6 ms·

Maybe after the shift from space-grade parts to commodity parts in (some) satellites, this will mean the same for parts of future mars missions. A Snapdragon 80

by danhor 5y ago

Maybe after the shift from space-grade parts to commodity parts in (some) satellites, this will mean the same for parts of future mars missions. A Snapdragon 801 as used here is certainly much cheaper, easier to work with and powerful than anything space-grade.

DrNuke 5y ago
It really depends on the space environmental work conditions, though… that’s the fundamental trigger for space grade quality.
nisegami 5y ago
From my limited knowledge of electronics-in-space, isn't ECC memory a hard requirement to correct for bit flips due to solar radiation/particles? But DDR5 having ECC be standard across the board may make this a non-issue.
- Cthulhu_ 5y ago
  Kinda, yes, but at the same time, servers here on earth have moved away from (expensive) ECC in favor of resilient software, disconnected services (job queue based) and redundancy. Cheaper to have a dozen computers that sometimes break than three with lower risk (but they still break).
  - zitterbewegung 5y ago
    For consumers no but for things like large simulations or even deep learning you want to have ECC memory. Intel only enables it on Xeons but AMD doesn't officially support it on their consumer chips but sometimes it works.
    - my123 5y ago
      It's locked down on AMD customer APUs, only not forcefully disabled on AMD customer CPUs that do not have integrated graphics. (Ryzen PRO parts have it officially supported)
solarkraft 5y ago
It’s also such great marketing. I wonder why the Zigbee alliance doesn’t communicate more that it’s used for communication between the rover and the helicopter.
- mschuster91 5y ago
  Probably because they'd get ridiculed to beyond the solar system that they can make a Mars rover fly, but it's a nightmare to get two Zigbee products from different vendors to talk to each other.
_joel 5y ago
There's also some RAD hardened components to check for bit-flips and restart the hardware that quickly it can do it mid-flight. Here's the full paper. https://trs.jpl.nasa.gov/bitstream/handle/2014/46229/CL%2317-6243.pdf https://trs.jpl.nasa.gov/bitstream/handle/2014/46229/CL%2317...
- flavius29663 5y ago
  That is so smart. I know we're talking about NASA, but I am still in awe. I read a bit of the paper, it's actually even more redundant than what you said: - snapdragon is only used for higher level functions of the flight: The SnapdragonTM processor has a 2.26 GHz Quad-core SnapdragonTM 801 processor with 2 GB Random Access Memory (RAM), 32 GB Flash memory, a Universal Asynchronous Receiver Transmitter (UART), a Serial Peripheral Interface (SPI), General Purpose Input/Ouput (GPIO), a 4000 pixel color camera, and a Video Graphics Array (VGA) black-and-white camera. This processor implements visual navigation via a velocity estimate derived from features tracked in the VGA camera, filter propagation for use in flight control, data management, command processing, telemetry generation, and radio communication. The SnapdragonTM processor is connected to two - for flight control they use 2 (redundant) automotive chips from Texas Instruments: TMS570LC43x high-reliability automotive processor operating at 300 MHz, with 512 K RAM, 4 MB flash memory, UART, SPI, GPIO. - these are all controlled by a radiation hardened board, the 3rd level of redundancy
  - _joel 5y ago
    They also used a $129.99 altimeter from Garmin (Lidar-Lite-V3), no fancy fancy. https://www.sparkfun.com/products/14032 https://www.sparkfun.com/products/14032 edit: https://www.garmin.com/en-US/blog/general/garmin-on-mars/ https://www.garmin.com/en-US/blog/general/garmin-on-mars/
- ryandrake 5y ago
  I wonder how much hardware marketed as "space grade" (or whatever the terminology is, I'm just a layman) is actually as hardened using these techniques and redundant as you'd expect. Or are they simply certified because the manufacturer's process went through some checkboxes. In the aviation world, for some components, the difference between aviation grade and non- is 1. cost and 2. a Certificate of Conformance from the manufacturer that says "this thing is provably aviation grade, trust us bro."
enkid 5y ago
I believe it is much cheaper to put something in LEO then it is to get it to Mars, so I don't know if the economics of Mars Missions have fundamentally shifted to cheaper probes the way they have for satellites.
- XorNot 5y ago
