Tools / Parts and analysis / Proton and heavy ion cases
Does a clean proton test rule out heavy-ion latchup?
What paired test data say about proton screening: case studies from the combined test corpus and the public table of parts tested with both protons and heavy ions, showing when a clean proton result predicts latchup immunity and when it does not.
Proton vs. Heavy-Ion SEL95 parts with both heavy-ion and proton SEL data8 case studies92 public paired partsProton Proxy Risk Toolmethodology
Scope:
Every part below is drawn from published sources (IEEE REDW and RADECS workshop and journal records, NASA GSFC test reports).
Records were re-extracted from the source documents and, where per-run data existed, refit with a Poisson-Weibull; heavy-ion onset is the uniform LET@10⁻⁸ crossing where a fit exists, else the author-stated threshold.
Numbers are quoted as reported; consult the source before relying on any of them for a design decision.
Companion tool: Proton Proxy Risk Tool · methodology.
Not on its own. Protons found latchup in 20 of the 37 heavy-ion-susceptible parts with a known onset, none of them with an onset above 20, and 7 of the 30 parts with an adequate proton null were still heavy-ion susceptible below LET 37.
Parts with both heavy-ion and proton SEL data
95
in the corpus
Heavy-ion-susceptible parts where protons found latchup
20 of 37
with a known heavy-ion onset
Median onset of the parts protons found
3.0
MeV·cm²/mg; 16 of 20 below LET 10
Adequate proton nulls on parts susceptible below LET 37
7 of 30
23%, 90% interval 13–38%
The picture in one paragraph
Of 95 parts in the corpus with both heavy-ion and proton SEL data, protons found latchup in 20 of the 37 heavy-ion-susceptible parts with a known onset. The ones they found have a median onset of 3.0 MeV·cm²/mg (16 of 20 below LET 10); no part with an onset above 20 has ever produced a proton latchup in this corpus. The ones they missed span the full range, and among the 30 parts with an adequate null proton test (≥ 50 MeV, ≥ 10¹⁰ p/cm²), 7 (23%, 90% interval 13–38%) were heavy-ion susceptible below LET 37 and 5 (17%) below LET 15. A proton null is evidence, not a threshold: it discounts low-onset, large-cross-section, shallow-volume behaviors and says nothing about the rest. The cases below show each of the mechanisms behind that sentence.
| Heavy-ion onset bin (LET@10⁻⁸) | Proton SEL observed | Adequate proton null | Detection fraction |
|---|---|---|---|
| 0–5 | 14 | 3 | 82% [64–93%] |
| 5–10 | 1 | 2 (+2 weaker nulls) | 33% [6–76%] |
| 10–20 | 5 | 2 (+2 weaker nulls) | 71% [41–91%] |
| 20–40 | 0 | 0 (+4 weaker nulls) | no events |
| ≥ 40 | 0 | 0 (+2 weaker nulls) | no events |
Cross-section ratio: where both cross sections were measured (n=5), σp(≥ 100 MeV)/σHI
has a median of 10-7.09 and spans 10-10.37–10-6.05, one to two decades
below the 1-in-289,000 (10-5.5) recoil-count ceiling of Ladbury & Lauenstein, because most recoils that do occur deposit less than the part's onset.
Case studies 8 parts, heavy ions and protons side by side
Each card pairs the heavy-ion characterization with the proton result, states what the Proton Proxy engine predicts from the heavy-ion curve alone, and draws the lesson. "SV" is the sensitive-volume depth the recoil must traverse; it is the single largest uncertainty in every proton-to-heavy-ion argument.
Xilinx XCZU9EG (Zynq UltraScale+ MPSoC)
Heavy ions
- Onset LET@10⁻⁸:
- 1.2 MeV·cm²/mg (Weibull L₀ 1.2, W 35, s 2.2)
- σsat:
- 1.5×10⁻³ cm²/device
- Nature:
- non-destructive, cleared by power cycle
Protons
- 30 MeV:
- no SEL (upper limit only)
- 50 MeV:
- SEL, σ = 4.1×10⁻¹² cm² (to 6×10¹² p/cm² ~1 Mrad)
- 100 MeV:
- 4.2×10⁻¹¹
- 200 MeV:
- 6.7×10⁻¹¹ cm²
- 0.15–12 MeV:
- no SEL (direct-ionization LET ≤ 0.44)
What the Proton Proxy engine says: From the heavy-ion curve alone the recoil emulator predicts σp(50 MeV) = 1.8–12×10⁻¹² cm² for SV depths 10–2 µm, the measured 4.1×10⁻¹² sits inside the band. At 200 MeV it predicts 0.7–3.8×10⁻¹¹ against a measured 6.7×10⁻¹¹: right order, but the emulator's energy scaling (×3–4 from 50 to 200 MeV) is flatter than this part's ×16.
