Turn measurements into reliable data

Avionics — open definition are the electronics that help operate the vehicle. Computers accept commands, run scheduled tasks, collect telemetry and coordinate equipment. A fast processor alone does not guarantee useful operations: the software, timing, communication interfaces and electrical environment must work together.

A science measurement can be correct at the sensor and still be lost through a full recorder or an incorrect time tag. Separate the original measurement from later processing. Compression, filtering and packet formatting should preserve the information the mission needs, with their effects documented for the people interpreting the result.

An invented instrument produces 2 megabits per second for 60 seconds. That is 120 megabits. Using decimal units and 8 bits per byte, the raw total is 15 megabytes. Headers, error-correction coding and other traffic would increase the transmission requirement; compression could change it in the other direction.

Suppose a recorder holds 120 MB of usable space and this instrument fills it at 0.25 MB/s with no simultaneous downlink. It fills in 480 s, or 8 minutes. These values describe a teaching scenario, not a particular spacecraft. A high radio rate is of little help while the ground station is out of view.

Radiation can upset stored bits or disrupt electronics; not every event has the same effect. Error-detecting and correcting codes can handle specified error patterns. A watchdog — open definition can notice that expected software activity has stopped and trigger a recovery action. Neither mechanism guarantees recovery from every hardware or software failure.

A restart may restore operation while also interrupting work. Preserve enough event information to understand what happened, and verify that recovery leaves actuators in an appropriate state. Repeating the same failed command forever is not meaningful autonomy. Fault handling needs bounded actions and a way for operators to regain control.

Follow the units through the recorder

A calculation that finishes after its deadline may be operationally wrong even if its number is mathematically correct. Test commands, missing inputs, saturated buffers and reset behavior with realistic interfaces. Separate commands that change hardware state from ordinary data handling so their effects are traceable.

Review a downlink plan as a queue: data enters, waits and leaves. Compare the peak occupancy with available storage, not just the daily average. A schedule can have a favorable daily total and still overflow during one long communication gap.