Using a Logic Analyzer to Debug X68000 Bus Timing
The Sharp X68000 is a rewarding machine to repair because its signals are accessible, its architecture is well documented, and a fault can often be narrowed down to a particular bus cycle. When a game loads to a blank screen, an expansion board fails to respond, or the system resets unpredictably, a logic analyzer can show whether the 68000 is issuing valid addresses and receiving the expected acknowledgements.
This work is especially useful on Japanese machines imported into Australia, where replacement boards and donor systems are less common than in Japan. A careful timing investigation can save a scarce motherboard from unnecessary chip replacement. It also helps distinguish a software fault from a weak power rail, a damaged buffer, a missing pull-up, or a peripheral that is holding the bus incorrectly.
Preparing The X68000 And Test Equipment
Begin with a stable machine. Bus timing measurements taken from a computer with a noisy or sagging supply can lead to false conclusions. The original X68000 power supply is designed for Japanese mains, nominally 100 V, while Australian outlets supply approximately 230–240 V. Use a properly rated step-down transformer and confirm the output with a multimeter before connecting the computer. A cheap travel adaptor is not a substitute for voltage conversion.
A logic analyzer with at least 16 channels is practical for basic work. You will want to observe several address lines, data lines, and control signals at the same time. A sample rate of 50 MHz is usable for slower activity, while 100 MHz or higher gives more confidence around signal edges. The analyzer should support 5 V TTL inputs, or have a suitable level interface. Many inexpensive USB analyzers are advertised with high sample rates but use limited-quality input circuitry, so verify their threshold specifications before trusting a marginal waveform.
Use a short ground connection between the analyzer and the X68000. Long flying leads can pick up noise and make a clean signal appear unstable. Connect probes with the machine powered down, avoid slipping between adjacent pins, and secure the cables so movement cannot short the board. If the case must remain open during testing, keep mains wiring enclosed and work through an isolation and current-limited setup where appropriate.
Finding The Right Bus Signals
The 68000 family uses a multiplexed control process in which the processor presents an address, asserts control lines, and waits for a device to complete the cycle. Useful signals include address strobe, read/write, upper and lower data strobes, data acknowledge, and several address bits. Depending on the board revision and test point access, you may monitor the CPU pins directly or use buffers and expansion connectors that expose the bus more conveniently.
Address strobe, commonly identified as AS, marks the valid portion of a bus cycle. R/W indicates the direction of transfer, while UDS and LDS select the upper and lower byte lanes. DTACK tells the processor that the addressed device has completed the transfer. A normal memory read should show a valid address followed by the appropriate strobes and then an acknowledgement. If DTACK never arrives, the CPU may remain waiting indefinitely, which explains a frozen boot screen or a machine that appears dead after reset.
The X68000 also uses system-specific decoding and peripheral timing around the 68000 bus. ROM, RAM, video hardware, sound hardware, and expansion devices may have different response delays. Do not assume that every cycle has identical spacing. First capture a known-good operation, such as reset activity or a ROM read, and use it as a reference. If you have a second working X68000, even a different model, it can provide useful comparative evidence while allowing for clock and board-revision differences.
Capturing A Repeatable Bus Cycle
Set the analyzer to trigger on a stable event rather than on any random transition. A falling edge on AS is a sensible starting point. For reset debugging, trigger on the release of RESET and capture several milliseconds around it. For a suspected expansion problem, trigger on a decoded address range or capture repeatedly while the failing software accesses the board.
Sampling should be fast enough to show the relationship between control lines, not merely their average state. At a 10 MHz processor clock, a full clock period is around 100 nanoseconds, but bus phases do not necessarily align neatly with the analyzer’s sample intervals. Use a suitable pre-trigger buffer so the address and control activity leading into the failure are visible. A capture that starts after AS has already changed may hide the important part of the cycle.
Keep the first capture small and understandable. Monitoring every CPU pin can produce a large waveform that is difficult to interpret. Start with AS, R/W, UDS, LDS, DTACK, RESET, a clock reference, and a few address lines. Add data lines once the cycle structure is clear. If the analyzer software supports protocol decoders, treat them as a convenience rather than an authority; a decoder can label a malformed cycle as a valid transaction if its assumptions are too broad.
Interpreting Common Failure Patterns
A missing DTACK is one of the most useful observations. If AS and the address bus look healthy but DTACK remains inactive, inspect address decoding, the selected device, buffer enable signals, and the device’s own supply. A broken trace or an open logic gate can prevent the acknowledgement from reaching the CPU. If DTACK is permanently asserted, the processor may complete every cycle too quickly or read invalid data, depending on the circuit.
Incorrect UDS or LDS behaviour can point towards a damaged data-buffer control path. A byte access should select the appropriate lane, while a word access generally involves both lanes. Compare read and write operations at the same address. If the address is correct but the data changes before the expected sampling point, investigate bus contention, a weak transceiver, or excessive capacitance on the line.
