Chapter 12 · Lesson 2 of 4 · about 12 minutes
Reconstructing exposure and keeping hypotheses alive
A large solar event makes radiation plausible, but it does not show what reached the part. The orbit, the shielding, the orientation and the operating state all enter. A local monitor is the best evidence when there is one; a remote monitor or a model describes the part only after the transfer to its location has been argued, and the uncertainty in that transfer stays with the result.
What you need to know
A large solar event in the news makes radiation plausible. It does not show what reached the part. The exposure at the device is reconstructed from the orbit and the time (geomagnetic position, anomaly or horn transit), from the solar and galactic conditions over the interval, from the shielding and orientation, and from any local monitor. Remote monitors such as the geostationary proton flux and the ground-based neutron monitors give context, but you have to argue the transfer through the geomagnetic cutoff and the shielding before they say anything about the part. The absence of a reported event excludes nothing. Trapped protons and galactic ions are present throughout the mission, and in a sun-synchronous orbit the anomaly passes, 12 % of the time with an instantaneous peak 50 times the orbit average, carry most of the proton exposure (chapter 1 lesson 4). Reconstruction is an inverse problem with few sensors, and several local spectra are generally consistent with the same observations. The result is a set of defensible scenarios, each with its uncertainty, rather than one curve.
A bit flip, a reset or a current step is a signature, and several mechanisms can produce the same one: a single-event effect, a software fault, a timing margin, an electrical transient, spacecraft charging and discharge, a thermal excursion, a command. The investigation starts with several hypotheses and asks what observation would favor or weaken each. For a radiation hypothesis that means a credible particle pathway to the affected structure (the right particle at the right place with enough energy), a rate consistent with the prediction and the exposure, and a signature consistent with the ground test. Each alternative has its own predictions to check. Negative evidence counts only when the missing signature should have been detectable. The hypothesis matrix records the status of each hypothesis (excluded, consistent, demonstrated) separately from the mitigation decision. A mitigation can be adopted for the consequence while the cause stays open, and a reset that cleared the symptom says nothing about where the symptom came from.
Worked example 12.2. The U1 current step: exposure and competing explanations
For the event of lesson 12.1 at 48° S, 20° W, the spacecraft was inside the AP-8 anomaly footprint at 600 km during a quiet period. There was no solar particle event in the geostationary record for the preceding 72 hours, and the galactic flux was near the solar-minimum level.
| Hypothesis | Predicts | Observed | Status |
|---|---|---|---|
| Heavy-ion-induced high-current event in U1 | rate of order 1.3 × 10⁻⁶ to 4.6 × 10⁻⁶ per day, anywhere on the orbit; a step within microseconds; recovery by the limiter | step within one sample; limiter trip; recovery per the ladder; the first such event in 410 days | consistent in signature but not in rate. The worst case predicts 1.9 × 10⁻³ events in 410 days, and one event bounds the rate above 6.2 × 10⁻⁵ per day at 95 % (lesson 12.3). Either the curve, the geometry or the environment is under-predicted, or the cause is something else |
| Proton-induced event in U1 | onset near LET 21 is above what proton recoils deposit; no proton events in the U1 ground test | the position in the anomaly favors protons | disfavored by the threshold, though the threshold is now itself in question. A proton run of U1 at the flight bias would settle it |
| Load transient from the recorder (software) | a current rise in step with a recorder mode change, over milliseconds, no trip | no mode command in the preceding minute; a sub-sample step | weakened |
| Charging-induced discharge on the board | a transient on several rails at once; correlation with eclipse exit and high-latitude electron flux | single rail; mid-latitude; no eclipse transition | weakened |
| Limiter false trip (its own transient) | a trip without a preceding current step | the step precedes the trip by 12 µs in the trace buffer | excluded by the trace |
The anomaly position supports both radiation hypotheses and the threshold argument separates them, but the rate fits neither. One event where the worst case expected 1.9 × 10⁻³ is a finding against the prediction; under the worst-case model the probability of at least one event in that time was about 0.2 %. The investigation therefore records "signature consistent with a high-current single event; rate inconsistent with the heavy-ion prediction; cause not demonstrated". The actions are a proton run and an angular heavy-ion run of U1, and a check of whether solar heavy ions contributed during the interval. The mitigation question is separate from the attribution. The limiter and the recovery path worked as demonstrated on the ground, and the post-trip inspection for latent damage from lesson 11.2 is now a flight question, answered by the U1 leakage trend.
Exercise
For each physical hypothesis, write down the time and position of the event, the monitor data used, the assumptions made in transferring that data to the part, and the local transport through the shielding.
Keep a hypothesis matrix like the one in worked example 12.2, with the observation that would discriminate each hypothesis, the observations that count against it, and its current status (excluded, consistent or demonstrated).
