Chapter 1 · Lesson 4 of 6 · about 12 minutes
Lifetime totals and critical windows
Some effects accumulate over the mission, and for those only the total matters. Others happen in an instant, and for those what matters is the rate at that instant and whether the function can tolerate an event then. The two questions take different quantities from the environment: a fluence integrated over the lifetime for the first, a peak flux over the critical window for the second. Mixing them up produces requirements that are too strict in one place and too loose in another.
What you need to know
Cumulative effects. Total ionizing dose and displacement damage add up. The requirement is the integral over the mission of the local spectrum, folded with the response, and the time structure within the mission generally does not matter except through annealing and the order of exposures. A mission extension adds in proportion, and a solar event adds a step.
Event effects. A single-event upset, a latchup, a transient on a regulator or a burnout is caused by one particle, and its probability is a rate times a time. For these the question is how often the event can happen during the window in which the function must not fail, and what the worst rate is during that window. A memory that is scrubbed every hour needs the upset rate per hour at the worst moment. A one-time pyro firing needs the probability of an event in a few seconds. A payload that must deliver data every day needs the rate during an anomaly pass and during a solar event, not the mission average.
Three time structures matter in Earth orbit.
- The anomaly pass. A low orbit meets trapped protons only on some passes and only for minutes. The peak flux on those passes is tens of times the orbit average.
- The solar event. Days of elevated proton and heavy-ion flux, at irregular intervals, with the largest events delivering in a few days what the trapped environment delivers in a year at some orbits.
- The storm. Outer-zone electron fluxes can rise by orders of magnitude within a day and decay over a week, which matters for internal charging and for dose at GEO and in the outer zone.
Worked example: five years of dose, and the week that delivers it
The question. A five-year sun-synchronous mission at 800 km. Which population sets the five-year dose behind 100 mil of aluminum, how much does one large solar event add, and what does the time structure of the proton exposure look like for a part whose requirement is written as a rate?
Inputs. The one-year trapped doses and the one-event August 1972 dose behind 100 mil of aluminum from the Space-RHA hardware panel (AE-8 MAX, AP-8 MAX, SHIELDOSE-2, solid sphere, event through the quiet-time cutoff of the orbit); AP-8 MAX protons above 10 MeV along the orbit from the same panel; the SPENVIS runs of lessons 1 to 3.
| Dose behind 100 mil Al, rad(Si) | LEO 420 km, 51.6° | Polar 800 km, 98.6° | GEO |
|---|---|---|---|
| Trapped, one year | 381 | 2,580 | 88,300 |
| Trapped, five years | 1,900 | 12,900 | 442,000 |
| One August 1972 event | 0.15 | 505 | 3,260 |
| Event as a share of five years | 0.008 % | 3.9 % | 0.7 % |
| Trapped protons above 10 MeV along the orbit (AP-8 MAX) | LEO 420 km, 51.6° | Polar 800 km, 98.6° |
|---|---|---|
| Orbit-averaged flux, cm⁻² s⁻¹ | 9.23 | 115 |
| Largest flux met on the track, cm⁻² s⁻¹ | 743 | 5,760 |
| Peak over average | 80 | 50 |
| Fraction of the time in nonzero flux | 8.5% | 12% |
Steps.
- Multiply the yearly trapped dose by the lifetime: 2,580 rad(Si) per year gives 12,900 rad(Si) in five years on the polar orbit, nearly all from electrons at this depth (lesson 5). This is a model output for a solid sphere; a project applies its own geometry and margin policy on top (chapter 3).
- Add the events. One August 1972 event adds 505 rad(Si) on the polar orbit, 4 % of the five-year trapped total. A 95 % active-year fluence from ESP would be the proper requirement basis (lesson 3). On the ISS orbit the same event adds a negligible 0.15 rad because the cutoff excludes it. At GEO it adds 3,260 rad, under one percent of a five-year electron dose of 442 krad. For dose, solar events are a correction at these orbits. For displacement damage in a solar cell or an optocoupler they can be the main term, because the trapped electrons that dominate the dose do far less displacement damage per particle than the protons.
