SEL Test-LET Requirement Tool

1. Overview: What Question Does This Tool Answer?

Suppose you are building a system from commercial, unhardened CMOS parts and you cannot heavy-ion test every part to destruction-level fluences. A practical compromise is a screening requirement: test every potentially SEL-susceptible part up to some LET, LET_T, and fly only parts that show no latchup below that level. The question is: how high does LET_T need to be so that the system, as a whole, meets a stated reliability over the mission?

This tool answers that question using the statistics of historical single-event latchup test data, the augmented CERN + JPL database of 66 unhardened CMOS part types, rather than data on your specific parts, which by assumption you do not yet have. It implements and extends the system-level method of Ladbury & Joplin (IEEE TNS 2026), built on the historical prior of Ladbury, Allen, Irom, Gaza, Vartanian, Barth & Hodson (IEEE TNS 2025). The extensions unique to this implementation are CREME96 environment-based rate integration (in place of the Petersen figure-of-merit), an episodic solar-flare exposure model, and an n-of-m constellation survivability criterion.

2. Quick Start

1. Pick a reliability criterion, the paper criterion (expected <1 SEL per mission) for comparability with the published method, the strict zero-SEL criterion for conservative single-spacecraft requirements, or the constellation criterion for n-of-m fleet survivability.
2. Set the confidence and reliability targets, e.g. 90% confidence, 99% reliability.
3. Enter the part count, either the number of SEL-susceptible parts directly, or total unhardened CMOS count times a susceptible fraction. For the constellation criterion this is the count per satellite.
4. Set mission duration and environment, GCR solar minimum is the default quiet-time GEO/interplanetary case; add flare episodes or choose the composite for solar-active exposure.
5. Compute. The large number is the required test LET in MeV·cm²/mg. Read the verdict sentence below it for the exact statistical statement being made, and check the cross-checks line for how other criteria/methods compare.

3. The Statistical Model in Plain Language

Each Monte Carlo trial builds one hypothetical system from parts that are statistically similar to the historical database:

• Onset LET (LET₀). Each part draws an onset LET from an empirical distribution of SEL-susceptible unhardened CMOS parts. Two priors are selectable under Advanced: the default Extended Historical SEL Priors (EHSP 2026-09) global onset CDF (739 uniform LET@1e-8 onsets from Poisson-MLE Weibull fits and onset brackets, mechanism-screened, all eras; F(37) = 0.68, F(75) = 0.93, F(100) = 0.98, closure extrapolated to 120 MeV·cm²/mg) and the original TNS 2025 66-part CDF (0.5–72 MeV·cm²/mg), retained for traceability to the published curves. The CDF table is directly editable. The derivation of the model is documented in the SEL Threshold Predictor methodology.
• Limiting cross section (σ_s). Sampled from the lognormal trend of the historical data: parts with low onset LET tend to have large saturated cross sections, and the scatter is wide (s_ln = 2.45).
• SEL rate. In the CREME96 environment modes, a full Weibull cross-section curve is constructed (shape parameters drawn from the 31 CERN database Weibull pairs) and convolved with the selected LET spectrum, a standard spectral rate integration. In the FOM modes, the published bound-and-reduce method is used instead: the Petersen figure-of-merit bound C_E·σ_s/LET₀² reduced by a beta-distributed ratio fitted to the same 31 Weibull pairs.
• Screening. Testing to LET_T removes parts with onset below LET_T; removed parts are replaced with fresh draws (themselves screened once). Independently, each part carries a 3.5% chance of remaining susceptible just above the test LET regardless of screening, the 90%-confidence bound implied by 0-of-66 database parts violating their screen. This "beyond-database floor" is what eventually limits how much screening can buy you.
• System rate. Part rates sum. Repeating over thousands of realizations of the prior produces a distribution of possible system rates, from which the tool evaluates your chosen criterion and bisects on LET_T to find the requirement.

4. Choosing a Reliability Criterion

Paper criterion: expected <1 SEL per mission (Pₛ)

Reliability is the fraction of Monte Carlo realizations whose expected SEL count over the mission stays below one. This is the definition behind Figs. 3–6 and the closed-form Eq. (4) of the system-level paper, so use it when you want numbers traceable to the published curves. It is the most permissive of the three criteria: it tolerates an expected count approaching one event per mission.

Strict criterion: zero SELs at confidence

Takes the confidence-level quantile of the system rate distribution and requires the Poisson probability of zero SELs over the mission to meet the reliability target. Every SEL is treated as mission-ending. At 99% reliability this tolerates a rate roughly 100× lower than the paper criterion (−ln 0.99 ≈ 0.01 expected events vs. 1), so the required LET_T is substantially higher, often pushed into the regime where the 3.5% beyond-database floor governs. Use it for single-string spacecraft where any latchup is unacceptable.

