space-rha.comShort courseChapter 3 · RHA requirements

Chapter 3 · Lesson 1 of 6 · about 10 minutes

RHA choices are connected

The mission needs a function to keep working. Which parts are used, how the architecture is built around them, what test evidence exists and what resources are available all shape how that claim is supported, and a change to one of them usually calls for another look at the rest.

RHA choices are connected.

The RHA trade diagram

Adapted from G. R. Allen, "Predicting On-Orbit Performance of Devices," RADECS 2025 Short Course, slide 11. Choose a scenario to see which regions a change touches, in order, and select a region for its trade and limits.

MEAL

Mission, Environment, Application and Lifetime. Before any requirement is written, record the trajectory, the duration, the operating modes and duty cycle, how critical each function is, and how much risk the project is prepared to accept. The requirements that follow are derived from this record, so it is worth getting it down early.

What a change touches

Extend the mission, add susceptible parts, lose a beam slot, add a recovery circuit, or propose board-level TID evidence in place of part data. Each of these changes the exposure, the number of opportunities for an event, or the evidence available, and so each calls for another look at the parts of the argument that depend on it. Tracing those connections tells you what to revisit. It does not choose a design or decide how much risk is acceptable; those remain project decisions.

The LEO shortcut fails

Orbit alone cannot justify a low SEE LET requirement. A complex LEO system with many susceptible parts and a long life can accumulate more opportunities for an event than a small, short mission in a harsher environment, even though each of its parts sees a milder one. The assurance a project needs follows from MEAL and from the consequences of a failure, and the orbit is only one input to that.

Change one choice, then revisit the argument

Another beam condition, a new part, more margin and a recovery circuit each answer a different question. When one of them is proposed, ask what it improves, what it costs, and which uncertainty it leaves in place.

Test depth has several dimensions. More fluence, more parts, more operating corners and more representative particles each address a different uncertainty, and one generally cannot stand in for another. Plan the test around the question that matters to the mission.

Compare evidence routes. Part irradiation, board testing, circuit analysis, system modeling and fault injection can complement one another. Before you treat one as a replacement for another, check how well its conditions match the application and what each one can actually observe.

3D scene

The RHA trade diagram in three dimensions

The nine regions of the trade diagram on a tetrahedral base. Select a region or a scenario, and separate the regions to see which connect.

  1. The apex is the claim the project is making: a function that must keep working, in its real operating windows. Reliability, availability, data quality and safety are different claims, and each asks for different evidence. Everything below the apex exists to support it, and a higher point on the diagram, stronger language or a larger test budget cannot make the claim more probable than its evidence allows.

  2. Pulled apart, the nine regions show the three sides of the source diagram: program constraints, assurance stringency, and device and architecture options. The two lower corners are where a program feels the trade: at the left, more technology options with less reliability assurance; at the right, less confidence under a tighter budget. The regions between them (test coverage, focused evidence, assurance criteria, tailoring to MEAL, test resources and technology limits) are the choices that move a project between those corners.

  3. Keep the orbit and the design and extend the mission: the lifetime term of MEAL changes. The lit path runs from tailoring to the claim, then to the criteria and to the evidence. More time generally means more opportunities for a single-event failure and more accumulated dose, so the evidence that covered the shorter mission may no longer cover the new use, and the phase-by-phase dose and event statistics need another look.

  4. The planned heavy-ion facility becomes unavailable before the design review. The path now runs from test resources through focused evidence and test coverage to the remaining uncertainty. The missing test changes what is known; it changes nothing about the hardware. Another facility or particle, a later test, narrower coverage, heritage, analysis or a design change each answer a different question, and whatever none of them covers stays visible as an open uncertainty.

  5. Adding recovery logic or a redundant channel reaches from design options through technology limits and test coverage back to the claim. Recovery can reduce what an interruption costs the mission, and it adds detection, restart and shared-resource questions of its own. A redundant channel is generally not independent when it shares power, reset, clocks, software or a vulnerable interface, so the mitigation needs its own evidence.