Takeaway. A part that latches at LET 1.2 is exactly the kind protons do find, but only above a proton energy threshold (here between 30 and 50 MeV), and the cross section keeps climbing to 200 MeV. A null at 30 MeV would have been worthless; the ratio σp/σHI ≈ 4×10⁻⁸ even at 200 MeV shows how few recoils reach the deposition this device needs.
Sources: Koga et al. NSREC REDW 2018; Davis et al. Aerospace compendia, REDW 2019 and 2021; Koga, Davis & Mabry, REDW 2020.
Cypress CY7C1069 (4 Mbit SRAM): flight-validated
Heavy ions
- Onset LET@10⁻⁸:
- 3.2 (Weibull L₀ 3.2, W 14.9, s 2.2)
- σsat:
- 0.37 cm²/device (whole-die)
- Bias dependence:
- none significant across three static conditions
Protons
- Ground:
- proton Weibull in energy: E₀ = 25 MeV, σsat = 1.2×10⁻⁸ cm²
- In flight (CARMEN/MEX on JASON-2, SAC-D):
- 1,140 SEL in the first campaign, 1.12 per device-day, 97.4% inside the South Atlantic Anomaly
- 14-year record:
- SEL rate tracks trapped-proton exposure across four orbits
What the Proton Proxy engine says: Predicted σp(200 MeV) = 5×10⁻⁹ cm² at 5 µm (2×10⁻⁸ at 2 µm) against the measured 1.2×10⁻⁸, within ×2.
Takeaway. The cleanest demonstration that proton-induced latchup is a mission-driving mechanism for low-onset, large-σ parts: trapped protons in the SAA delivered more than one latchup per device per day, and the ground proton curve predicted it. For a part like this the proton test is not a proxy at all; it is the environment.
Sources: Bezerra et al. RADECS 2007, 2009, 2011; Bezerra et al. IEEE TNS 2023 (RADECS 2022); Coronetti et al. REDW 2024 (HEARTS@CERN).
Samsung K6R4016V1D (4 Mbit SRAM)
Heavy ions
- Onset LET@10⁻⁸:
- 7.2 (2024 fit); other campaigns quote 12–18, a rising curve with a low-LET tail attributed to ion nuclear reactions
- σsat:
- 0.10 cm²/device (laser mapping: a 170× more sensitive array region drives it)
- Mixed field:
- σ varies >10× with spectral hardness (tungsten in the metallization)
Protons
- 200 MeV:
- SEL, σsat = 4.3×10⁻¹⁰ cm² fitted proton energy threshold 50.9 MeV; no events at the two lowest energies
- In flight (PROBA-II):
- 0.4 SEL/device-day, about 2× the CREME96 prediction from room-temperature ground data (flight at ~45 °C)
What the Proton Proxy engine says: With the steep fitted curve (W 0.9) and σsat 0.10 the emulator over-predicts σp at shallow depth (2×10⁻⁸ at 2 µm) and under-predicts at 10 µm (5×10⁻¹¹); the measurement (4.3×10⁻¹⁰) is bracketed at about 7 µm. The SV depth you assign is the dominant uncertainty; the tool exposes it as an input for that reason.
Takeaway. Three lessons in one part: proton SEL has an energy threshold (51 MeV) set by recoil deposition versus SV depth; heavy-metal metallization hardens the mixed-field response; and a room-temperature proton curve under-predicted flight by 2× because the spacecraft ran hot. Temperature enters the SEL model as a ×0.74 onset factor for exactly this reason.
Sources: D'Alessio et al. RADECS 2013 (PROBA-II); Cecchetto et al. IEEE TNS 2019; Mattos et al. IEEE TNS 2024 (PROBA-V); Coronetti et al. REDW 2024; Cecchetto et al. RADECS REDW 2025.
