is usually bad (not as bad as for Li-ion cells), but full discharge of NiCd cells is much less so. (they tend to do well in long-term storage in the near-fully-discharged state and have more problems if left to slowly self-discharge)

Those cases are also dependent on very low current draw from the CMOS RAM circuit (and realtime clock if present). My PCChips/ILON/etc 286 and 386SX boards seem to behave that way and are also among the few I have that hold CMOS data for several minutes with no power applied at all (just board-level capacitance), though my M205 286 board does that better than the M396. (though user-configured hard drive parameters seem to be saved indefinitely even when other settings are lost; that seems common to 486 boards of the same period, at least using the AMI BIOS ... not quite sure how that works unless there's a tiny bit of EEPROM present)

Those boards also have the easy jumper-selected charging circuit, so easy set-up for a CR2032 installation, though case in point to misplacing things: I couldn't find my set of coin cell sockets, but did have a working barrel battery and soldering station on-hand, so that M205 has the latter in it for now. (I'll probably change that when I pull it out of the case it's mounted in, but for now it works)

appiah4 wrote on 2020-02-05, 06:48:

I've gone through several different options after starting this thread and settled on these as the most viable options:

a) Just connect a 3xAA or 3xAAA external battery holder
b) Buy this and solder it on.

Good choice ! You do not want to use Lithium rechargeable on an old motherboard as they were mostly designed for NiCad. The charge circuit for Lithium rechargeable is different and requires circuitry to monitor charge level and stop the charge as the battery achieves full charge. Ever ask why it is Lithium rechargeable batteries (in cell phones, etc) that explode and or start a fire but you never hear about NiCads doing that. Also "Can I replace NiCd with lithium ion? The earliest lithium-ion batteries were not backward-compatible with nicad battery tools, but that's changed over time. ... Due to advances in battery circuitry, however, chargers are not forward-compatible; the charger that came with a nicad tool will not work with lithium-ion batteries."

Some mobos have a jumper setting (could be factory soldered though) for a non-rechargeable cells, these won't explode but also - depending on how the diodes are wired - might be drained even with PSU voltages being present.
In general these days I just remove the battery completly and connect a 3xAA or AAA battery pack via the "external "header. Which might also require replacing if the original NiCd spill has corroded it by the way. Mobos are stored short-term with that pack connected so that I don't loose BIOS HDD settings but for long-term storage I remove it. I'd rather play around with BIOS for 10 minutes, even if I forget the HDD params, than have to deal with another spill and corrosion damage.

AA/AAA battery packs have many advantages, easy replacement, easy installation (unless it's a laptop) since wires can be long, no soldering, and 4x pack will work too, even 2x might work with fresh batteries. No issues with charging, or at the very least I've not seen a mobo that could recharge via the connector.

Deunan wrote on 2020-03-01, 11:52:

Some mobos have a jumper setting (could be factory soldered though) for a non-rechargeable cells, these won't explode but also - depending on how the diodes are wired - might be drained even with PSU voltages being present.
In general these days I just remove the battery completly and connect a 3xAA or AAA battery pack via the "external "header. Which might also require replacing if the original NiCd spill has corroded it by the way. Mobos are stored short-term with that pack connected so that I don't loose BIOS HDD settings but for long-term storage I remove it. I'd rather play around with BIOS for 10 minutes, even if I forget the HDD params, than have to deal with another spill and corrosion damage.

AA/AAA battery packs have many advantages, easy replacement, easy installation (unless it's a laptop) since wires can be long, no soldering, and 4x pack will work too, even 2x might work with fresh batteries. No issues with charging, or at the very least I've not seen a mobo that could recharge via the connector.

if you use energizer lithium AAs you can leave them in there forever and never worry about leakage