space-rha.comShort courseChapter 3 · RHA requirements

Chapter 3 · Lesson 6 of 6 · about 10 minutes

A requirement someone can verify

A requirement that can be verified names the hardware or function that must perform, the exposure and conditions it must perform under, the measurable limit, and how the evidence will show that the limit is met.

Write a requirement someone can verify.

Rewrite the weak version

A clause that says only "radiation tolerant" cannot be verified, because it gives nothing to measure. A verifiable clause names the hardware or function, the exposure or event, the operating envelope, the measurable limit and the route by which it will be verified.

Draft matrix

A draft verification matrix lists each clause alongside the evidence it has and the evidence it still needs. A clause whose arithmetic falls within the budget is still unverified until the evidence behind the arithmetic is shown to apply and has been reviewed.

Build the requirements and verification package

The package links each requirement to its parent, its owner, the analysis it is based on, the conditions, the verification method and the evidence. Analysis, test, inspection and demonstration are all legitimate methods; choose for each requirement the one that answers its question, and say why.

Evidence strategy. Each kind of evidence answers a different question. Irradiating the part measures how susceptible it is. Circuit analysis connects that degradation or event to what the circuit does. Fault injection shows whether the detection and recovery work when an event occurs. A fault-injection test does not measure the cross section, so it cannot replace the irradiation.

Package contents. The package should carry the environment and transport configuration the numbers came from, the inputs, the clauses and the list of unresolved items, together with the verification matrix in a form that reviewers can edit. A complete package is still a draft. It becomes a verified mission result only when the evidence in it has been reviewed and accepted.

3D scene

Closing the argument

The dependency network again, read from verification back to the mission claim.

  1. A clause that says only "radiation tolerant" cannot be verified, because it gives nothing to measure. A verifiable clause names the hardware or function that must perform, the exposure or event, the operating envelope and the measurable limit. The lit node is the allocation: the function the clause belongs to.

  2. The clause also says how the evidence will show that the limit is met. Irradiating the part measures how susceptible it is; circuit analysis connects that to what the circuit does; fault injection shows whether detection and recovery work when an event occurs, and it does not measure the cross section, so it cannot replace the irradiation. Analysis, test, inspection and demonstration are all legitimate methods; choose the one that answers the question, and say why.

  3. The package links each requirement to its parent, its owner, the analysis it rests on, the conditions, the verification method and the evidence, so that a reviewer can trace it from the mission need down to the evidence and back. A matrix with every field filled in is still a draft; it becomes a verified mission result only when the evidence in it has been reviewed and accepted.

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

Prepare a draft requirements package for one function that a reviewer could trace from the mission need down to the evidence and back.

A matrix with every field filled in still needs evidence that applies to the application, and it still needs a technical review.

Going deeper

Closing the argument and controlling change

Deviations and residual risk are decided by the project authority against the approved mission criteria. When the requirements are baselined, record the assumptions, the conditions under which the evidence applies and the hardware configuration along with them, since a later change will be checked against those.

Reopen triggers. A mission extension, a change of trajectory, a change of part, process or lot, a shielding update, a different workload, a change in voltage or temperature, or new recovery behavior can each require the assessment to be reopened.

Next chapter. The next chapter, on mechanisms and technologies, connects these criteria to the device physics and to the functions that are susceptible. The requirements should guide that investigation, and they can be revised when the evidence changes.

Takeaway

Make the acceptance criterion explicit before the test.

Check yourself

A part passed its qualification test. What else is needed before the requirement is closed for the flight configuration?

That the tested configuration is the flight one: same die revision and lot where the effect depends on it, the same bias and modes, and the mitigation as implemented. Closure is a statement about the flight hardware, and a test on something else supports it only through an applicability argument.

A firmware update changes the scrub interval after verification. What reopens?

The requirements whose verification assumed the old interval: the uncorrectable-error rate and the availability that depends on it. The other closures stand. Change control is what finds the ones that reopen.

A fault-injection campaign shows that the watchdog recovers from every injected upset, and it is offered as closing the single-event requirement on the processor. What does it close, and what does it not?

It shows whether the detection and recovery work when an event occurs, which is the question fault injection answers. It does not measure the cross section, so it cannot replace the irradiation that measures how susceptible the part is, and it does not by itself connect the event to what the circuit does, which is the job of circuit analysis. Each kind of evidence answers a different question, and the requirement should say which method answers its question and why; the rate part of the requirement stays open until the irradiation evidence is in.

A verification matrix for a function has every field filled in and the arithmetic falls inside the budget. Why does the lesson still call it a draft?

A clause whose arithmetic falls within the budget is still unverified until the evidence behind the arithmetic is shown to apply to the application and has been reviewed. The package becomes a verified mission result only when the evidence in it has been reviewed and accepted, and a matrix with every field filled in still needs evidence that applies to the application and a technical review. Deviations and residual risk are then decided by the project authority against the approved mission criteria, which is a decision the matrix records rather than makes.

A TID test is run first and the pass/fail limit is chosen afterward from the data. Why does the lesson ask for the acceptance criterion before the test?

A verifiable clause names the measurable limit and the route by which it will be verified, and the test is designed around that limit: the parameters monitored, the modes exercised and the conditions set. A data-sheet violation, a loss of circuit margin and a loss of mission function are three different endpoints (lesson 3.3), and a limit chosen after the data may not be the one the application needs, which leaves the clause unverified however clean the result looks. Making the criterion explicit before the test is what lets the evidence be judged against the requirement rather than the requirement against the evidence.

Further reading