- Look at the time structure of the protons. On the polar orbit the trapped-proton flux above 10 MeV is nonzero only 12% of the time and peaks at 50 times the orbit average; on the ISS orbit, 8.5% of the time and 80 times the average. A proton-sensitive function that must work continuously sees the peak, not the mean.
- Compare one event with one year. At 800 km the trapped protons deliver 5.3 × 10⁹ per cm² above 10 MeV in a year. The 95 % worst event, after the orbit's transmission of 0.21, delivers about 5.3 × 10⁹ per cm² in a few days, essentially a year's worth of protons at once (ratio 0.99). Proton-induced upset rates during such an event are correspondingly higher than the mission average, and that is the rate an availability requirement has to be written against. At the ISS orbit the corresponding share is 1 %, so there the anomaly passes set the peak instead.
Result. The five-year polar mission needs three kinds of requirement. The cumulative dose requirement is of order 13 krad(Si) behind 100 mil from the trapped environment, plus the active-year solar term at the project's confidence level. The displacement-damage requirement is driven by the proton fluences of both the anomaly and the events. The event-rate requirements are written for the anomaly pass and for the design solar event, not for the mission average.
Going deeper
Flux, fluence and the units that go with them
Flux is particles per unit area per unit time, fluence its time integral. Both come as integral (above an energy or LET) or differential (per unit energy or LET) quantities, and both come as directional (per steradian) or omnidirectional. An orbit-averaged omnidirectional integral flux above 10 MeV in cm⁻² s⁻¹ is a different quantity from a differential directional flux in cm⁻² s⁻¹ sr⁻¹ MeV⁻¹. The two differ by factors of 4π and by an energy bin, so label every quantity with all four attributes. Model outputs that quote "per m² per sr per s" for an isotropic field convert to omnidirectional flux through a sphere by a factor 4π (lesson 1).
Counting statistics in a short window
For a Poisson process of rate r, the probability of at least one event in a window T is 1 − exp(−rT), which is close to rT when rT is small. A latchup rate of 10⁻⁴ per device-day gives a 3.6 % probability per device-year and about 0.1 % over a ten-day critical window, provided the rate is the rate for that window. During a solar event the rate can be larger by orders of magnitude for the days of the event. The same expression gives the counts expected during a beam test and the fluence needed to demonstrate a rate bound with no events (chapter 3).
Scrubbing and the critical window
A memory protected by single-error correction and periodic scrubbing fails when a second upset lands in a word that already holds one, before the scrub clears it. The failure rate scales with the square of the upset rate times the scrub interval, so a tenfold increase in the upset rate during an event raises the uncorrectable rate a hundredfold. A scrub interval that was comfortable at the mission average may be inadequate during the event. This is one way the time structure of the environment reaches the system design.
The common mistake
The common mistake is to verify an event-rate requirement against a mission average. "Fewer than one upset per year on average" is met by a part that upsets ten times during the worst week and never otherwise, and that part may be unacceptable in a function that cannot tolerate a burst. Write the window into the requirement: the rate during an anomaly pass, during the design solar event, during the critical operation, with the environment case named for each.
Takeaway
Cumulative effects take the lifetime integral and event effects take the worst rate in the critical window. Specify both, with the environment case that produces each.
Check yourself
A proton-sensitive part is used in a function that runs only during a ten-minute daily downlink on the 800 km polar orbit. Which environment case does its requirement need?
The peak trapped-proton flux during an anomaly pass (the downlink may coincide with one), and the design solar event, each with the probability of an event in ten minutes computed from the rate for that case. The mission-averaged flux is not relevant to this function, since it would understate the risk by a factor of about fifty on this orbit.
