On-orbit radiation failures and published data?

I was tagged in a LinkedIn post recently asking about typical failure rates for "radiation-induced faults, sources of data, impact of altitude and Error Correction Codes, etc." I get asked this a lot, so here is my usual answer for anyone wrestling with this question:

1) There are no easy rules of thumb:

Radiation's impact varies based on the component, application, and orbit. You can't generalize without having test data for a component or system.

2) Sources of on-orbit data for failure rates:

Acquiring good data for the causes of satellite failures is challenging. It’s not always easy to characterize how something failed, and most companies don’t share this data. Some valuable references from NASA [1] discuss NASA and university cube-to-smallsat failures. Some older references discuss high reliability MEO/GEO spacecraft [2,3] and some organizations have data available for a fee [4].

NASA plot showing a sample of cube and smallsat failure rates [1]

3) The most likely cause of an early, radiation induced mission failure:

Devices resetting too frequently, if your radio or onboard computer resets 10 times a day, a satellite is pretty much unusable.

To ensure you don’t have the early radiation induced issues, you don’t need rad hard parts or complex testing. The above issue can be ruled out with one relatively easily board level test (similar to other environmental tests). It won’t guarantee high reliability over 10 years, but quick, cheap and simple tests can ensure a demonstration system won’t freak out the second it reaches orbit.

4) Estimating on orbit event rates:

Published test data for COTS parts, when paired with simulation tools like OMERE or SPENVIS, can be revealing. Recent test data for the Xilinx Zynq-7000 [5] estimates 3 resets/day at 800 km orbits. In this scenario, a key factor is the absence of built-in Error Correction Code, which means bit flips in critical memory can cause the device to reset.

Analysis of test data showing Zynq-7000 on-orbit event rates [5]

5) How do event rates change with orbit?

The Petersen figure of merit method [6] is an approximation technique used to correlate ground test data with on-orbit event rates. It’s a really convenient way to plot visual trends and show how bit flips and reset rates change with orbit. It's not 100% accurate but serves as a useful rule of thumb.

Petersen's on-orbit rate coefficients [6]

Not a perfect answer, but it’s the best I have. Would love to get other people's opinions :)

References:

[1] NASA “Small-Satellite Mission Failure Rates” 2016

[2] NASA “Spacecraft System Failures and Anomalies Attributed to the Natural Space Environment” 1996

[3] HC Koons et al, “The Impact of the Space Environment on Space Systems” Journal: Proceedings of the 6th Spacecraft Charging Conference 1998

[4] Seradata “Satellite Anomalies & Failures”

[5] "Estimating the Error Rates of Xilinx Zynq-7000 APSoCs in Space Radiation Environments"

[6] E. L Petersen, "The SEU Figure of Merit and Proton Upset Rate Calculations," IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 45, NO. 6, DECEMBER 1998