Building a SCSI bus analyzer for the X68000

The Sharp X68000 relies on SCSI for hard disks, magneto-optical drives, scanners, removable media, and a wide range of external storage devices. When a system refuses to boot, loses a drive during access, or reports intermittent read errors, the visible symptom rarely identifies the actual fault. The problem may be a damaged cable, incorrect termination, a stuck phase signal, a device that never asserts REQ, or software waiting for a response that the bus never receives.

A SCSI bus analyzer turns those invisible electrical and protocol events into a useful trace. It records the eight data lines, parity, control signals, and timing relationships while the X68000 communicates with a peripheral. The resulting capture can show whether a command was issued correctly, whether the target responded, and exactly where a transaction stopped.

A practical analyzer does not need to become a full SCSI controller. Its first job is observation. By remaining electrically passive and concentrating on reliable timestamped sampling, a home-built instrument can answer difficult repair questions without adding another uncertain device to the bus.

Understand the X68000 SCSI environment

The original SCSI bus used by many X68000 configurations is an 8-bit, single-ended parallel bus. Its signals are generally active low and use open-collector or open-drain behavior, allowing several devices to pull a line low while termination supplies the high level. The main control signals include BSY, SEL, C/D, I/O, MSG, REQ, ACK, ATN, and RST. The data group contains D0 through D7 and a parity line.

SCSI communication is divided into phases. During arbitration, an initiator or target competes for control of the bus. Selection identifies the intended device, command and message phases exchange protocol information, and data phases transfer payload bytes. Status and message-in phases complete the transaction. The C/D, I/O, and MSG lines form the phase code, while REQ and ACK provide the request and acknowledge handshake.

This matters because a logic analyzer configured only for ordinary digital buses can misread the activity. A SCSI line may remain high through a passive pull-up or terminator and drop low only when a device asserts it. A transition is meaningful in relation to the other lines and the current phase. The analyzer therefore needs enough channels to capture all signals together, preferably with parity and the reset line included.

SCSI timing also varies by device and transfer mode. A basic asynchronous transaction can be understood at relatively modest sample rates, but synchronous transfers may use shorter REQ-to-ACK intervals. The instrument should sample considerably faster than the narrowest pulse it needs to distinguish. A 20 MHz or faster digital capture path is a sensible starting point for a diagnostic tool, while deeper memory is often more valuable than an extreme sample rate.

Design for passive observation

The safest connection method is a short interposer placed between the X68000 and its SCSI cable or peripheral. The analyzer taps each signal through a high-impedance input rather than driving the bus. Avoid ordinary indicator LEDs connected directly to SCSI lines: even a small load can alter a marginal bus, and a visible LED circuit may introduce unexpected current paths between devices.

A suitable front end can use fast CMOS buffers with input protection, provided the selected parts tolerate the bus voltage and do not clamp normal SCSI levels. Series resistors at the analyzer connector can limit fault current, while carefully chosen protection devices guard against static discharge. The protection network should not add enough capacitance to soften REQ, ACK, or arbitration edges.

The analyzer should have a clear ground reference connected to the SCSI ground conductors. Keep the ground lead short and use a ground plane or broad return path on the circuit board. Long flying wires are convenient for a first prototype but can pick up noise and create ringing. A keyed connector and strain relief are worthwhile because an incorrectly aligned SCSI tap can damage equipment.

Powering the analyzer independently is preferable. A USB supply or isolated bench supply prevents the instrument from consuming current from the X68000 or changing the voltage drop along an already questionable cable. If the analyzer shares ground with a computer used for USB capture, check for ground loops before connecting vintage equipment. A resettable fuse and a clearly marked power switch add inexpensive protection.

Before attaching the instrument to a working machine, test every input with a signal source or a known-good logic device. Confirm that an idle high bus appears high, that a pulled-low signal is detected cleanly, and that no analyzer input presents a measurable drive voltage. The analyzer should be an observer under all normal conditions.

Choose the capture architecture

There are two practical architectures. A microcontroller-based design is accessible and can provide USB, an SD card, or a serial console. Its weaknesses are interrupt latency, limited simultaneous sampling, and difficulty preserving exact timing when the processor is busy formatting data. It works well for slow diagnostic captures and event counters, but it may miss short handshake relationships.