An upset analysis for a scrubbed memory on the 800 km polar orbit uses the orbit-averaged trapped-proton flux, 115 cm⁻² s⁻¹ above 10 MeV, and shows a comfortable margin on uncorrectable errors. Why is that result generally not trustworthy for the anomaly passes or for a solar event?
The uncorrectable rate of a scrubbed memory scales with the square of the upset rate times the scrub interval, so a tenfold rise in the upset rate raises the uncorrectable rate a hundredfold. On this orbit the trapped-proton flux is nonzero only 12 % of the time and peaks at 5,760 cm⁻² s⁻¹, 50 times the average, and the 95 % worst solar event delivers about a year of protons, 5.3 × 10⁹ per cm², in a few days. The scrub interval has to be sized for the worst window, and the requirement has to name that window and its environment case.
A latchup rate of 10⁻⁴ per device-day is quoted as the mission average. What probability does that give over a ten-day critical operation, and what condition does the lesson attach to the arithmetic?
For a Poisson process the probability of at least one event in a window T is 1 − exp(−rT), close to rT when rT is small, which gives about 0.1 % over ten days and 3.6 % per device-year. The condition is that r must be the rate for that window: during a solar event the rate can be larger by orders of magnitude for the days of the event, and on an anomaly pass the flux is tens of times the orbit average, so a mission average understates any window that overlaps either. The same expression gives the fluence a beam test needs to demonstrate a rate bound with no events (chapter 3).
The worked example shows one August 1972 event adding 505 rad(Si) on the polar orbit, 4 % of the five-year trapped dose behind 100 mil, and calls solar events a correction for dose. Why does the same lesson say they can be the main term for a solar cell or an optocoupler?
Because those parts degrade by displacement damage, and the trapped dose at this depth is nearly all from electrons, which do far less displacement damage per particle than protons. For displacement damage the quantity that matters is the proton fluence, and at 800 km the 95 % worst event delivers about as many protons above 10 MeV as a year of trapped protons (5.3 × 10⁹ per cm², ratio 0.99), so the event term is comparable to the trapped term rather than a few percent of it. The cumulative requirement should use the ESP active-year fluence at the project confidence level (lesson 3), and the two effects generally need separate lines in the sheet.
An availability requirement inherited from an ISS program is written against the anomaly pass and is reused unchanged for the 800 km polar orbit. What changes between the two orbits?
On the ISS orbit the solar-event share of the yearly protons is about 1 %, so the anomaly passes set the peak; the largest flux on the track is 743 cm⁻² s⁻¹, 80 times the average, with nonzero flux 8.5 % of the time. On the polar orbit the anomaly peak is 5,760 cm⁻² s⁻¹, 50 times an average that is itself about ten times higher, with nonzero flux 12 % of the time, and the 95 % worst event delivers essentially a year of protons in a few days (ratio 0.99). The requirement therefore needs a second window, the design solar event, and a higher anomaly-pass rate, each with its environment case named.
Further reading
- M. A. Xapsos et al., "Inclusion of radiation environment variability in total dose hardness assurance methodology," IEEE Trans. Nucl. Sci., vol. 64, no. 1, pp. 325-331, 2017. doi:10.1109/TNS.2016.2607021.
- P. Jiggens et al., "Long-term destructive SEE risk and calculations using multiple worst-case events versus modelling," IEEE Trans. Nucl. Sci., vol. 61, no. 4, pp. 1695-1702, 2014. doi:10.1109/TNS.2014.2302994.
- NASA, "Avionics Radiation Hardness Assurance (RHA) Guidelines," NASA/TM-20210018053, 2021. NTRS 20210018053.
- Space-RHA, "Space radiation environment in 3D," space-rha.com/space-radiation-environment (the hardware panel supplies the dose figures used here).
- G. P. Ginet et al., "AE9, AP9 and SPM: new models for specifying the trapped energetic particle and space plasma environment," Space Sci. Rev., vol. 179, pp. 579-615, 2013.