CPU364 "OK" LED Goes Dark at Power-Up: How to Capture the Serial Kernel Diagnostic Log
Engineers maintaining legacy GE Fanuc Series 90-30 systems often face a frustrating symptom. The "OK" lamp flashes once, then dies. The "FAIL" lamp stays lit. In industrial automation, this pattern usually points to a firmware fault or a watchdog timeout. The following guide explains how to retrieve the kernel diagnostic log through the dedicated RS-232 port. It also covers baud rate settings, binary capture rules, and data interpretation.
Why the LED Sequence Matters for PLC Power-Up Failure
At power-up, the CPU364 runs a fixed diagnostic routine. Normally, the "OK" LED blinks fast while other lamps stay off. A single short flash followed by darkness means the routine started but never finished. Meanwhile, the solid "FAIL" lamp confirms a fatal fault state. This condition differs from a key-switch STOP position or a configuration mismatch. Therefore, treat it as a serious hardware or firmware event. The kernel crash dump captures the operating system state at the exact moment of failure.
Connect the Serial Cable Before You Cycle Power
Timing is critical for serial diagnostics. You must connect the cable and start monitoring before power is applied. The diagnostic access point sits on the 9-pin D-sub connector on the PLC power supply. However, do not confuse it with the RJ-11 Station Manager port on the CPU faceplate. Use a null-modem serial cable to link this port to a PC or laptop. This step is essential for any factory automation site running aging control systems.
Terminal Settings for Reliable Binary Capture
Communication parameters must match exactly: 19200 baud, 8 data bits, no parity, and 1 stop bit. Disable flow control entirely. Most importantly, set your terminal software to raw binary mode, not text mode. Text mode translates control characters and inserts line breaks. As a result, it corrupts the binary integrity of the kernel dump. Save the stream with a .bin or .dmp extension. Typical dump sizes range from 64 KB to 256 KB, depending on the firmware revision.
Trigger the Kernel Dump and Watch the Terminal
With logging active, apply power to the CPU364 rack. The dump usually transmits automatically once the fault is detected. However, some firmware revisions need a manual escape sequence or diagnostic command. Watch the terminal window right after power application. If no data appears within 10 seconds despite a confirmed FAIL condition, try the manual trigger from the firmware release notes. The terminal will print the literal string "Kernel Crash Dump" alongside the binary payload. Do not disconnect the cable until transmission stops completely.

Verify Capture File Integrity Before Analysis
File integrity checks are mandatory after capture. Confirm the saved file size falls within the expected range for your firmware revision. A file under 64 KB likely means an incomplete capture or a terminal configuration error. For example, ASCII mode may truncate the dump or inject corrupted control sequences. Moreover, check that the PC did not sleep or suspend the USB-to-serial adapter. Field experience shows that cheap USB-serial adapters with unstable drivers cause many capture failures. A hardware RS-232 port or an industrial-grade adapter is strongly recommended.
Secondary Diagnostic Paths When Serial Capture Fails
If the serial dump cannot be obtained, other options exist. The Ethernet Station Manager interface provides exception logs and PLC fault tables. However, it cannot capture kernel-level panics. After connecting via Ethernet, read both the CPU fault table and the I/O fault table immediately. Fault codes such as MOD-001 indicate a missing module. Checksum failures point toward memory corruption. In addition, verify the internal battery voltage exceeds 3.0 V DC. A depleted battery is a known cause of configuration loss and boot failure on aging CPU364 hardware.
Age-Related Failure Modes in Industrial Control Systems
Field reliability data reveals clear patterns after 15 to 20 years of service. The most common failure modes include SRAM backup loss, solder-joint fatigue on the BGA-mounted Ethernet controller, and on-board flash memory failure. If the kernel dump shows a flash checksum error, module replacement is the only viable long-term remedy. From an engineering standpoint, every kernel panic deserves attention. Treat it as a terminal warning sign, not an isolated event. Proactive replacement protects critical processes in any DCS or PLC environment.
Documentation and Long-Term Reliability Planning
Record all diagnostic findings in the site maintenance log. Note the date, time, firmware revision, dump file size, and any fault codes from the binary analysis. This baseline data supports trending and predictive maintenance. The CPU364 is discontinued, and spare inventory is limited. Therefore, plants still relying on this platform should plan a migration path to the CPU372 or CPU374. For critical applications where downtime costs exceed replacement costs, keep a pre-configured spare CPU364 module on the shelf. Furthermore, remove any CPU364 exhibiting repeat kernel panics after a successful cold restart from production duty.
Practical Application Scenario
Consider a water treatment plant running a Series 90-30 rack with a CPU364. One morning, the "OK" lamp flashes once and the "FAIL" lamp stays solid. The maintenance team connects a laptop to the power supply serial port before the next power cycle. They capture a 128 KB binary dump at 19200-8-N-1 in raw mode. Analysis reveals a flash checksum error. The team replaces the module with a pre-configured spare and schedules a migration to the CPU374. Downtime stays under two hours, and the plant avoids an unplanned outage.
Key Technical Takeaways for Field Engineers
The serial kernel dump remains the only method for capturing pre-boot firmware failures on the CPU364. Always use raw binary capture mode at 19200-8-N-1. Connect before power-up and verify file size against the 64–256 KB expected range. If serial capture is impossible, use the Ethernet Station Manager exception log and PLC fault tables as secondary evidence. Given the documented aging-related failure modes, treat every kernel panic as a terminal warning sign.

Frequently Asked Questions
Q1: Why does the CPU364 "OK" LED flash once and then go dark?
A single brief flash means the power-up diagnostic routine started but failed to complete. The solid "FAIL" lamp confirms the CPU entered a fatal fault state. Common causes include firmware corruption, flash memory failure, or a watchdog timeout.
Q2: Which serial port should I use for the kernel diagnostic log?
Use the 9-pin D-sub connector on the PLC power supply. Do not use the RJ-11 Station Manager port on the CPU faceplate. A null-modem cable is required between this port and your PC.
Q3: What terminal settings are required for a valid dump?
Set 19200 baud, 8 data bits, no parity, 1 stop bit, and disable flow control. Most importantly, use raw binary capture mode. Text mode corrupts the binary payload and may truncate the file.
Q4: How large should the captured dump file be?
Typical dumps range from 64 KB to 256 KB, depending on the firmware revision. A file smaller than 64 KB usually indicates an incomplete capture or a terminal configuration error.
Q5: What if serial capture fails completely?
Use the Ethernet Station Manager interface to read exception logs and PLC fault tables. Also check the internal battery voltage, which should exceed 3.0 V DC. If the kernel dump shows a flash checksum error, replace the module.