  I suspect we're limited more by communications capability then anything else: pretty much a hand-held radio will reach up to LEO if someone's listening (short-wave definitely does and that's off-the-shelf at consumer prices pretty much). Whereas Mars requires the Deep Space Network to reasonably be able to maintain communications - so your very cheap probes still need super-highgrade radios to be useful (the landers for example still generally pack an antenna that can reach all the way back to Earth even when they use uplinks - the helicopter is relaying via the lander). What we're missing on Mars is reliable communications for probes: there's no common backhaul to Earth that you can just hook into. But I'd be willing to bet that's coming: if Starlink works on Earth, there's no real reason it shouldn't work on Mars. At which point you can go with the "swarm of cheap hardware" idea because it can hand off the need for the power, hardware and logistics and backing to communicate inter-planetary.
  - jvanderbot 5y ago
    > Whereas Mars requires the Deep Space Network to reasonably be able to maintain communications - so your very cheap probes still need super-highgrade radios to be useful That's not technically true. They maintain fancy radios that can reach Earth as a backup. Data transfer is done via the Mars Relay network. https://mars.nasa.gov/news/8861/the-mars-relay-network-connects-us-to-nasas-martian-explorers/ https://mars.nasa.gov/news/8861/the-mars-relay-network-conne... Ingenuity uses this as well. > if Starlink works on Earth, there's no real reason it shouldn't work on Mars. Ground-to-low orbit comms relays have been in operations for half a century. And decades at Mars already. Starlink is for high throughput communications from ground to ground on the same planet with always-on connectivity and global coverage. Their sattelites are engineered cheap and replenished with cheap lift from SpaceX. Totally different problem calculus at Mars and doesn't address "backhaul" to Earth. You'd need longer-lived high throughput optical terminals or better, preferably at Areostationary orbits. Checkout work by Briedenthal and Edwards on the matter.
darkwater 5y ago
I'm not a space expert at all but my wild guess is that doing small, careful tests for consumer grade hardware (and software) to be used on Mars is a way to lower costs of a possible future settlement, by leveraging economies of scale here on Earth.
holoduke 5y ago
Anyone knows if the copter suffer already from issues and reboots already?
johnwalkr 5y ago
I actually just wrote the policy on this topic for a space project I work on. It really depends. It’s often not cheaper to use such commodity stuff unless you can tolerate a lot of risk. A lot of normal and automotive components are used in space, but it’s mostly simple stuff that’s relatively easy to qualify. A snapdragon cpu will basically never be fully qualified as a single component because you cannot control each state, let alone test it in each state (I am oversimplifying but bear with me). What you can do is apply overcurrent detection, external watchdog timers, redundancy and other techniques to mitigate the effects of radiation. Then, you can almost fully test that mitigation in theory but not really because it will cost you millions of dollars with huge technical risk and the part will probably be obsoleted or made by a different fab next time you need it. So, you choose older proven easily-hard stuff where you can, or for anything mission critical or related to safety. Better to just buy the $200k rad hard cpu and save your millions. Then, you are left with only bonus items such as this helicopter that simultaneously do not have many reliability requirements (because it’s not mission critical) and are not possible with traditional space grade technology due to performance requirements. “Easier to work with” is only true when you take risks too. Yeah, you can start developing right away using a devkit and Linux, but if you find your project needs reliability or safety requirements, you may find you are spending person-years of software engineering time if it’s even possible at all. You might be required for the system to be deterministic and avoid dynamic memory allocation, for example. Interestingly, students making cubesats don’t “know better” yet. And while the overall failure rate is probably bad on a per mission basis, it’s probably good on a per kg or per dollar basis. There’s a ton of cubesats and cubesat payloads that have raspberry pi’s inside for Leo missions.
- kanwisher 5y ago
  Yeah but part of why they used a high speed arm, is so they could do realtime AI fo determining where to land and what not. If you use a RAD hardened chip you are limited to what 400-500mhz chips from over a decade ago. Fine if your doing a rover that doesn't need to do realtime detection and avoidance of stuff.
pkaye 5y ago
However they still use a rad-hardened FPGA to run some of the critical aspects. This paper discusses a lot of the internals. https://rotorcraft.arc.nasa.gov/Publications/files/Balaram_AIAA2018_0023.pdf https://rotorcraft.arc.nasa.gov/Publications/files/Balaram_A...
jessriedel 5y ago
A bigger effect is that, by the time the next helicopter is ready to launch, SpaceX Starship is likely to drop the cost of delivering cargo to Mars by an order of magnitude or more. That will massively disrupt the mass vs. capability trade-offs we see in all Martian rovers and aircraft.

