5 ms·

This article does a good job outlining why these systems can't just be replaced with a Raspberry Pi or an Arduino: Durability. At one point I was interested in

This article does a good job outlining why these systems can't just be replaced with a Raspberry Pi or an Arduino: Durability. At one point I was interested in replacing the ECU in my old '88 Suburban. The ECU is probably the only "real" bit of electronics in the whole vehicle. And yet after almost 35 years it's still doing fine. The Arduino based ECU I was looking into, as it turns out, is a great hobby project, but not so good for reliability due to the harsh conditions. Still, I think electronics in modern vehicles are completely over the top, but then again... I drive a 30+ year old Suburban. I'm prone to thinking that way :D

dylan604 5y ago
The space shuttle used a hardened 386 (or something similar) that cost a bazillion dollars. The drone currently on Mars used commidity off the shelf chips (snapdragon IIRC). The harsh conditions thing is not something to laugh at, but I also think we might have over-engineered somethings a bit.
- Retric 5y ago
  The surface of Mars being further from the sun and inside an atmosphere has vastly fewer issues with radiation.
  - geerlingguy 5y ago
    But more issues with temperature swings—ideally radiation wouldn't be the main concern here in Earth's atmosphere.
    - Retric 5y ago
      Rovers regulate internal temperatures to keep electronics happy. Ex: https://mars.nasa.gov/mer/mission/rover/temperature/ https://mars.nasa.gov/mer/mission/rover/temperature/
  - Symmetry 5y ago
    In both cases you're well protected against the solar wind by distance and atmosphere in the case of the Mars rover and by the Van Allen belts in the case of the Space Shuttle. But both still have to worry about cosmic rays. Mars's thin atmosphere helps a bit with that but not nearly as much as you'd like. EDIT: I suspect a lot of what lets you use Snapdragons in the rover is the idea of making error resistant systems rather than error resistent devices. The same thing that let us replace a single $10,000 hard driver with ten $100 hard drives in a RAID with the same performance and reliability despite the individual drives being much worse.
    - Retric 5y ago
      That’s not really accurate the sun causes real issues in LEO. See chapter 2 starting on page 9 here: https://commons.erau.edu/cgi/viewcontent.cgi?article=1101&context=edt https://commons.erau.edu/cgi/viewcontent.cgi?article=1101&co...
      - dylan604 5y ago
        As an anecdotal bit of support, I seem to recall recently SpacX loosing a number of satellites in their LEO constellation because of sun activity
        mschuster91 5y ago
        Yes, but that was not due to EM activity, it was because the solar activity made the atmosphere go up higher than usual which meant that the atmospheric drag was too strong for the engines to overcome.
  - addingnumbers 5y ago
    We can't teleport something from one atmosphere to another, though. It's a long trip.
    - Retric 5y ago
      That’s a risk, but the chip is off for the trip which makes a significant difference. Fail safe is much easier if you can run diagnostics without needing the system to be doing it’s job.
  - temp0826 5y ago
    Biggest difference is the lack of risk to human life, tbh
  - dclowd9901 5y ago
    Also, human lives aren’t at stake if things go awry.
- Isinlor 5y ago
  SpaceX is also using commodity hardware, but they duplicate it 3x with majority vote for error correction.
  - froh 5y ago
    do you happen to have pointers evidencing this handy? IS this also true for Dragon?
    - TrainedMonkey 5y ago
      It is, see https://stackoverflow.blog/2021/12/27/dont-push-that-button-exploring-the-software-that-flies-spacex-starships/ https://stackoverflow.blog/2021/12/27/dont-push-that-button-...
      - froh 5y ago
        Thanks! I still hope to learn how they deal with infamous "bi directional requirements Tracing" and similar bureaucracies in the name of dependability. How do they evidence engineering scrutiny, how do the embed it into their processes in such a way that they _evidence_ scrutiny and deliberate engineering _without_ massive paper safety tigers, _without_ tons of slow bureaucracy?
    - xyzzy21 5y ago
      Some of it is classified but you can often find info on certain open sites at NASA or SNL.
- tablespoon 5y ago
  > The space shuttle used a hardened 386 (or something similar) that cost a bazillion dollars. Their main computers were a custom variant of the IBM 360 mainframe. They also duplicated it 5x: 4x with majority vote for error correction, plus a backup running separately-developed software. https://en.wikipedia.org/wiki/IBM_System/4_Pi https://en.wikipedia.org/wiki/IBM_System/4_Pi > The drone currently on Mars used commidity off the shelf chips (snapdragon IIRC). The harsh conditions thing is not something to laugh at, but I also think we might have over-engineered somethings a bit. It should be noted that drone is 1) an engineering experiment and 2) no one will die if it malfunctions, so they can afford to be a bit sloppier.
technothrasher 5y ago
I designed a drop in replacement ignition ECU for Ferrari 308 cars from the 1980's (the Magnum PI model), as after almost 40 years a lot of them are now failing. When I started the project, I had no idea how miserable an environment the car actually was. It wasn't hard to get a prototype that basically worked, but I had to go through quite a few revisions before I got something that could withstand the torture of everyday use (and things like electrically dirty jump starters that would completely fry my units until I put a protective power supply in it against +/-100VDC surges.)
calvinmorrison 5y ago
My daily is also an 88. It's running very standard Bosch of that era. At some point I took my 93 (same model) and swapped it with newer fuel management, ecu, etc out of a 2005 Saab. Same valve cover allows you to the stock coils and setup, a timing sensor solves that, and boom, good to go. Well sorta, turns out the 88 is far more reliable and consistent - except for a few hard to find parts... namely the distributor is a total piece of crap after that many years... and it's not very efficient. I get 50% more performance and 50% more gas milage on the uprated one. Difference in reliability is probably... once every 25k versus once every 50k though. Those bosch systems were all over volvo, saab, porsche, VW, Audi, etc and are dead reliable but not very hackable.

dylan604 5y ago

The space shuttle used a hardened 386 (or something similar) that cost a bazillion dollars. The drone currently on Mars used commidity off the shelf chips (snapdragon IIRC). The harsh conditions thing is not something to laugh at, but I also think we might have over-engineered somethings a bit.

