EEC510 electron beam controller: blown tantalum capacitor, failed switching regulator and unstable internal USB link

A component on the controller's I/O board had visibly burst. The unit also showed communication errors with its I/O board; after a restart, communication came back only briefly.

• The burst part was a 68 µF 20 V tantalum capacitor in the board's power supply.
• The LM2575S-ADJ step-down switching regulator next to it had also failed.
• With the power supply repaired, the unit starts: BIOS, Linux and the controller software (EEC500 v3.007) load, and the network port links at 1 Gbps.
• Communication fault reproduced: at some starts the FT4232HL USB-to-serial bridge on the I/O board was not enumerated, so its four USB serial ports did not appear and the controller software could not connect to the I/O board.
• The internal link was traced on the unit: PC board header CN18 → white internal USB cable → I/O board J10 → FT4232HL (U1) → H8S controller (IC19).
• Voltages around U1, measured against J9 GND, read about 1.80 V and 3.6 V with no steady loss of supply at the points checked.
• With the link up, the I/O board answered correctly (firmware version and status replies), so the bridge-to-controller path itself works.
• Some test cables failed USB enumeration (descriptor errors) while others worked; the original CN18 path still failed some cold starts after the BIOS change.

• Power supply: not confirmed from the record. A shorted tantalum capacitor can overload the regulator next to it, but which of the two parts failed first was not established.
• Communication: not confirmed. The fault is an intermittent USB enumeration problem on the internal CN18 → J10 path at start-up; which part (header, cable, or the timing on either board) is to blame was not proven.

• Replaced the 68 µF 20 V tantalum capacitor.
• Replaced the LM2575S-ADJ switching regulator.
• Changed the BIOS USB setting: xHCI enabled, USB 2.0 (EHCI) support disabled.
• Disconnected the original white internal cable from J10 only (left in place at CN18) and connected J10 to the PC board's second USB-A port with a new USB A-to-B cable.
• No board, chip or EEPROM was replaced for the communication fault, and no firmware, system or control program was rewritten; the key program, kernel and start-up files match the original image.
• Also delivered a separate read-only monitoring tool that shows the link, USB state, I/O values and logs over the network.

• Power-on test passed: the unit started up and its display and indicator worked.
• Three cold starts (power off 30 s, 30 s and 120 s): each time the four USB serial ports appeared on their own, the controller software connected, the READY lamp stayed lit, and no USB errors or disconnects were logged.
• 35 status samples across the three starts all showed the I/O board connected, with fresh data each time.
• Not tested: the full electron-beam control functions on the customer's complete machine (remote control, crucible drive, water, high voltage, filament and all physical I/O).

Both failed power-supply parts replaced, and the controller now starts communication with the I/O board on every cold start tested, using a rerouted USB link and no changes to the original software. The root cause of the intermittent fault on the original USB path is not yet confirmed.