CREME96 has been the standard tool for SEE rate prediction for decades. It’s embedded in institutional processes, referenced in standards, and used as the baseline in countless flight heritage arguments. But the space radiation environment models, the device technologies, and our understanding of SEE mechanisms have all evolved significantly since 1996.
Some specific concerns worth discussing:
Environment models. CREME96’s GCR model is based on the Nymmik model with specific solar modulation assumptions. More recent models (e.g., ISO 15390, CREME-MC’s updated environments, or DLR’s models) incorporate additional solar cycle data and may give different flux predictions, particularly at high LETs and during solar minimum.
The RPP assumption. CREME96’s rate calculation uses the rectangular parallelepiped (RPP) sensitive volume model. For modern devices with complex 3D geometries, FinFET structures, and multiple sensitive nodes, the single-RPP assumption can be a poor approximation. MRED and other Monte Carlo tools handle geometry more realistically but require much more setup effort and information about the device structure.
Proton direct ionization. CREME96 wasn’t designed to handle proton direct ionization SEE, which is increasingly relevant for advanced technology nodes (sub-28nm). The nuclear reaction framework doesn’t capture the direct ionization contribution for devices with very low LET thresholds.
The counterargument: CREME96 is well-understood, produces conservative results for most cases, and has decades of flight correlation. “Conservative and understood” beats “more accurate but uncertain” in many mission assurance contexts. Switching tools also means losing direct comparability with heritage analyses.
Where do you come down on this? Is CREME96 still your go-to, or have you moved to other tools? Has anyone done direct comparisons between CREME96 and CREME-MC (or SPACE RADIATION, OMERE, etc.) for the same device and orbit?