An FPGA or CPLD design offers deterministic sampling and parallel trigger logic. An FPGA can store a compact record in block RAM, decode bus phases in real time, and stream captures through USB or an attached microcontroller. A CPLD may be enough for trigger generation and signal compression, with external RAM handling deeper storage. The development effort is greater, yet the result is better suited to timing-sensitive SCSI troubleshooting.

A useful compromise is a small FPGA with a USB bridge or UART-connected microcontroller. The FPGA samples the bus continuously into a circular buffer. When a trigger occurs, it preserves a configurable amount of pre-trigger and post-trigger data. The microcontroller then transfers the frozen capture to a host computer. This separation keeps acquisition timing independent from file handling.

Each sample need not be stored as a raw 16-bit word. A compact record can contain the eight data bits, parity, nine control signals, and a timestamp or elapsed-cycle value. For a first version, raw samples are easier to validate. Later firmware can perform run-length encoding, storing a signal value and the number of sample clocks for which it remained unchanged. Since many SCSI lines are idle for long periods, compression can greatly extend capture duration.

Timestamp resolution should be documented clearly. A sample clock gives a fixed grid, but a transition-timestamp design can offer more useful timing data with less memory. The decoder must distinguish the delay from REQ falling to ACK falling, the release of ACK, and the time before the next byte. These intervals often expose a slow or stalled device.

Decode transactions instead of only displaying waveforms

A waveform viewer is valuable, but a protocol-aware decoder turns a confusing trace into a sequence of events. The decoder should identify bus-free periods, arbitration, selection, reselection, and each SCSI phase. It can then group data bytes around REQ and ACK edges and label likely command descriptor blocks, status bytes, and messages.

For asynchronous transfers, the target generally asserts REQ when a byte is ready and the initiator responds by asserting ACK. The direction is indicated by I/O, while C/D and MSG identify the phase. The decoder must account for active-low logic and should display both the physical signal level and the logical meaning. Confusing “line low” with “signal asserted” is a common source of false interpretations.

Selection deserves special treatment. During selection, the initiator asserts BSY and SEL while placing its own ID and the target ID on the data lines. The selected target responds by asserting BSY and releasing SEL. A trace that shows SEL without a corresponding response points toward a target, cable, termination, or ID issue. A trace with two devices asserting conflicting IDs may indicate an installation error or hardware damage.

Parity errors should be reported, but not treated as proof of a single faulty chip. A bad cable, weak termination, ground noise, or a sampling edge that is too close to the data transition can all produce a parity mismatch. The tool should record the sampled data, parity bit, phase, and timing so the user can compare repeated failures.

A practical decoder can export CSV and a human-readable event log. The log might state that the initiator selected ID 3, sent a six-byte command, entered data-in, received 512 bytes, and stopped after a missing REQ. It should preserve raw samples alongside the interpretation, because an incorrect decoder must never hide the original evidence.

Analyzer feature Basic diagnostic build More capable build Why it matters
Input channels Data, parity, and core control lines All data, parity, control, and reset lines Complete phase and byte interpretation
Sampling 10–20 MHz 40 MHz or higher Captures handshake timing and short glitches
Memory Small circular RAM Deep RAM or external SDRAM Preserves pre-trigger context and long transfers
Triggering RST, SEL, or REQ Phase, target ID, command byte, timeout Isolates intermittent failures
Output UART or removable storage USB streaming and decoder software Easier capture review and archiving
Electrical interface Protected high-impedance buffers Buffered interposer with selectable protection Reduces loading and protects vintage hardware
Analysis Raw waveform and byte display Phase decoder, parity checks, timing statistics Converts signals into repair evidence

Build triggers around real fault conditions

A trigger on reset is useful when investigating boot failures. It captures the bus immediately after the X68000 initializes its SCSI hardware and shows which device responds first. A trigger on SEL helps isolate a particular target, especially if several devices share the chain. A trigger on a missing handshake requires a timeout counter: if REQ or ACK does not arrive within a selected interval, the analyzer freezes the preceding and following context.

Command-aware triggering makes the instrument much more useful. The firmware can watch for a command phase and compare the first byte against common SCSI commands such as INQUIRY, TEST UNIT READY, REQUEST SENSE, READ, or WRITE. It can then capture only transactions involving a selected target or command. This prevents memory from filling with routine bus activity before the interesting failure occurs.

A manual trigger input is still valuable. Connect it to a push button or a safe external signal so the operator can mark a physical event, such as inserting media or moving a cable. A second input can accept a service switch or diagnostic pulse. Never connect an unverified external signal directly to the X68000 bus; isolate it with the analyzer’s own input circuitry.