DDC BU65170G1 (MIL-STD-1553 BC/RT hybrid)
Heavy ions
- Processor die:
- SEL onset 3.3, σsat 5.6×10⁻² cm²
- Transceiver dice:
- no SEL to LET 55.9
Protons
- 80 MeV:
- no SEL to 10¹¹ p/cm²
- Higher energies:
- SEL, σsat = 2×10⁻¹¹ cm² at 190 MeV
What the Proton Proxy engine says: Taken at face value (onset 3.3, σ 0.056, 2–5 µm SV) the emulator expects 120–400 events at 80 MeV and 10¹¹ p/cm²; zero were seen. The null therefore excludes that shallow-SV model, the die's latch path needs deposition that only higher-energy recoils supply, and the 190 MeV cross section (2×10⁻¹¹) is 30–350× below the point prediction. Multi-die packages are the engine's weakest geometry.
Takeaway. A proton null at one energy followed by latchup at a higher one: the recoil energy spectrum matters as much as fluence. When a proton null conflicts with a known low heavy-ion onset, the resolution is usually depth (a buried latch path) or package geometry, not a mistake in either test.
Sources: Malou et al. RADECS 2007 (compendium of TID and SEL results for candidate spacecraft electronics).
National LM2991 and the ISS component survey
Heavy ions
- Onset:
- 5.85 MeV·cm²/mg (²⁰Ne), the lowest of 38 component types in the survey; project requirement was 36
- σ:
- not reported in the corpus record
Protons
- 250 MeV, 13 part types re-tested:
- no SEL in any type to 2×10¹¹ p/cm² including the LM2991
What the Proton Proxy engine says: Assuming a modest σHI of 10⁻³ cm² the expected proton event count at 2×10¹¹ is 6 (2 µm), 0.75 (5 µm) or 0.04 (10 µm): the null is entirely consistent with a deep latch path in a power IC. The tool would report the low-onset models as unbounded for SV depths ≥ 5 µm.
Takeaway. The archetype of the problem: a part with the lowest heavy-ion threshold in a 38-type survey sailed through a 250 MeV proton screen at 2×10¹¹. Power ICs have deep, well-defined pnpn paths that recoils rarely reach, a proton null on a regulator says almost nothing about its heavy-ion latchup.
Sources: Reinecke, Pedersen & Harboe-Sørensen, RADECS 2003, "Heavy Ion and Proton SEL Characterization on Selected EEE Component Types Used in ISS Equipment".
Freescale/Everspin MR2A16A (4 Mbit MRAM)
Heavy ions
- Onset:
- 6.6–7 MeV·cm²/mg (range-limited thinned samples; latch sites in peripheral logic, not the array)
- σsat:
- 6×10⁻⁴–3×10⁻³ cm²/device
Protons
- 200 MeV, three samples:
- no SEL to 10¹¹ p/cm² each (room temperature, nominal bias)
- GSFC:
- no SEL during proton exposures; report notes the test was not at worst-case temperature or voltage
What the Proton Proxy engine says: Expected proton events at 10¹¹: 0.33 (2 µm), 0.01 (5 µm), ~0 (10 µm). A null was the expected outcome even though the part latches at LET 7 with a cross section near 10⁻³ cm² the test simply could not see it.
Takeaway. Three samples, 10¹¹ protons each, zero events, and a heavy-ion onset of 7. This is the case to keep in mind when someone proposes a proton screen as SEL qualification: at 200 MeV and this fluence the expected event count for a shallow SV was below one, and the test was run cold and at nominal bias on top of that.
Sources: Nuns et al. RADECS 2007 and IEEE TNS 2008; Nguyen & Irom, RADECS 2007; Oldham, Pham & Friendlich, NASA GSFC test report 2008.
IDT 7201T FIFO: the epitaxial-thickness series
Heavy ions
- 12 µm epi:
- latched at the first ion tried, LET 26.6
- 10 µm:
- ≤ 26.6 (possible threshold)
- 8 µm:
- 50–60
- 6 µm:
- no SEL to LET 80, σ < 10⁻⁶ cm²
Protons
- 63 MeV (UC Davis):
- no SEL on any epi variant
What the Proton Proxy engine says: With onsets of 27 and above, the recoil spectrum at 63 MeV cannot reach the required deposition at any SV depth: the emulator predicts zero proton events for every variant. The proton null is uninformative here, and the tool reports the whole low-onset model grid as unbounded.
