Odin/Dallas RTC modules with integrated battery (WHL #107)
Have you ever desoldered one of those ancient RTC modules from an equally ancient PC mainboard?
There was a time in the 1990s where board manufacturers abandoned the old leaky NiCd buffering cells and instead of hopping onto the CR2032 lithium train. They used Dallas modules (back when Dallas was still Dallas and not Maxim…which also ceased to exist in 2021 when they were bought by Analog) that served as real time clocks and battery backed SRAM for some 100 bytes of BIOS settings.
Well yeah, these are terrible as the battery isn’t replaceable since it is moulded into the bulky case. One can’t get in, and one can’t get them out either.
Common models were the Dallas DS1287, DS12885, DS12887, DS12B887, DS1387, DS14287, as well as the knockoff ODIN OEC12C887, OEC12C887A, VIA VT82887, and newer Benchmarq BQ3287, BQ4287. The modernized DS12C887/DS12C887 and DS12C887A/DS12C887A+ (+ being RoHS compliant, chip is the same) are still available today from reputable sellers, so I would assume they’re still in production and not NOS with 20+ years of age. Analog didn’t care to refresh the datasheet and purge the Maxim OR EVEN THE DALLAS NAME from it, but that’s just Analog being Analog.
Well anyway, I got one of these Odin modules on a Pentium 1 board as shown above, and since the default BIOS settings declare “none” on all four PATA channels instead of just “Auto”, and the darn thing resets and loses data right after “saving” these settings, I’m out of luck and need to replace the module.
So I desoldered it.
And then I realized that this was likely socketed from the factory.
I’m an idiot.
Date code 9618 matches up with 9615 from some UMC chip, 9622 from two TMtechs and late 95s from others. That one however probably isn’t a date code, or did I miss the Undecember/Duodecember in 1995 to make a 9557 datecode? 
On the bright side, there are open source projects that do combine a compatible chip with an accessible lithium battery, namely the ones from Necroware, which are also remixed with the more common (=cheaper, higher capacity and better availability) full-size CR2023 BIOS cells. These go for around 20 to 25€ shipped from other (non-commercial) enthusiasts via eBay, or one can spin their own with the provided schematics and Gerbers.
Or, and that’s the route I’m willing to go for the moment: There are also knockoffs available on AliExpress starting at 2.50€ a piece with choice shipping (free with a total order value of at least 10€ across all items). I don’t know for sure and I don’t really care which chip they actually use, injecting code via some ancient RTC protocol into an offline DOS machine is not a security concern to me. They might have a terrible operation life, but I need a working BIOS for board and card testing for a couple of weeks to months, and once for sale, I will include the original chip and a detailed description on how to deal with this issue going forward.
There also is the option to file down the case of the original RTC module to be able to solder onto the original battery contacts, but boy does that look ugly and botched. I’d rather spend 3 bucks for a one-year temporary solution than present that hack to a potential buyer of the board.
Will report back once the module has arrived and is installed.
I encountered one of these Dallas devices when a former employer called me and asked if I could fix her computer. It was a 386. I examined it, found the Dallas RTC module, and I replaced it, Dallas still being in business at the time. It may have had a cable-tie around it to hold it in place. That fixed the computer, and I never heard any more about it and assume she has moved on to better computers since then.
If you decide to take apart the old module, be careful around that lithium battery.
I don’ think a fully depleted lithium battery of any chemistry (MnO2, SOCl2, FeS) poses a risk at that size, but I do extract live 3-ish volt lithium cells on a regular basis. Like for Christmas I disassembled a 6S10P 18650 e-bike battery that had a dead row at the end. Solid nickel tabs all around, four proper welds per cell end. That was a solid four hour thrill ride, they are still very spicy in 10P configuration, even drained down to sub-10% state of charge
The original RTC module is still here and untouched, but I have my doubts about the replacement that showed up very recently…
[…] Two weeks ago, I wrote about a Dallas/Odin RTC module with integrated battery that caused me some trouble on an old Pentium board. The BIOS defaults to no hard disks, and the module isn’t rechargeable, so once it’s empty, the board will no longer be able to boot directly from hard disk. So I checked my options…and bought from AliExpress, because I was feeling adventurous […]