A timing fault may also be power-related. Electrolytic capacitors that have aged can increase ripple or reduce stability, especially during disk access or when several peripherals start together. Before interpreting subtle nanosecond differences, inspect the supply rails and known failure points; the motherboard capacitor work provides useful background for this stage of preventative repair.
| Capture pattern | Likely area to inspect | Useful next measurement |
|---|---|---|
| AS and address activity, no DTACK | Decoder, selected device, DTACK path | Probe chip-select and acknowledge inputs |
| DTACK stuck active | Shorted logic, failed buffer, incorrect pull-up | Check DTACK at the CPU and source device |
| Valid reads with corrupted data | RAM, ROM, data transceiver, contention | Compare CPU-side and memory-side data lines |
| UDS or LDS missing | Byte-lane decode and buffer enables | Capture both strobes during word and byte accesses |
| RESET repeatedly asserted | Supply supervisor, reset circuit, power fault | Monitor RESET alongside 5 V during startup |
| Clean CPU cycle but no expansion response | Connector, address decode, expansion-board logic | Trace the selected address and board acknowledge |
The table is a starting point rather than a diagnosis. A signal that looks late may be correct for that device class, and an apparently noisy line may be a probing artefact. Measure the same signal at both ends of a path whenever possible. If DTACK is present at an expansion board but absent at the CPU, the fault is in the return path, buffer, connector, or track between those points.
Comparing Good And Bad Captures
A known-good capture is more valuable than a theoretical timing diagram because it includes the actual board revision, clock source, and loading conditions. Save captures with the machine model, test software, probe points, and supply measurements. The X68000 family includes several revisions, and a timing difference between an ACE, XVI, Compact, or earlier model may be normal rather than evidence of damage.
For a machine that boots intermittently, capture the same address region during both successful and failed starts. Look for the first divergence. A failed read may show the correct address and control strobes but a different data value. Alternatively, the CPU may stop issuing cycles because RESET is being asserted again, a bus error is occurring, or a peripheral has not released a control line.
Software can help produce repeatable activity. A small diagnostic program that reads a fixed ROM location, writes a pattern to RAM, or polls an expansion register makes it easier to trigger and compare captures. Keep the test routine simple and avoid relying on the subsystem under investigation. For example, testing a memory expansion through a program loaded from that same expansion can obscure whether the fault is in loading, execution, or the target address range.
Australian repairers often need to work around limited local stock. A replacement 74-series logic IC may be available through Jaycar or an electronics distributor, while an exact Sharp custom chip may need to come from a donor board purchased through eBay Australia, Gumtree, or a Japanese auction proxy. A timing capture can justify replacing a common buffer first instead of risking a scarce custom LSI.
Avoiding Measurement Errors
Logic analyzers show digital decisions, not the full electrical waveform. A line with ringing may cross the analyzer threshold several times, producing apparent glitches that are not valid logic transitions. If a signal is borderline, check it with an oscilloscope if available. Examine rise time, overshoot, undershoot, and the voltage level at the receiving device rather than relying solely on a decoded trace.
Probe loading matters on old hardware. CMOS inputs may tolerate a light connection, but a crowded test point, long ground lead, or accidental contact with a neighbouring pin can alter the very timing being measured. Use spring grounds or short ground clips, keep wires separated, and avoid attaching a large bundle to a delicate expansion connector. Photograph the setup so it can be reconstructed after the board has been moved.
Do not perform resistance or continuity checks on a powered board, and never connect an analyzer ground to an unknown mains-referenced point. The computer’s low-voltage ground is not the same as making the open power supply safe. If the supply needs repair, disconnect it from the motherboard and follow appropriate Australian electrical safety requirements. Work on a bench in Sydney, Melbourne, Brisbane, or elsewhere with the same discipline used for any vintage computer carrying hazardous voltages.
Turning Timing Evidence Into A Repair
Once a faulty transaction is identified, divide the circuit into stages: CPU output, address decoding, device selection, data transfer, and acknowledgement. Probe each stage in sequence. This prevents random chip swapping and shows whether the failure is generated at the source or introduced along the route. A logic probe can confirm a stuck state, but the analyzer reveals when that state becomes wrong.
After replacing a component or repairing a trace, repeat the original capture under identical conditions. Confirm that the repaired cycle now has the expected address, byte selection, data stability, and DTACK timing. Then test cold startup, warm reset, floppy or hard-disk access, sound, video, and any fitted expansion board. A repair that passes one bus transaction may still fail when temperature, loading, or access order changes.
Keep the original captures and notes with the machine. Record the analyzer model, sample rate, probe locations, clock frequency, supply voltage, and software used to generate the activity. Such records are valuable when a future fault appears and contribute to the wider preservation effort around the X68000. They can also help another enthusiast in Perth, Adelaide, or regional Australia avoid repeating the same investigation.
A logic analyzer turns an apparently mysterious X68000 failure into a sequence of observable events. Start with safe power, clean probing, and a known-good reference; then follow AS, address selection, byte strobes, data, and DTACK through the failing cycle. Share well-documented captures and repair notes with the X68000 community so these techniques remain available for the next rare motherboard that needs a second life.
Nereid-X Expansion Board
A personally-produced LAN+USB+Memory expansion board for Sharp X68000 series computers. Multiple production runs were offered, including a final batch and a later revival reproduction run.
Power Supply Repair
X68 power supply repair and modification services were offered by the site owner, with documentation shared through diary entries spanning 2001–2006.
Server & Networking
Notes on FreeBSD administration, ISP changes, server migration, and networking topics. The site itself ran on FreeBSD with the hns diary system and Namazu search integration.
Get in touch
X68K.NET connects Sharp X68000 enthusiasts through community links and shared projects. Reach out with questions about the Nereid project or X68 resources.