Takeaway. A single device family spanning onset 27 to >80 purely through epi thickness, and a proton test that could not tell the variants apart. Epi (or well) depth sets both the heavy-ion onset and the depth of the sensitive volume; the two quantities that decide proton coverage move together, and in the wrong direction for proton screening.
Sources: Crabtree & LaBel, GSFC heavy-ion test report (BNL, 1994); Seidleck & LaBel, GSFC proton test report (UC Davis, 1994); LaBel et al. NSREC REDW 1995.
STMicroelectronics STM32F103RGT6 (Cortex-M3 MCU)
Heavy ions
- Campaign 1:
- immune to SEL up to the maximum LET of 60 (room temperature)
- Campaign 2 (2024):
- "SEL observed: no"
Protons
- 75 MeV, 85 °C, worst case of four conditions:
- SEL, σ = 4.3×10⁻¹¹ cm²/device (10¹¹ p/cm²)
What the Proton Proxy engine says: A device declared heavy-ion immune to LET 60 cannot latch under 75 MeV protons in this model, predicted σp is zero at any depth. The measured 4.3×10⁻¹¹ is not an engine failure; it is a conditions mismatch: the heavy-ion immunity was established cold, the proton events were found hot.
Takeaway. Immunity claims carry their test conditions. The Extended Historical SEL Priors (EHSP) map room-temperature nulls to hot-equivalent LET by a factor 0.74 for this reason, and the proxy tool's Bayesian block uses hot-equivalent susceptibility priors. Read "immune to LET 60" as "immune to LET 60 at 25 °C and nominal bias".
Sources: Wind et al. RADECS 2022 (ESA CORHA study); Wind et al. RADECS REDW 2024.
Also in the corpus:
TI ADS1271 (24-bit ADC; heavy-ion onset ~10, σsat 5–7×10⁻⁵): proton SEL at 4×10⁻¹² cm² (230 MeV) growing more than 40× between 100 MeV and 24 GeV at CHARM, driven by tungsten vias, the silicon-recoil emulator predicts essentially nothing, the tool's high-Z option reproduces the measurement within ×2.
Microsemi M2GL010T Igloo2: high-current reset events at 40–50 MeV protons, none at 20 MeV (heavy-ion L₀ 2.0, σ 0.05).
Xilinx XC7Z020 Zynq-7000: VccAUX latchup at LET 16 under heavy ions; 105 MeV protons to 1.1×10¹¹ found nothing, the authors themselves framed the null as "equivalent to 21.5 years of heavy-ion exposure up to LET ~10", which is the coverage concept this tool computes.
Zarlink GP2021 (GPS correlator): non-destructive micro-latch at LET 15–20 with σ > 0.01 cm²; under 190 MeV protons a rare event at 4×10⁻¹³ cm².
Microchip dsPIC30F6014A (ExoMars 2020): heavy-ion L₀ 3.0, W 35, s 1.9, σ 0.078; proton cross section fitted versus energy from 29 to 200 MeV and combined with the heavy-ion rate for the mission estimate.
All public paired parts 92 parts
The full public table (92 parts). Click a column header to sort. "Adequate" marks null tests with ≥ 50 MeV and ≥ 10¹⁰ p/cm²; "(fit)" onsets are LET@10⁻⁸ from the Poisson-Weibull fit, "(author)" are as stated in the source. Internal test-report pairs are not listed (they enter the aggregates above only).