by danhor 5y ago

DrNuke 5y ago

It really depends on the space environmental work conditions, though… that’s the fundamental trigger for space grade quality.

nisegami 5y ago

From my limited knowledge of electronics-in-space, isn't ECC memory a hard requirement to correct for bit flips due to solar radiation/particles? But DDR5 having ECC be standard across the board may make this a non-issue.

Cthulhu_ 5y ago
Kinda, yes, but at the same time, servers here on earth have moved away from (expensive) ECC in favor of resilient software, disconnected services (job queue based) and redundancy. Cheaper to have a dozen computers that sometimes break than three with lower risk (but they still break).
- zitterbewegung 5y ago
  For consumers no but for things like large simulations or even deep learning you want to have ECC memory. Intel only enables it on Xeons but AMD doesn't officially support it on their consumer chips but sometimes it works.
  - my123 5y ago
    It's locked down on AMD customer APUs, only not forcefully disabled on AMD customer CPUs that do not have integrated graphics. (Ryzen PRO parts have it officially supported)

solarkraft 5y ago

It’s also such great marketing. I wonder why the Zigbee alliance doesn’t communicate more that it’s used for communication between the rover and the helicopter.

mschuster91 5y ago
Probably because they'd get ridiculed to beyond the solar system that they can make a Mars rover fly, but it's a nightmare to get two Zigbee products from different vendors to talk to each other.

_joel 5y ago

There's also some RAD hardened components to check for bit-flips and restart the hardware that quickly it can do it mid-flight. Here's the full paper. https://trs.jpl.nasa.gov/bitstream/handle/2014/46229/CL%2317-6243.pdf https://trs.jpl.nasa.gov/bitstream/handle/2014/46229/CL%2317...

flavius29663 5y ago
That is so smart. I know we're talking about NASA, but I am still in awe. I read a bit of the paper, it's actually even more redundant than what you said: - snapdragon is only used for higher level functions of the flight: The SnapdragonTM processor has a 2.26 GHz Quad-core SnapdragonTM 801 processor with 2 GB Random Access Memory (RAM), 32 GB Flash memory, a Universal Asynchronous Receiver Transmitter (UART), a Serial Peripheral Interface (SPI), General Purpose Input/Ouput (GPIO), a 4000 pixel color camera, and a Video Graphics Array (VGA) black-and-white camera. This processor implements visual navigation via a velocity estimate derived from features tracked in the VGA camera, filter propagation for use in flight control, data management, command processing, telemetry generation, and radio communication. The SnapdragonTM processor is connected to two - for flight control they use 2 (redundant) automotive chips from Texas Instruments: TMS570LC43x high-reliability automotive processor operating at 300 MHz, with 512 K RAM, 4 MB flash memory, UART, SPI, GPIO. - these are all controlled by a radiation hardened board, the 3rd level of redundancy
- _joel 5y ago
  They also used a $129.99 altimeter from Garmin (Lidar-Lite-V3), no fancy fancy. https://www.sparkfun.com/products/14032 https://www.sparkfun.com/products/14032 edit: https://www.garmin.com/en-US/blog/general/garmin-on-mars/ https://www.garmin.com/en-US/blog/general/garmin-on-mars/
ryandrake 5y ago
I wonder how much hardware marketed as "space grade" (or whatever the terminology is, I'm just a layman) is actually as hardened using these techniques and redundant as you'd expect. Or are they simply certified because the manufacturer's process went through some checkboxes. In the aviation world, for some components, the difference between aviation grade and non- is 1. cost and 2. a Certificate of Conformance from the manufacturer that says "this thing is provably aviation grade, trust us bro."