Retric 5y ago
The surface of Mars being further from the sun and inside an atmosphere has vastly fewer issues with radiation.
- geerlingguy 5y ago
  But more issues with temperature swings—ideally radiation wouldn't be the main concern here in Earth's atmosphere.
  - Retric 5y ago
    Rovers regulate internal temperatures to keep electronics happy. Ex: https://mars.nasa.gov/mer/mission/rover/temperature/ https://mars.nasa.gov/mer/mission/rover/temperature/
- Symmetry 5y ago
  In both cases you're well protected against the solar wind by distance and atmosphere in the case of the Mars rover and by the Van Allen belts in the case of the Space Shuttle. But both still have to worry about cosmic rays. Mars's thin atmosphere helps a bit with that but not nearly as much as you'd like. EDIT: I suspect a lot of what lets you use Snapdragons in the rover is the idea of making error resistant systems rather than error resistent devices. The same thing that let us replace a single $10,000 hard driver with ten $100 hard drives in a RAID with the same performance and reliability despite the individual drives being much worse.
  - Retric 5y ago
    That’s not really accurate the sun causes real issues in LEO. See chapter 2 starting on page 9 here: https://commons.erau.edu/cgi/viewcontent.cgi?article=1101&context=edt https://commons.erau.edu/cgi/viewcontent.cgi?article=1101&co...
    - dylan604 5y ago
      As an anecdotal bit of support, I seem to recall recently SpacX loosing a number of satellites in their LEO constellation because of sun activity
      - mschuster91 5y ago
        Yes, but that was not due to EM activity, it was because the solar activity made the atmosphere go up higher than usual which meant that the atmospheric drag was too strong for the engines to overcome.
- addingnumbers 5y ago
  We can't teleport something from one atmosphere to another, though. It's a long trip.
  - Retric 5y ago
    That’s a risk, but the chip is off for the trip which makes a significant difference. Fail safe is much easier if you can run diagnostics without needing the system to be doing it’s job.
- temp0826 5y ago
  Biggest difference is the lack of risk to human life, tbh
- dclowd9901 5y ago
  Also, human lives aren’t at stake if things go awry.
Isinlor 5y ago
SpaceX is also using commodity hardware, but they duplicate it 3x with majority vote for error correction.
- froh 5y ago
  do you happen to have pointers evidencing this handy? IS this also true for Dragon?
  - TrainedMonkey 5y ago
    It is, see https://stackoverflow.blog/2021/12/27/dont-push-that-button-exploring-the-software-that-flies-spacex-starships/ https://stackoverflow.blog/2021/12/27/dont-push-that-button-...
    - froh 5y ago
      Thanks! I still hope to learn how they deal with infamous "bi directional requirements Tracing" and similar bureaucracies in the name of dependability. How do they evidence engineering scrutiny, how do the embed it into their processes in such a way that they _evidence_ scrutiny and deliberate engineering _without_ massive paper safety tigers, _without_ tons of slow bureaucracy?
  - xyzzy21 5y ago
    Some of it is classified but you can often find info on certain open sites at NASA or SNL.
tablespoon 5y ago
> The space shuttle used a hardened 386 (or something similar) that cost a bazillion dollars. Their main computers were a custom variant of the IBM 360 mainframe. They also duplicated it 5x: 4x with majority vote for error correction, plus a backup running separately-developed software. https://en.wikipedia.org/wiki/IBM_System/4_Pi https://en.wikipedia.org/wiki/IBM_System/4_Pi > The drone currently on Mars used commidity off the shelf chips (snapdragon IIRC). The harsh conditions thing is not something to laugh at, but I also think we might have over-engineered somethings a bit. It should be noted that drone is 1) an engineering experiment and 2) no one will die if it malfunctions, so they can afford to be a bit sloppier.

technothrasher 5y ago

I designed a drop in replacement ignition ECU for Ferrari 308 cars from the 1980's (the Magnum PI model), as after almost 40 years a lot of them are now failing. When I started the project, I had no idea how miserable an environment the car actually was. It wasn't hard to get a prototype that basically worked, but I had to go through quite a few revisions before I got something that could withstand the torture of everyday use (and things like electrically dirty jump starters that would completely fry my units until I put a protective power supply in it against +/-100VDC surges.)

calvinmorrison 5y ago

My daily is also an 88. It's running very standard Bosch of that era. At some point I took my 93 (same model) and swapped it with newer fuel management, ecu, etc out of a 2005 Saab. Same valve cover allows you to the stock coils and setup, a timing sensor solves that, and boom, good to go. Well sorta, turns out the 88 is far more reliable and consistent - except for a few hard to find parts... namely the distributor is a total piece of crap after that many years... and it's not very efficient. I get 50% more performance and 50% more gas milage on the uprated one. Difference in reliability is probably... once every 25k versus once every 50k though. Those bosch systems were all over volvo, saab, porsche, VW, Audi, etc and are dead reliable but not very hackable.