Visual status indicators help during long captures. A power LED confirms that the analyzer is operating, while a capture or trigger LED shows that an event has occurred. If the X68000 itself is being modified for clearer diagnostics, the reset and power LED guide provides a useful example of adding front-panel visibility without confusing the machine’s original signals.

Use captures to diagnose hardware

Start with a known-good transaction and save it as a reference. Record the X68000 model, SCSI host adapter, target ID, cable arrangement, termination location, device type, and software used. Vintage systems often contain several revisions of firmware and interface hardware, so a capture without this context is difficult to reproduce.

A bus that remains busy after reset may have a device holding BSY or a termination fault that prevents a line from returning high. SEL asserted with no response suggests a selection problem, while repeated selection and reselection can indicate a target that is alive but unable to complete its command. If the command phase succeeds and the data phase stalls, inspect REQ, ACK, phase changes, and the target’s status response before replacing the drive.

Cable faults often appear as isolated parity errors, extra transitions, or failures that change when the cable is moved. Incorrect termination may produce ringing or marginal voltage levels rather than a completely dead bus. An oscilloscope is useful for analog edge quality, but the digital analyzer shows whether the protocol actually breaks and which byte was affected.

The analyzer also helps separate software from electronics. If the X68000 sends a valid command and the target returns a consistent CHECK CONDITION with a sensible sense response, the bus may be functioning and the problem may be media or driver-related. If the host issues malformed phase sequences or never releases ACK, the investigation should move toward the host adapter, system software, or a timing incompatibility.

For safety, test one change at a time. Capture the original behavior, power down before altering termination or cabling, then repeat the same command. Keep raw files and annotations together. A small archive of traces from working disks, failed disks, and different X68000 models can become more useful than a single sophisticated instrument.

Extend the tool without turning it into a controller

Once passive capture is stable, add features that improve preservation and collaboration. A host utility can annotate traces, calculate inter-byte delays, search for command patterns, and export screenshots or text reports. A compact JSON or binary format can retain metadata such as sample rate, trigger condition, and target ID alongside the signal data.

Network access is convenient for transferring captures from a workshop machine. The X68000 community has also developed expansion hardware such as Nereid-X, which illustrates how modern connectivity and custom hardware can support continued work with the platform. That kind of expansion should remain separate from the analyzer’s electrical observation path; convenience features must not compromise passive behavior.

A future version could include a differential or analog measurement option for diagnosing signal integrity, but it should not replace the digital capture path. It could also provide a programmable trigger language, automatic SCSI phase diagrams, or a library of known-good traces. These features are most valuable after the basic hardware has demonstrated accurate sampling and safe connection.

Document pin assignments, buffer part numbers, protection components, firmware versions, and known limitations. Include photographs of the interposer and clear markings for cable orientation. The goal is a repeatable repair instrument that another X68000 owner can build, verify, and trust.

Practical recommendations for the first prototype

  • Begin with a passive interposer, protected inputs, and complete signal capture before adding protocol transmission or bus-driving functions.
  • Use an FPGA or CPLD when exact REQ and ACK timing matters; reserve a microcontroller for storage, USB, and user-interface tasks.
  • Provide pre-trigger memory, a reset trigger, a selection trigger, and a timeout trigger for missing handshakes.
  • Validate the decoder against a known-good disk and retain raw captures beside every interpreted log.
  • Keep the analyzer electrically independent from the X68000 and verify that it never drives or significantly loads a SCSI signal.

A carefully built SCSI bus analyzer can become a permanent part of an X68000 repair bench. Start with observation, prove the electrical design on a sacrificial or well-understood setup, and publish the schematics, firmware, decoder, and example traces. Sharing those records will make future failures easier to diagnose and preserve practical knowledge that might otherwise disappear with aging hardware.

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.

A two-ink risograph print in muted slate-blue and charcoal on off-white paper, showing a stylized desktop computer monitor beside a circuit board with soft geometric trace lines, conveying a calm retro-computing workshop atmosphere. A two-ink risograph print in deep purple and dark grey on cream stock, depicting a compact expansion card with connector ports and subtle Japanese technical annotations, evoking a hobbyist electronics bench. A two-ink risograph print in teal and charcoal on warm white paper, showing a server rack silhouette with soft network-line motifs and a small weather icon, suggesting a personal server room corner.

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.