  6. Using a board irradiation in place of part data lights the path from test coverage through technology limits and the criteria to the remaining uncertainty. A board test can show that the assembled function works under the tested conditions. It does not show every part's behavior in every use, so the source penetration, the local dosimetry, the dose-rate and bias dependence and the operating modes exercised have to be compared with the application before it stands in for part evidence.

The original space-rha.com environment and mission renderer (NASA imagery and spacecraft models, credits in the view), with the course scenes drawn by the same renderer. Trails, clocks and counts are illustrative; the scene computes no dose or rate.

Exercise

Take one proposed change to a test plan or a design and write down what it improves (coverage, precision or design tolerance), what it costs, and which uncertainty remains.

Choose the evidence for the uncertainty it can resolve.

Going deeper

The assumption behind the evidence

Every piece of evidence supports a mission claim through a set of assumptions: that the tested samples represent the flight lot, that the test source stands in for the flight environment, that the tested operating modes cover the flight modes, that the events are independent, and that the model matches the hardware closely enough. A quantitative result is only as relevant as those links, so find the weakest one before you rely on the number.

Conditional risk. An assumed event rate is a long-run average under a stated exposure, and it is not the same thing as a measured probability of failure. Before you read anything into the number, write down the exposure case it assumes, the response model, the critical window during which an event matters, and what recovery is credited.

Early involvement. RHA should be part of the design loop before the parts, the board layouts and the recovery architecture are fixed, because changing any of them later is expensive. A change whose consequences fall outside the existing allocation should go back to the project for a decision.

Takeaway

Start with the behavior that matters to the mission, then ask how the evidence supports it.

Check yourself

A part has no radiation data and the schedule allows no test. Which regions of the trade diagram can still be worked?

Those that change the consequence rather than the likelihood: redundancy, recovery, derating, operating restrictions, and acceptance of a documented risk. The likelihood remains unknown and is recorded as such; the design can still make the function tolerant of the event.

What does a tested part with a known rate buy the trade that an untested one does not?

A likelihood that can be compared with the consequence. With it the project can decide between tolerating the events, reducing their rate with a different part or more shielding, and spending on mitigation; without it every option is argued on consequence alone.

A short mission is offered as the reason to skip destructive single-event testing of a new part. Which region of the trade diagram answers that, and what does it say?

The tailoring region. Mission, Environment, Application and Lifetime determine which effects, operating modes and consequences need attention, and the limit on that region is that a benign orbit alone does not justify a low LET screen and a short mission does not remove high-consequence single-event mechanisms. A shorter mission generally reduces the number of opportunities for an event; it does not change what the event does to the function, so the tailoring has to be made to MEAL and consequence together, with an explicit reason for each choice.

A test plan proposes more fluence at the one LET already tested in order to raise confidence. Which uncertainty does that reduce, and which does it leave untouched?

More fluence reduces the counting uncertainty for the sampled condition, and only that. Coverage and precision are separate: a million additional ions at one condition do not establish another mode, lot, ion range or failure mechanism, and more samples, more operating corners or more representative particles each address a different uncertainty that one generally cannot stand in for another. The next test increment should be spent on the uncertainty that matters to the claim, which is a question about the mission rather than about the beam.

The planned heavy-ion slot is lost before the design review, and a proton test at an easier facility is offered in its place. What does the trade diagram say about the substitution?

The missing test changes what is known; it leaves the hardware's susceptibility unchanged. An easier facility is a useful substitute only when its particles, range, dose rate, geometry and instrumentation address the same question, or when the remaining gap is understood, and a proton result does not generally establish a heavy-ion destructive-SEE bound. Another facility or particle, a later test, narrower coverage, heritage, analysis and a design change each answer a different question, and whatever none of them covers stays visible as an open uncertainty rather than disappearing with the beam slot.

Further reading