Repairing the X68000 Real Time Clock with DS3231 and I2C

The Sharp X68000 keeps system time through a small Real Time Clock circuit on its mainboard. After three decades, those original RTC chips often drift, lose time when the machine is powered down, or fail outright because the backup battery has leaked across the board surface. Replacing the stock module with a DS3231 breakout gives the machine a temperature compensated oscillator that holds time accurately across long shutdowns and warm Australian summers alike, which matters when the X68000 is doing development work, running ZMUSIC playlists, or hosting scheduled demos.

A typical replacement pulls together four ingredients: a DS3231 module, four jumper wires, a small Philips screwdriver, and an I2C driver that speaks the same protocol the original RTC used. The procedure does not require hot air rework or a soldering microscope, which makes it one of the more approachable hardware jobs on the platform. Pulling the lid off the chassis, identifying the RTC location, and tapping into the existing I2C lines is something a careful hobbyist in Brisbane, Perth, or Hobart can finish in a single afternoon.

Diagnosing RTC Failure in the Sharp X68000

The factory Real Time Clock in early X68000 models is a discrete design wired around a 32.768 kHz crystal and a handful of CMOS logic gates. Timekeeping depends on the crystal trim, on a charging circuit for the internal NiCd or lithium cell, and on stable voltages when mains power drops away. Three decades of capacitor drift, board flex, and battery leakage take their toll. X68K enthusiasts have documented multiple failure modes: clocks that reset to midnight on every boot, dates that jump forward by years, and boards with green corrosion creeping under the original battery holder near the MPU area.

When the backup cell leaks, KOH electrolyte creeps along the legs of adjacent ICs and eats through copper traces. A machine that still boots cleanly in Melbourne can lose the year by mid afternoon because the corroded trace behaves like a slow resistor, dragging the RTC supply voltage below its hold threshold. Restoration work normally starts with neutralising the residue using isopropyl alcohol and a soft brush, then verifying the I2C lines still respond with a logic analyser before any new module goes in.

The other common fault is plain old frequency drift. Crystals age, and a 32.768 kHz part that ran within five parts per million when new can run tens of PPM off after twenty years of service. Over a year that adds up to several minutes of accumulated error, which is enough to confuse Human68k when stamping file timestamps during a long compile session. A swapped in module with active temperature compensation removes that drift almost entirely, leaving only the much smaller intrinsic error of the DS3231 chip itself.

What Makes the DS3231 Different from a Stock Crystal Module

The DS3231 is a drop in upgrade over stock RTC implementations because it integrates the crystal, the oscillator trimming network, and the digital counter inside one package. Maxim, now part of Analog Devices, designed it for loggers, point of sale terminals, and any product where an external crystal would be too inaccurate. The chip measures die temperature and adjusts its load capacitance on the fly, which holds time within two minutes per year across the full commercial temperature range. For an Australian machine that lives through hot Canberra summers and cold Canberra winters in a home office, that compensation matters in a way it would not in a climate controlled server room.

Several hobby boards expose the bare part, but most Australian builders use a pre assembled module from Jaycar or Altronics that adds an AT24C32 EEPROM, a CR2032 battery holder, and a charging circuit for that cell. The module routes the I2C lines through 10k pull up resistors already fitted on the board, so a pack of four female to female DuPont leads is enough to hook everything up. Price at the till in Sydney or Adelaide sits between eight and fifteen AUD depending on brand, and an invoice from a local seller keeps freight faster than waiting on a slow boat from Akihabara or relying on a friend forwarding a care package.

Power draw is one of the quieter wins. The DS3231 sips around three microamps from its backup cell while mains power is absent, which means a fresh CR2032 will keep the clock running for the better part of a decade between swaps. The original X68000 RTC would not hold date settings for more than a few months once its internal NiCd dried out, so the runtime gain feels almost generous by comparison and removes a recurring item from the maintenance checklist.

Connecting the I2C Bus to the X68000 Mainboard

Hooking the DS3231 to the X68000 involves four wires: ground, plus 5V from a standby rail, SDA, and SCL. On most ACE and later CZ models, the existing battery backed RTC socket carries those lines already, and the easiest path is to tap the underside of that footprint. Early XVI boards route the lines through a different connector, so builders hunting for the test points often grab the service manual PDF and trace the 32.768 kHz crystal back to its associated CMOS gate.