Going deeper
Exposure reconstruction is an inverse problem
With few sensors, several local spectra can fit the same observations. Geomagnetic access, the anisotropy of the event, transport through the shielding and the detector response each add uncertainty. The useful comparison is across a set of physically defensible scenarios. Fitting one environment curve to the anomaly hides how little the local field is constrained, so the alternatives are kept when the constraint is weak.
Two remote records that agree with each other can still share the same model and the same assumptions about the transfer to the spacecraft.
Preferring a hypothesis takes more than coincidence
The preferred hypothesis is the one that explains the evidence better than the alternatives, under stated assumptions about the prior and about the measurement. A rare anomaly seen through incomplete telemetry may not support a calibrated probability for any hypothesis at all. Negative evidence helps only when the missing signature should have been detectable. The unresolved branches are kept, and excluded, consistent and demonstrated are kept as three different statuses.
A reset that worked, or one symptom reproduced in the laboratory, does not by itself identify the mechanism that started the flight event.
The common mistake
The environment at the time of the event cannot by itself tell a radiation upset from a software, electrical or charging event.
With incomplete telemetry the root cause can stay unresolved even after a workable mitigation has been found.
Takeaway
- A radiation explanation needs a credible pathway, a particle or a dose, to the affected structure and the affected function.
- Keeping the alternatives open stops a plausible radiation story from getting ahead of the evidence.
Check yourself
A reset coincides with a reported solar event. What has to be shown before the event is the explanation?
A pathway: that the particles of the event reached the part at that orbit position and shielding (through the cutoff, for a low-inclination orbit), that the affected part has a credible response to them, and that the rate is consistent. The coincidence by itself is only context.
Why are the alternatives kept after a radiation hypothesis becomes consistent?
Because consistent is not demonstrated. The same signature has other causes whose predictions have not been excluded, and the next event will discriminate between them only if the record was designed to.
A ground test reproduced the current step with a laser at the suspected block. What has it demonstrated?
That the block can produce that signature under injected charge. It has not shown that the flight event originated there, or which particle caused it. It strengthens the pathway argument and is recorded as that.
The U1 event occurred inside the anomaly footprint, and a colleague takes the position as strong evidence for a radiation cause. How much does the position weigh?
It supports a radiation hypothesis without establishing it. In a sun-synchronous orbit the anomaly passes are 12 % of the time with an instantaneous peak 50 times the orbit average and carry most of the proton exposure (chapter 1 lesson 4), so a proton-driven event would be expected there. But the heavy-ion hypothesis predicts events anywhere on the orbit, and the proton hypothesis is disfavored by the onset near LET 21, above what proton recoils deposit. The position is context: it separates neither the two radiation hypotheses from each other nor radiation from the alternatives, which the sub-sample step, the single rail and the absence of a mode command do.
The limiter tripped and the recorder recovered by the ladder, so the project proposes closing the investigation as "radiation event, mitigated". What is wrong with that record?
The mitigation decision and the hypothesis status are kept separate. The limiter and the recovery path worked as demonstrated on the ground, and that is recorded; the cause is not demonstrated, because the signature is consistent with a high-current single event while the rate is not: the worst case predicted 1.9 × 10⁻³ events in 410 days, and one event bounds the rate above 6.2 × 10⁻⁵ per day at 95 %. The record therefore reads "signature consistent; rate inconsistent with the heavy-ion prediction; cause not demonstrated", with the actions attached (a proton run, an angular heavy-ion run, a check of solar heavy ions). A reset that cleared the symptom says nothing about where the symptom came from.
Why does the hypothesis matrix give each alternative its own prediction, rather than listing only what would confirm the radiation hypothesis?
Because several mechanisms produce the same signature, and a hypothesis is weakened or excluded only by an observation it predicted and that was not seen. The load-transient hypothesis predicted a rise over milliseconds in step with a mode change and no trip, and the observation was a sub-sample step with no mode command in the preceding minute; the charging hypothesis predicted a transient on several rails at eclipse exit, and the observation was a single rail at mid-latitude with no eclipse transition; the false-trip hypothesis predicted a trip without a preceding step, and the trace buffer alone excluded it. Negative evidence counts only when the missing signature should have been detectable, which is why the telemetry design of lesson 12.1 decides what the matrix can resolve.
Further reading
- D. Falguère et al., "In-flight observations of the radiation environment and its effects on devices in the SAC-C polar orbit," IEEE Trans. Nucl. Sci., vol. 49, no. 6, pp. 2782-2787, 2002.
- R. Harboe-Sørensen et al., "Observation and analysis of single event effects on-board the SOHO satellite," IEEE Trans. Nucl. Sci., vol. 49, no. 3, pp. 1345-1350, 2002.
- NASA, "Avionics Radiation Hardness Assurance (RHA) Guidelines," NASA/TM-20210018053, 2021. NTRS 20210018053.