| Part | Class | Heavy-ion status | HI onset / clean-to | Proton result | Emax (MeV) | Fluence (p/cm²) | T (°C) | σp (cm²) | Adequate | Sources |
|---|---|---|---|---|---|---|---|---|---|---|
| STM32F103RGT6 | MCU/Processor | immune to tested LET | > 60 | SEL | 75 | 1.0e+11 | 85 | 4.3e-11 | radecs paper:2022 | |
| 8116400-60PJ | DRAM/SDRAM | immune to tested LET | > 80 | null | 63 | 25 | gsfc test report:1994 | |||
| ADS1281 | ADC/DAC | immune to tested LET | > 87 | null | 230 | 6.9e-14 | redw paper:2013 | |||
| ADS5424 | ADC/DAC | immune to tested LET | > 55 | null | 195 | gsfc test report:2009 | ||||
| AM7968-125DC | Other | immune to tested LET | > 53 | null | 196 | 3.0e+10 | 25 | yes | gsfc test report:1997 | |
| CDCLVC1310 | Power | immune to tested LET | > 62 | null | 184 | 1.0e+11 | yes | redw paper:2022 | ||
| Comhvt (90nm TCV commercial-like SRAM, h | SRAM | immune to tested LET | > 120 | null | 198 | 1.0e+11 | 125 | yes | redw paper:2009 | |
| DG412 | Logic/Interface | immune to tested LET | > 86 | null | 230 | 2.0e+12 | yes | redw paper:2014 | ||
| EDE2108ABSE | DRAM/SDRAM | immune to tested LET | > 30 | null | 200 | 25 | 1.0e-09 | redw paper:2010 | ||
| FM22L16 | SRAM | immune to tested LET | > 74 | null | 198 | 1.3e+11 | 75 | yes | gsfc test report:2007 | |
| H5PS2G83AFR-S6C | DRAM/SDRAM | immune to tested LET | > 30 | null | 200 | 25 | 2.0e-11 | redw paper:2010 | ||
| HV583 | Power | immune to tested LET | > 55 | null | 63 | -243 | gsfc test report:2002, gsfc test report:2003, redw paper:2003 | |||
| IBM 45nm SOI SRAM test chip | SRAM | immune to tested LET | > 77 | null | 63 | redw paper:2009 | ||||
| IL611 | Logic/Interface | immune to tested LET | > 97 | null | 68 | 2.0e+10 | 25 | yes | radecs redw paper:2024 | |
| IO65LPHVT_TF3V3_1V8_2V5_3V3_FS_50A_7M4X0 | Other | immune to tested LET | > 60 | null | 230 | 5.0e+10 | 125 | yes | redw paper:2014 | |
| K4T2G084QA | DRAM/SDRAM | immune to tested LET | > 30 | null | 200 | 25 | 2.0e-11 | redw paper:2010 | ||
| K9F4G08U0A | Flash/NVM | immune to tested LET | > 87 | null | 200 | redw paper:2008 | ||||
| LEON5FT/NOEL-VFT SoC Test Chip | MCU/Processor | immune to tested LET | > 92 | null | 230 | 1.0e+11 | yes | radecs redw paper:2023 | ||
| LT1499HS#PBF-ND | Amplifier | immune to tested LET | > 60 | null | radecs paper:2022 | |||||
| LTC6240HVCS8#PBF-ND | Amplifier | immune to tested LET | > 60 | null | radecs paper:2022 | |||||
| LTC6400-20 | ADC/DAC | immune to tested LET | > 50 | null | 198 | 1.0e+12 | 25 | yes | gsfc test report:2009 | |
| LX25 | FPGA/SoC | immune to tested LET | > 75 | null | 195 | redw paper:2007 | ||||
| Low Power Pentium MMX | MCU/Processor | immune to tested LET | > 10 | null | 200 | 1.5e+13 | yes | redw paper:2001, redw paper:2002 | ||
| MPFS250T | FPGA/SoC | immune to tested LET | > 68 | null | 200 | 2.0e+12 | 100 | 8.7e-12 | yes | radecs paper:2022, redw paper:2022 |
| MT47H256M8 | DRAM/SDRAM | immune to tested LET | > 30 | null | 200 | 25 | 1.0e-09 | redw paper:2010 | ||
| PLL_PG_1201x_6P_CMOS065LP | Other | immune to tested LET | > 60 | null | 230 | 5.0e+10 | 125 | yes | redw paper:2014 | |
| Pentium III | MCU/Processor | immune to tested LET | > 28 | null | 200 | 3.3e+11 | yes | gsfc test report:2000, gsfc test report:2001, gsfc test report:2002, redw paper:2001 | ||
