Rad-hardened compute for thin constellations: lessons from 6 months in orbit

Rad-hardened compute lessons from 6 months in orbit

AI & Edge · 20 min read  ·  07 APR 2026

Fig. 01 — On-orbit compute stack telemetry readout. LaythTech constellation arc, six-month operations review.

LaythTech Engineering
On-orbit Systems · LaythTech

After six months of on-orbit operations with our first constellation arc, the engineering team documents what we got right — and what we'd change — about the compute stack.

Six months into the operation of our first three-satellite constellation arc, the LaythTech engineering team has accumulated enough on-orbit data to draw meaningful conclusions about the compute architecture choices we made during the design phase.

What We Got Right

The hybrid compute architecture — rad-hard primary processor paired with a watchdog-monitored COTS accelerator in a shielded enclosure — has performed better than our pre-launch FMEA predicted. Single-event upsets (SEUs) in the COTS accelerator have averaged 1.2 events per spacecraft per month, all successfully handled by the watchdog reset mechanism without human intervention.

What We Would Change

The memory architecture — separating mission-critical flight software storage on radiation-hardened NOR flash from payload data storage on COTS NAND with triple-redundancy — has been particularly effective. However, the thermal management subsystem was undersized for the equatorial crossing thermal cycles we are observing.

Filed under AI & Edge Saudi Space Deep Tech
Partner with us

Build Saudi Arabia's
future in space.

Layth Tech collaborates with governments, tech firms, universities, and industry leaders to enhance space systems and create national opportunities. If you need satellite infrastructure, space connectivity, technology transfer, or strategic partnerships, Layth Tech is here to help.

Partner with LaythTech →
LaythTech satellite