3 ms·

There is virtually no functional difference between a flight qualified processor and a ground development board. The cost difference is due to the overall relia

There is virtually no functional difference between a flight qualified processor and a ground development board. The cost difference is due to the overall reliability requirement for flight parts, for one. If you lose your spacecraft processor you lose your spacecraft, which probably cost you a couple of hundred million dollars. So the yield rate for spacecraft CPUs is quite low, like single percentage low. The chips that fail the qual inspection become ground development units. In addition the manufacture, storage, and movement of each chip is very carefully managed and documented, which adds a surprising amount to the cost. This allows you to determine root causes in case of a failure, which then lets you understand whether other spacecraft using similar parts are in danger. Radiation tolerance is an important part of this, both in terms of lifetime exposure tolerance and in terms of single event upset susceptibility.

exikyut 8y ago
Wow, thanks for the TIL. Suddenly I understand the fuss about the early processors (eg, the Forth ones) that went into spacecraft in the 70s and 80s. It wasn't just the architecture that was good, it was the electrical/physical design that was noteworthy (which, to excuse the terrible pun, the CPUs are still writing home about :D). Heh, I wonder how much the ground development boards cost? Probably way too much to find one affordably on eBay, say... Also - how do you figure out if a processor is going to last for ages? How do you test it?! Hmm, I get the impression some spacecraft probably have full failover capability as well. And now I wonder what systems use CPUs in lockstep _and_ have full independent failover...
- GlenTheMachine 8y ago
  A ground development board would be on the order of $50,000, give or take. You'll also need a chassis and power supply of course. “Also - how do you figure out if a processor is going to last for ages?” Really big semiconductor junctions help. Low heat production also helps. And then there's this thing called “heritage”. If it worked on the last fifteen missions, odds are decent it'll work on the sixteenth. Many spacecraft have fully redundant processors boards. Whether that helps you depends on whether you think chip failures are statistically uncorrelated. Which they are, sometimes. But not if there is a design flaw or manufacturing defect.

exikyut 8y ago

Wow, thanks for the TIL. Suddenly I understand the fuss about the early processors (eg, the Forth ones) that went into spacecraft in the 70s and 80s. It wasn't just the architecture that was good, it was the electrical/physical design that was noteworthy (which, to excuse the terrible pun, the CPUs are still writing home about :D). Heh, I wonder how much the ground development boards cost? Probably way too much to find one affordably on eBay, say... Also - how do you figure out if a processor is going to last for ages? How do you test it?! Hmm, I get the impression some spacecraft probably have full failover capability as well. And now I wonder what systems use CPUs in lockstep _and_ have full independent failover...

GlenTheMachine 8y ago
A ground development board would be on the order of $50,000, give or take. You'll also need a chassis and power supply of course. “Also - how do you figure out if a processor is going to last for ages?” Really big semiconductor junctions help. Low heat production also helps. And then there's this thing called “heritage”. If it worked on the last fifteen missions, odds are decent it'll work on the sixteenth. Many spacecraft have fully redundant processors boards. Whether that helps you depends on whether you think chip failures are statistically uncorrelated. Which they are, sometimes. But not if there is a design flaw or manufacturing defect.