| Pentium MMX | MCU/Processor | immune to tested LET | > 10 | null | 120 | 1.6e+13 | yes | redw paper:2000, redw paper:2001 | ||
| PolarFire SoC MSS | FPGA/SoC | immune to tested LET | > 37 | null | 64 | 1.0e+11 | 25 | yes | redw paper:2024 | |
| SN65LVCP15 | Logic/Interface | immune to tested LET | > 60 | null | redw paper:2011 | |||||
| SN65LVCP23 | Logic/Interface | immune to tested LET | > 60 | null | redw paper:2011 | |||||
| SN65LVDS100 | Logic/Interface | immune to tested LET | > 60 | null | redw paper:2011 | |||||
| SN65LVDS20 | Logic/Interface | immune to tested LET | > 60 | null | redw paper:2011 | |||||
| SN65LVDS250 | Logic/Interface | immune to tested LET | > 60 | null | redw paper:2011 | |||||
| TLE4945L | Amplifier | immune to tested LET | > 103 | null | 68 | 2.0e+10 | yes | radecs paper:2022 | ||
| TLK2711 | Logic/Interface | immune to tested LET | > 60 | null | 250 | redw paper:2008 | ||||
| TMS570LS3137 | MCU/Processor | immune to tested LET | > 55 | null | 200 | 1.4e+11 | yes | redw paper:2018 | ||
| TPS7H1101-SP | Power | immune to tested LET | > 63 | null | 1000 | 1.0e+10 | yes | redw paper:2020 | ||
| XC2V1000 | FPGA/SoC | immune to tested LET | > 63 | null | 120 | 2.1e+11 | yes | redw paper:2004 | ||
| XC3S50 | FPGA/SoC | immune to tested LET | > 10 | null | 120 | 2.7e+11 | yes | redw paper:2004 | ||
| XC4VLX25 | FPGA/SoC | immune to tested LET | > 55 | null | 198 | redw paper:2006, redw paper:2008 | ||||
| XC6VLX240T | FPGA/SoC | immune to tested LET | > 10 | null | 105 | redw paper:2012 | ||||
| XC7Z030 | FPGA/SoC | immune to tested LET | > 60 | null | 9.9e+11 | radecs redw paper:2019 | ||||
| XCKU060 | FPGA/SoC | immune to tested LET | > 80 | null | 200 | 1.0e+12 | 125 | yes | radecs redw paper:2019 | |
| XCVC1902 | FPGA/SoC | immune to tested LET | > 80 | null | 105 | 2.0e+12 | 120 | yes | radecs paper:2022, redw paper:2021, redw paper:2022, redw paper:2023 | |
| M2GL010T-1FCC484 | FPGA/SoC | susceptible | 0.7 (fit) | SEL | 50 | redw paper:2017 | ||||
| IS61LV5128AL-12 | SRAM | susceptible | 1.0 (author) | SEL | 52 | radecs paper:2011, radecs paper:2013, radecs redw paper:2024 | ||||
| TC55VD836 | SRAM | susceptible | 1.0 (fit) | SEL | 200 | redw paper:2003 | ||||
| XCZU9EG | FPGA/SoC | susceptible | 1.2 (author) | SEL | 200 | 6.0e+12 | 6.7e-11 | yes | redw paper:2018, redw paper:2019, redw paper:2020, redw paper:2021 | |
| HM628512 | SRAM | susceptible | 1.7 (author) | SEL | radecs paper:2009, radecs paper:2011, radecs paper:2023 | |||||
| L64811 | MCU/Processor | susceptible | 1.7 (author) | SEL | redw paper:1993 | |||||
| STM32L152RET6 | ADC/DAC | susceptible | 2.0 (fit) | SEL | 75 | 1.0e+11 | 85 | 1.8e-08 | radecs paper:2022 | |
| BS62LV1600EIP55 | SRAM | susceptible | 2.4 (author (TNS 2021 Weibull)) | SEL | 200 | 3.0e+11 | 3.5e-08 | radecs paper:2016, radecs paper:2018, radecs paper:2024 | ||
| Brilliance SRAM (specific part number no | SRAM | susceptible | 2.4 (author) | SEL | 200 | radecs paper:2020, radecs paper:2021 | ||||
| BS616LV1611 | SRAM | susceptible | 3.0 (fit) | SEL | 185 | 3.5e-07 | radecs paper:2019, radecs paper:2021, radecs paper:2023 | |||
| dsPIC30F6014A-30I/PF | MCU/Processor | susceptible | 3.0 (author) | SEL | 200 | radecs redw paper:2018 | ||||
| CY7C1069 | FPGA/SoC | susceptible | 3.2 (author) | SEL | 100 | 1.2e-08 | radecs paper:2009, radecs paper:2011, radecs paper:2023, redw paper:2011 | |||
