9
$\begingroup$

Is it feasible to design a flight computer for a small CubeSat spacecraft destined for lunar orbit that is controlled exclusively by one or more non-radiation-hardened processors? What are the main design considerations to make such an architecture radiation tolerant, without using special radiation hardened components?

What would be the expected risk of single-event upsets and total ionizing dose effects over a one-year time span?

What type of redundancy or voting scheme, if any, would be recommended for the processors?

$\endgroup$
  • $\begingroup$ How much computing power do you need and how much electrical power can you use? Old processors are COTS. The 65C02 (Apple II, Rockwell System65, etc) is build with a process using much much larger features than today's small processors. $\endgroup$ – C. Towne Springer Feb 17 '15 at 19:25
  • $\begingroup$ Only low processing power is needed, and low electrical power is available, since it is a CubeSat. How would the 65C02 perform in such an environment? $\endgroup$ – Cosmo Harrigan Feb 18 '15 at 1:05
  • $\begingroup$ Yes, my company has. Not leaving as an answer because that feels inappropriate, but it is an answer to the feasibility part of the question. You'll need to elaborate on the SEU and TID part of your question. Are you assuming quick injection into lunar orbit followed by a one-year mission? How many passes through the van Allen belts? Risk of SEU is high, but radiation tolerant architectures can recover; is that acceptable? $\endgroup$ – Adam Wuerl Feb 18 '15 at 3:36
  • $\begingroup$ Suppose the spacecraft would be released after the van Allen belts have already passed; the key part of the question is: what are the main design considerations to make an architecture radiation tolerant, without using radiation hardened electronics? $\endgroup$ – Cosmo Harrigan Feb 18 '15 at 5:08
  • 1
    $\begingroup$ @CosmoHarrigan Interesting question, I have been thinking a lot about it recently. I hope you do not mind a bounty for this one. $\endgroup$ – s-m-e Feb 21 '15 at 14:39
10
+100
$\begingroup$

Yes, it’s probably possible, but to determine specific feasible approaches more information is needed about your mission. Key questions:

  1. What is the transfer and operational orbit? This information is needed to make analytical predictions about the probability of single-event-effects (SEE) and the total ionizing dose (TID) as a function of mission time. Good overview of radiation from JHU.

  2. What shielding, if any, does the larger spacecraft provide? Shielding knocks down the derived environments to those the processor and other susceptible electronics will actually see.

  3. What type of computer is required? Key requirements include operations per second, memory, storage, etc. Different types components and technologies respond to radiation environments differently. Your performance requirements may rule out entire types of technology.

  4. When does your computer need to operate? Is it on during launch, transit through the South Atlantic Anomaly, Van Allen Belts, or other localized areas of high radiation environments? It’s probably an easier mission if it’s only turned on once it reaches Lunar orbit.

  5. Probably others I’m forgetting that I’ll add based on comments.


Now that we've enumerated some complexities, let's blithely brush that away and talk about a potential architecture. You specifically want a processor that's not radiation hardened, but there are a variety of other radiation tolerance approaches you can use.

  1. More shielding. This time around the computer or processor. Aluminum is a good, cheap, and lightweight shield material. Tantalum is another high-performance shield material.

  2. De-rating. Use components below their specified values. This makes them more radiation tolerate. MIL-STD-975M has de-rating advice.

  3. Voting. Have multiple copies of your most susceptible components (probably processors or memory) that vote to determine truth. Voted EEPROMS can be a good way to store a golden master of key information that can be restored if error-detecting memory finds a flipped bit that it cannot repair.

  4. Watchdog timers. Latch-up is a radiation failure mode that can be recovered from by power-cycling. Have a robust analog watch dog timer that isn't radiation susceptible that must be pet periodically by the processor. If the processor hangs the timer resets the power to clear any latches. Of course this requires your mission be robust to having the computer go down for a little bit as things reboot.

  5. If you can't have the computer go down you may want a redundant processor running in a shadow mode and automatic fail-over to the secondary computer.

  6. You should test the snot out of all of this.

$\endgroup$
  • $\begingroup$ Could you elaborate on algorithms and potential designs for voting? $\endgroup$ – Cosmo Harrigan Feb 23 '15 at 2:04
  • 2
    $\begingroup$ What you’re asking is the topic of graduate-level courses in aerospace and computer engineering and could fill an entire book or set of academic papers. $\endgroup$ – Adam Wuerl Feb 23 '15 at 2:18
  • 1
    $\begingroup$ @AdamWuerl True enough. I guess there already is a bunch of relevant literature out there. Could you try to point to some? $\endgroup$ – s-m-e Feb 24 '15 at 19:10
  • $\begingroup$ @AdamWuerl Thank you for the thoughtful and informative answer. If you have some suggested references specific to relevant voting algorithms that you would be willing to expand upon in the answer, that would be helpful. Specifically, there exist consensus protocols to handle unreliability in the field of distributed computing research, and I am curious which parts of that research area are considered most applicable to this problem domain. $\endgroup$ – Cosmo Harrigan Feb 25 '15 at 8:32
  • $\begingroup$ I can't comment on the applicability of distributed computing. That's outside my expertise. I’m also not an expert in computer architectures. It was only touched on briefly in my graduate program. Professionally, my knowledge isn't much deeper than the above, which is what I've gleaned working with people who have implemented such schemes. $\endgroup$ – Adam Wuerl Feb 26 '15 at 5:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.