Constellation criterion: n-of-m satellites SEL-free

For a fleet of m identical satellites, each carrying the specified number of susceptible parts, requires that at least n of m remain SEL-free through the mission, at the stated confidence and reliability. Two structural assumptions matter:

• Shared design ⇒ correlated uncertainty. All satellites carry the same part types, so each Monte Carlo realization draws one set of part characteristics and applies it fleet-wide. A bad prior draw is bad for every satellite simultaneously. This is why constellation reliability is not simply the single-satellite result raised to a binomial power, treating satellites as independent would be optimistic.
• Independent occurrences. Given the design, actual SEL events in flight are independent Poisson processes per satellite. Per realization the per-satellite survival is p = exp(−R·T_M) and constellation survival is the binomial tail P(≥n of m).

Because the binomial tail is monotone in R, the confidence quantile is applied to the rate distribution and the tail evaluated there. Setting n = m = 1 exactly recovers the strict criterion. Redundancy earns real credit: accepting the loss of 2 of 8 satellites typically reduces the required LET_T by 15–20 MeV·cm²/mg relative to a single-satellite zero-SEL requirement at the same confidence.

5. Environments and the Flare Exposure Model

Continuous environments

GCR Solar Minimum (galactic cosmic rays behind 100 mil aluminum, CREME96, solar-quiet worst case for GCR) and ISS (400 km, 51.6° geomagnetically shielded) are applied for the full mission duration. These are the appropriate baselines for long-duration missions. GCR solar minimum is the default and matches the "GEO / interplanetary" intent of the source paper's C_E = 400.

Episodic (flare) environments

Worst Week, Worst Day, and Peak 5-minute are CREME96 solar particle event environments. They are transient by definition, and applying them for a full mission would be physically meaningless, a "10-year worst week" overstates the exposure by a factor of 500. The tool therefore treats them as episodes: you specify the expected number of events per year, and each event contributes its flux for its episode duration only (7 days, 1 day, and 5 minutes respectively). Because Poisson exposures add, this folds into an exact mission-averaged rate:

All reliability math downstream (expected counts, zero-SEL Poisson, constellation binomial) remains exact over the mission, and the mission-duration sweep chart stays meaningful, longer missions accumulate proportionally more episodes.

Composite mission

GCR + worst-week solar events applies GCR solar minimum continuously plus your stated number of worst-week episodes per year. This is the recommended design case for solar-active mission phases. Be aware that for SEL the flare episodes usually dominate: even one worst week per year typically contributes more to the requirement than ten years of quiet-time GCR.

Petersen FOM modes

The two FOM options reproduce the source-paper engine exactly (verified bit-identical): the figure-of-merit bound with C_E = 400 (GEO/interplanetary GCR) or a custom C_E, reduced by the beta-distributed bound-to-rate ratio. Use these for direct comparison with the published curves. When a CREME96 environment is primary, the tool automatically reports the FOM equivalent in the cross-checks line. The two methods are statistically consistent where they should be, at LET₀ = 10 MeV·cm²/mg the mean spectral-to-bound rate ratio in GCR solar minimum (0.18) matches the mean of the paper's beta distribution, but the FOM is a bound, so FOM-mode requirements typically run 10–20 MeV·cm²/mg higher than GCR spectral-mode requirements.

6. Input Reference

7. Reading the Results

The hero number is the required test LET in MeV·cm²/mg. "none" means the untested system already meets the target; a dash with a red badge means no amount of heavy-ion screening reaches the target (the 3.5% floor dominates) and you should consider redundancy, SEL-immune part selection, or mitigation credit instead.

The verdict sentence states exactly what the number means, including the criterion, confidence, environment, and exposure model. Read it carefully, the same LET value under different criteria means very different things.

The badges flag conditions worth noticing: the active environment, part counts outside the 5–50 range studied in the source paper, and results above the highest onset LET in the database (≈72 MeV·cm²/mg), which are extrapolations governed by the beyond-database floor.

The metric tiles give the system (or per-satellite) rate at the confidence level, expected SEL count over the mission, achieved reliability, and, in constellation mode, the expected number of surviving satellites and P(≥n of m).

The cross-checks line reports the same problem solved by the other methods: the closed-form Eq. (4) fit (FOM-based paper criterion), the Monte Carlo under the alternative criterion, and, when a CREME96 environment is primary, the Petersen-FOM equivalent. Large disagreements are informative: paper vs. strict shows the price of "zero events" vs. "less than one expected"; spectral vs. FOM shows the conservatism of the bound.