| BU65170G1 | MCU/Processor | susceptible | 3.3 (author) | SEL | 190 | 1.0e+11 | 2.0e-11 | yes | radecs paper:2007 | |
| HM65162 | SRAM | susceptible | 4.5 (author) | SEL | radecs paper:1995, radecs paper:1996, radecs paper:1999 | |||||
| K6R4016V1D-TC10 | SRAM | susceptible | 7.2 (author) | SEL | 200 | 1.0e+12 | 45 | 4.3e-10 | radecs paper:2013, radecs paper:2024 | |
| ADS1271 | ADC/DAC | susceptible | 10.0 (author) | SEL | 230 | 4.0e-12 | radecs paper:2018, redw paper:2013 | |||
| GP2021 | Other | susceptible | 14.3 (fit) | SEL | 190 | 1.2e+12 | 4.3e-13 | gsfc test report:2003 | ||
| SRAM B | SRAM | susceptible | 15.0 (author) | SEL | 200 | 4.6e+11 | 25 | 1.3e-08 | radecs paper:2017, radecs paper:2025 | |
| CLARO8v3 | Amplifier | susceptible | 17.2 (fit) | SEL | 24000 | 3.0e+15 | yes | redw paper:2018 | ||
| SRAM C | SRAM | susceptible | 18.0 (author) | SEL | 200 | 1.0e+10 | 25 | radecs paper:2017, radecs paper:2025 | ||
| RT PolarFire | FPGA/SoC | susceptible | 1.3 (author) | null | 64 | 1.0e+12 | 100 | 9.5e-13 | yes | radecs redw paper:2020, radecs redw paper:2021 |
| MR2A16A | Flash/NVM | susceptible | 4.3 (fit) | null | 200 | 1.0e+11 | yes | gsfc test report:2008, radecs paper:2007, radecs paper:2008 | ||
| LM2991 | Power | susceptible | 5.8 (author) | null | 250 | 2.0e+11 | yes | radecs paper:2003 | ||
| 7201T | Logic/Interface | susceptible | 6.1 (fit) | null | 63 | 25 | gsfc test report:1994, radecs paper:1995 | |||
| AD7664 | ADC/DAC | susceptible | 7.0 (author) | null | 1.8e+12 | 5.0e-13 | gsfc test report:2003, redw paper:2003 | |||
| ADXL354 | Sensor/Imager | susceptible | 9.7 (author) | null | 200 | 1.1e+10 | yes | gsfc test report:2018, gsfc test report:2019 | ||
| 4Mb SRAM (256k x16 / 512k x8) | SRAM | susceptible | 10.0 (author) | null | 200 | redw paper:2019 | ||||
| SiT8003 | Sensor/Imager | susceptible | 12.0 (author) | null | 1000 | redw paper:2015 | ||||
| XC7Z020-1CLG484C | FPGA/SoC | susceptible | 16.0 (author) | null | 105 | 1.1e+11 | yes | redw paper:2015 | ||
| FM20L08 | Flash/NVM | susceptible | 19.9 (fit) | null | 200 | 1.0e+11 | yes | radecs paper:2007, radecs paper:2008 | ||
| Texas Instruments 45 nm bulk CMOS SRAM t | SRAM | susceptible | 30.0 (author) | null | 198 | gsfc test report:2010 | ||||
| K9F8G08U0M | Flash/NVM | susceptible | 35.0 (author) | null | radecs paper:2011, radecs paper:2012 | |||||
| M67204EV-50 | Logic/Interface | susceptible | 37.1 (author) | null | 63 | 25 | gsfc test report:1996 | |||
| TC58F401F-10 | Flash/NVM | susceptible | 37.5 (author) | null | 149 | redw paper:2003 | ||||
| R1LV1616RBG-7SI | DRAM/SDRAM | susceptible | 53.9 (author) | null | 230 | 1.0e-13 | redw paper:2008, redw paper:2013 | |||
| TMS320C25 | MCU/Processor | susceptible | 80.0 (author) | null | 250 | radecs paper:1996 | ||||
| R3000A | MCU/Processor | susceptible | 26.9 (author) | unclear | redw paper:1992 | |||||
| NAND01GW3B2ANGE | Flash/NVM | susceptible | 55.0 (author) | unclear | 200 | gsfc test report:2006, redw paper:2007 | ||||
| Lyontek SRAM (unspecified P/N) | SRAM | no threshold data | SEL | 200 | radecs paper:2020, radecs paper:2021 | |||||
| R3000 | MCU/Processor | no threshold data | SEL | 200 | redw paper:1992 | |||||
