Hacker News Reader

Comment by mikeash | Hacker News Reader

Don't they use the latest and greatest tech that has been made spaceworthy? It's not just a matter of whether it's been tested a lot. Off-the-shelf electronics just won't work for these things, because they're not sufficiently hardened against radiation. From what I can dig up, it looks like Cassini uses 20-30MHz 16-bit CPUs, which is really not bad at all for a probe whose design started around the early 1990s.

View on HN

NASA does things this way but SpaceX explicitly does not. Rather than using 1 space hardened part, SpaceX usually uses 3 non-space hardened parts (processor, memory, bus, etc) that all do exactly the same calculations simultaneously and then compare every single answer and ignore it when one out of 3 parts disagree. This allows them to be both cheaper and much less power hungry because they operate at least 2 process nodes smaller than NASA hardware. I've seen data that says that while each individual processor will experience space related errors quite often, the whole 3 part voting system is at least as accurate as the way NASA does it.

It's not a SpaceX-only scheme, nor is it novel technology. It's got a long history in fighter jets, as well as modern passenger planes for example (Boeing 777).

Triple Mode Redundancy: https://en.wikipedia.org/wiki/Triple_modular_redundancy

Also, the longest life span SpaceX has ever really needed is maybe the Dragon, which is a month-long mission. Cassini has been in space for 20 years. Very different design requirements.

There's also a difference between a space probe, built once and having a mission that spans years or decades, and a rocket that gets built more often and operates only in Earth orbit (especially the first stage has to worry far less about radiation flipping bits).

This isn't solely a difference between NASA and SpaceX, but also a difference in mission requirements.

NASA must be well aware of the triple redundancy logic you're describing since they themselves used it on the Saturn V in the 60's.

https://en.wikipedia.org/wiki/Saturn_Launch_Vehicle_Digital_...

> Don't they use the latest and greatest tech that has been made spaceworthy?

Sure, and by the end of that process (testing, construction, launch, coast, arrival), it winds up being "extremely well tested, older technology".

But that implies that being "extremely well tested, older technology" is an afterthought, not a conscious choice. In other words, if brand-new hardware were magically radiation-hardened, they wouldn't disqualify that hardware for consideration by dint of being new. Especially since newer hardware tends to be more energy-efficient than older hardware, which is a very important consideration for these missions.

And no, this isn't implying that they're going to launch a mission without any testing at all. But spacecraft are specced and designed years before they actually launch; if you want to launch a craft in 2030, you need to pick your hardware in 2025. The intervening years are plenty of time to gain assurance in the reliability of the hardware, regardless of whether that hardware was released in 2025 or 1995.

No, in 2025 you're going to be picking from the hardware that's been already approved for space flight at that point. Reduces likelihood of discovering several years into building the probe that you're going to have to redesign everything to work with a different CPU.

Was talking with a JPL engineer at a party a few weeks back, he happened to mention that current state-of-the-art is the RAD750, essentially a radiation-hardened powerpc G3.

The RAD750 might be flight proven but I wouldn't call it the 'current state-of-the-art' in radiation hardened computing. Its technology is over a decade old.

Reply on news.ycombinator.com