1. When one particle changes everything

Michael Crichton’s The Andromeda Strain returns as the recalled work because its crisis begins with a microscopic event that propagates through a system not designed to contain it. A single-event upset is the orbital equivalent: one ionizing particle, one flipped bit, and a computation that no longer matches reality. The desktop must be tested to see whether it can catch and correct such errors before they become failures.

This entry explains why the single-event upset recovery test matters.

2. Why radiation-hardened design is not enough

The radiation reading arc established that the desktop should use a hybrid strategy: rad-hard or rad-tolerant parts where failure is catastrophic, and COTS compute with system-level mitigation where performance and cost dominate. That mitigation is a design claim. The SEU recovery test is the experiment that validates it.

Without SEU testing, the desktop might discover these problems only after launch:

  • A single bit flip in a guidance parameter produces a slowly diverging attitude estimate.
  • A corrupted command packet is executed because the checksum was not verified at the right layer.
  • An FPGA configuration bit flips and changes the behavior of a motor controller.
  • A multi-bit upset exceeds the correction capability of a simple ECC code.
  • A corrected error is logged but the trend is never analyzed, so cumulative degradation goes unnoticed.

3. The failures that only appear when bits are flipped

SEU injection exposes behaviors that do not appear in nominal testing:

  • ECC corrects the bit but the correction latency causes a real-time task to miss its deadline.
  • A scrubber detects and repairs an FPGA configuration upset, but the repair itself glitches an output.
  • A redundant voter sees two identical wrong answers and one right answer, then follows the majority.
  • A “corrected” error is silently ignored until a second error in the same word becomes uncorrectable.
  • A safe-mode trigger is itself corrupted, so the platform cannot retreat to safety.

4. What this test must cover

The SEU recovery test must exercise:

  • Single-bit errors in data memory, code memory, registers, and caches.
  • Multi-bit errors in the same word or adjacent words.
  • Errors in configuration memory for FPGAs and SoCs.
  • Errors in critical state versus non-critical state.
  • Errors during high-load and low-load operational phases.
  • Errors in the recovery mechanisms themselves.

The goal is to prove that the platform’s layered defenses detect, correct, contain, and report realistic SEUs.

5. What this changes

  • Single-event upset recovery testing is identified as the next qualification step for the desktop.
  • The test must verify correction and containment, not just detection.
  • The next entry will define the test matrix.