8. The Three Charts

• System failure probability vs. test LET. How the failure probability (under your criterion) falls as the screening level rises. The marker is the solved requirement at your target. The curve is roughly exponential in LET_T until the beyond-database floor flattens it.
• System SEL rate vs. confidence. The bounding rate distribution over realizations of the prior, untested vs. screened at the solved LET_T. The horizontal gap between the curves is the screening benefit, typically two to three orders of magnitude. The note below reports the fraction of screened realizations with zero remaining susceptible parts.
• Required LET_T vs. mission duration. The requirement swept over mission length at fixed confidence/reliability/part count (reduced-trial solves; expect slight jitter). In episodic-flare modes the episode count scales with duration, so this curve remains meaningful. The dashed closed-form curve appears when inputs are inside the paper's fit validity range.

9. Worked Examples

Values below are approximate (Monte Carlo with default settings; small variations with trials and seed are expected). All use 20 susceptible parts, 99% reliability, 90% confidence.

Example A: single GEO spacecraft, 5-year mission, quiet sun

GCR Solar Minimum, paper criterion, EHSP 2026-09 prior: LET_T ≈ 39 (≈36 with the TNS 2025 prior; the EHSP prior's heavier high-LET tail adds a few MeV·cm²/mg). The FOM cross-check gives ≈57, so the figure-of-merit bound costs roughly 18 MeV·cm²/mg of test requirement relative to spectral integration. Switching to the strict zero-SEL criterion raises the spectral requirement to ≈64: the price of "no events" versus "less than one expected event."

Example B: 10-year mission with solar activity

GCR-only, paper criterion: ≈47. Adding one worst-week episode per year (composite): ≈68. The flare episodes, 70 days of worst-week flux out of 3652 mission days, dominate the requirement. This is the correct episodic treatment; naively running worst-week flux for the whole decade would overstate the flare exposure by a factor of about 50 and push the requirement far past 100 MeV·cm²/mg.

Example C: 8-satellite constellation, 6-of-8 at 5 years

GCR Solar Minimum, constellation criterion, 20 parts per satellite: LET_T ≈ 48, with per-satellite zero-SEL probability ≈94% and expected survivors ≈7.5 of 8 at the solution. Compare ≈64 for a single satellite under the strict criterion: tolerating the loss of 2 of 8 satellites buys roughly 16 MeV·cm²/mg of screening relief. Requiring all 8 (n = m = 8) pushes the requirement to ≈90, fleets with no redundancy margin are harder than single spacecraft, because there are 8× as many parts that all must behave.

Example D: technology mix

A 20-part satellite built as 10 bulk-CMOS + 4 ADC/DAC + 2 SRAM-FPGA + 4 SOI parts has an effective N_P of about 8.7 and requires LET_T ≈ 33 (paper criterion, 5-yr GCR), versus ≈39 if all 20 were counted as susceptible and ≈34 at a generic 50% fraction. An all-SOI design drops to ≈19. The mix mode makes the technology-selection trade visible directly in the screening requirement.

10. Advanced Settings

11. Accuracy, Caveats, and Limitations

• This is a prior-based screening tool, not a rate predictor for known parts. If you have measured Weibull parameters for your parts, use the SEE Rate Assessment Tool's part-level methods instead, the historical prior is for parts you have not yet tested.
• Results above the prior's observed range are extrapolation. The highest onset LET is about 72 MeV·cm²/mg in the TNS 2025 prior and about 100 in the EHSP 2026-09 prior (closure extrapolated to 120); above the active prior's range the requirement is governed by the 3.5% beyond-database floor and should be read as "screening alone is marginal here." Note the 3.5% floor is retained from the published method (0-of-66 bound) even under the larger EHSP prior, which is conservative given that prior already quantifies its own high tail.
• Validation. Against the source papers this implementation reproduces the untested rate distributions of Fig. 2, the >200× screening rate reduction at 20 MeV·cm²/mg, and required-LET solutions typically within 10–15% (on the conservative side) of the paper's fitted curves. The FOM engine is bit-identical to the original implementation; the closed-form coefficients follow the figure-validated fits (correcting three apparent typos in the printed equations).
• Weibull shape independence. Spectral modes sample (w, s) independently of onset LET, consistent with the site's SEE Rate tool. To the extent shape correlates with onset in reality, low-onset parts' rates may be somewhat overestimated; this mainly affects untested-rate displays, not the solved requirement (which is set near the screening level).
• Environment scope. All CREME96 tables are behind 100 mil aluminum. Very different shielding depths, deep-space vs. magnetospheric trajectories, or solar-cycle averaging require externally generated spectra.
• SEL only. The prior, the susceptibility fractions, and the criteria are specific to destructive latchup in unhardened CMOS. Do not reuse the numbers for SEU, SET, or SEB/SEGR requirements.
• Statistical, not physical, screening. The model assumes screening at LET_T is performed to adequate fluence and temperature per your test standard; test escapes beyond the modeled 3.5% floor are not included.

12. References