The I2C address the DS3231 responds to is 0x68, which is the same address the original RTC chip answered, so software that already polled the clock will generally find the replacement without any modification. Pull up resistors matter because the X68000 mainboard does not provide strong ones on those lines, and the DS3231 module usually supplies them on its own PCB. Wiring should be kept short to avoid ringing on the rising edges of SCL, and a ground return that shares the same return path as the mainboard avoids ground loops when the system is also driving its floppy bus at speed.

Once the module is mechanically parked inside the chassis with a small piece of double sided foam, it stays put through Australia Post shipping and through being moved between meetups. Builders in Adelaide retro computing groups have started using hot glue dots on the corners of the PCB instead of foam, which is more permanent but still serviceable. Either approach works, and a label stuck to the inside of the lid helps anyone opening the machine later understand why an unfamiliar chip is sitting near the original RTC footprint. Routing the leads along existing cable runs also keeps them clear of the cooling fan, where vibration would eventually fatigue the wires.

Sourcing Components and Initial Calibration Locally

A first build costs more in time than in parts. Common modules available at local outlets come with one of two potential pitfalls: counterfeit DS3231 chips that drift faster than the genuine part, or modules with an absent charging resistor that overdrives the CR2032 over months of constant charge. Spotting a fake without an oscilloscope is hard, but a sanity check is to log the seconds register over a 24 hour window and see whether the machine loses more than a few seconds relative to NTP. Modules sold by Aus Electronics Direct and most of the larger Jaycar branches tend to use genuine silicon, but anything ordered from a marketplace stall warrants a quick paper test before trusting it with another decade of timekeeping.

Drivers vary depending on what software stack the machine is running. Human68k does not expose a generic I2C driver on its own, so the common pattern is to pair the DS3231 with an attached microcontroller that pumps the time across a serial port, or to use the Nereid-X expansion to expose the RTC register block over a shared memory window. The Nereid-X path is attractive because it lets an attached PC tool set the clock at boot time over USB, which is faster than fighting with the little CONFIG.SYS editor for date entry every time the clock drifts.

A good first test is to set the time, cycle mains power ten times, and confirm the clock survives each cycle without resetting. After that, leaving the machine powered down for a week and reading the time back on power up verifies the backup cell really is delivering current and the wiring holds. Builders have reported a few weeks of accumulated error under two seconds across more than a year of cycling, which is roughly the rated performance of the chip and a substantial step up from what the stock module managed. Logging the drift in a spreadsheet gives a baseline for future comparisons if a new firmware revision or a different module brand is tried down the track.

Comparing Stock and Replacement RTC Options

Replacing the X68000 RTC does not require locking into a single solution. The choice depends on how accurate the time needs to be, how much labour the owner wants to put in, and whether an expansion board is already installed. A machine that only needs rough date stamping can stay on the original circuit after cleaning and battery replacement, while a development workstation benefits from a digital module with active compensation.

Solution	Typical Accuracy	Backup Battery	I2C Address	Skill Needed	Approx Cost (AUD)
Original discrete RTC	±20 ppm or worse	Internal NiCd, prone to leak	0x68	None	$0 (already in machine)
DS3231 module with EEPROM	±2 ppm	External CR2032, easily swapped	0x68	Four jumper wires	$8 to $15
DS3231 plus Nereid-X bridge	±2 ppm	External CR2032	0x68	Wiring and configuration	$25 to $40 total
Network time via LAN board	Network dependent	None	n/a	Software only	Free if LAN hardware exists

For a hobbyist machine that only needs to keep the date roughly right, leaving the original circuit alone is acceptable as long as the battery has been checked for leakage and the board is clean. For development work, scheduled captures, or anything that stamps files with reliable timestamps, the DS3231 module delivers the accuracy and the runtime without forcing a full mainboard rework. Pairing the module with the Nereid-X adds the ability to set the clock from a connected PC toolchain, which keeps the workflow close to what modern developers expect from a network attached workstation.

Retro computing preservation in Australia leans heavily on small communities that share schematics, repair logs, and sometimes parts. The Melbourne retro computing group and the Sydney Japanese computer collectors run regular meetups where an afternoon repair can be wrapped up with borrowed equipment and second opinions. Anyone tackling an RTC swap for the first time can often find a contact in those groups willing to walk through the wiring over a video call, which makes the whole job feel less lonely than it does late at night, on the floor of a home office, with a torch clamped between the teeth.

For full build notes, schematics, and a longer write up of the DS3231 swap procedure with photos, the full project documentation is where most of the technical depth on the platform lives. Drop a message through the contact form if you want to share your own calibration results or add a note about the parts that worked under your conditions.