| AMD K7 | MCU/Processor | no threshold data | null | 190 | 3.3e+11 | yes | gsfc test report:2000, gsfc test report:2001 | |||
| CMV12000 | Sensor/Imager | no threshold data | null | 105 | 4.0e+11 | yes | redw paper:2017 | |||
| Cypress SRAM (on test daughter board) | SRAM | no threshold data | null | 230 | 6.5e+12 | yes | radecs paper:2018 | |||
| MX30LF4G18AC-TI | Flash/NVM | no threshold data | null | 200 | 1.0e+11 | 85 | yes | radecs redw paper:2021 | ||
| MX68GL1G0G | Flash/NVM | no threshold data | null | 200 | 2.0e+11 | 85 | yes | radecs redw paper:2021 | ||
| PC28F00AM29EW | Flash/NVM | no threshold data | null | 200 | 2.0e+11 | 85 | yes | radecs redw paper:2021 | ||
| TC58NVG2S0HTAI0 | Flash/NVM | no threshold data | null | 200 | 1.0e+11 | 85 | yes | radecs redw paper:2021 |
How it works
Why a proton test sees only part of the latchup risk
1
Heavy ions give the onset
The heavy-ion curve gives the LET where latchup starts, here the LET at which the fitted cross section crosses 10⁻⁸ cm² or the threshold the authors state, and the saturated cross section.
2
Protons act through recoils
A proton can trigger latchup only through a nuclear recoil that deposits more than the onset inside the sensitive volume, so the outcome depends on the proton energy, the fluence and how deep that volume sits.
3
A null covers part of the risk
A clean proton test discounts low-onset, large-cross-section, shallow-volume behaviors and generally says little about the rest. The Proton Proxy Risk Tool works out what a given test covers.
How this connects to the tool Proton Proxy Risk Tool
The Proton Proxy Risk Tool turns the physics in these cases into numbers for your test: the recoil LET-equivalent spectrum your fluence and energy delivered, which Weibull behaviors that spectrum could and could not have found at your sensitive-volume depth, the worst-case heavy-ion rate that remains, and, in the Bayesian block, how much a null result should move your belief about the part, starting from the EHSP susceptibility priors and cross-checked against the corpus numbers on this page. Validation of the engine against these same parts is reported in the help page.
Method pairing, classification, scripts
Parts were paired by normalized part number across the corpus (aliases merged); proton outcomes were classified from the
extracted result statements at sentence level (negated latchup statements and stated upper limits count as nulls) with 14 hand adjudications recorded in
the build script; one mis-digitized heavy-ion curve (an energy-axis proton plot) was replaced by the author-reported Weibull. Scripts:
build_proton_hi_pairs.py, build_public_aggregates.py, proxy_validation_harness.js.
References
R. Ladbury and J.-M. Lauenstein, "Use of Proton SEE Data as a Proxy for Bounding Heavy-Ion SEE Susceptibility," IEEE TNS 64(1), 2016.
R. Ladbury, J.-M. Lauenstein and K. P. Hayes, "Use of Proton SEE Data as a Proxy for Bounding Heavy-Ion SEE Susceptibility," IEEE TNS 62(6), 2015.
D. M. Hiemstra and E. W. Blackmore, "LET Spectra of Proton Energy Levels From 50 to 500 MeV and Their Effect on SEE Rate Prediction," IEEE TNS 50(6), 2003.
Per-case sources are listed on each card.
For your own part and test, the Proton Proxy Risk Tool works out which heavy-ion behaviors a proton null rules out and what risk remains.Open the Proton Proxy Risk Tool