enkid 5y ago

I believe it is much cheaper to put something in LEO then it is to get it to Mars, so I don't know if the economics of Mars Missions have fundamentally shifted to cheaper probes the way they have for satellites.

XorNot 5y ago
I suspect we're limited more by communications capability then anything else: pretty much a hand-held radio will reach up to LEO if someone's listening (short-wave definitely does and that's off-the-shelf at consumer prices pretty much). Whereas Mars requires the Deep Space Network to reasonably be able to maintain communications - so your very cheap probes still need super-highgrade radios to be useful (the landers for example still generally pack an antenna that can reach all the way back to Earth even when they use uplinks - the helicopter is relaying via the lander). What we're missing on Mars is reliable communications for probes: there's no common backhaul to Earth that you can just hook into. But I'd be willing to bet that's coming: if Starlink works on Earth, there's no real reason it shouldn't work on Mars. At which point you can go with the "swarm of cheap hardware" idea because it can hand off the need for the power, hardware and logistics and backing to communicate inter-planetary.
- jvanderbot 5y ago
  > Whereas Mars requires the Deep Space Network to reasonably be able to maintain communications - so your very cheap probes still need super-highgrade radios to be useful That's not technically true. They maintain fancy radios that can reach Earth as a backup. Data transfer is done via the Mars Relay network. https://mars.nasa.gov/news/8861/the-mars-relay-network-connects-us-to-nasas-martian-explorers/ https://mars.nasa.gov/news/8861/the-mars-relay-network-conne... Ingenuity uses this as well. > if Starlink works on Earth, there's no real reason it shouldn't work on Mars. Ground-to-low orbit comms relays have been in operations for half a century. And decades at Mars already. Starlink is for high throughput communications from ground to ground on the same planet with always-on connectivity and global coverage. Their sattelites are engineered cheap and replenished with cheap lift from SpaceX. Totally different problem calculus at Mars and doesn't address "backhaul" to Earth. You'd need longer-lived high throughput optical terminals or better, preferably at Areostationary orbits. Checkout work by Briedenthal and Edwards on the matter.

darkwater 5y ago

I'm not a space expert at all but my wild guess is that doing small, careful tests for consumer grade hardware (and software) to be used on Mars is a way to lower costs of a possible future settlement, by leveraging economies of scale here on Earth.

holoduke 5y ago

Anyone knows if the copter suffer already from issues and reboots already?

johnwalkr 5y ago

I actually just wrote the policy on this topic for a space project I work on. It really depends. It’s often not cheaper to use such commodity stuff unless you can tolerate a lot of risk. A lot of normal and automotive components are used in space, but it’s mostly simple stuff that’s relatively easy to qualify. A snapdragon cpu will basically never be fully qualified as a single component because you cannot control each state, let alone test it in each state (I am oversimplifying but bear with me). What you can do is apply overcurrent detection, external watchdog timers, redundancy and other techniques to mitigate the effects of radiation. Then, you can almost fully test that mitigation in theory but not really because it will cost you millions of dollars with huge technical risk and the part will probably be obsoleted or made by a different fab next time you need it. So, you choose older proven easily-hard stuff where you can, or for anything mission critical or related to safety. Better to just buy the $200k rad hard cpu and save your millions. Then, you are left with only bonus items such as this helicopter that simultaneously do not have many reliability requirements (because it’s not mission critical) and are not possible with traditional space grade technology due to performance requirements. “Easier to work with” is only true when you take risks too. Yeah, you can start developing right away using a devkit and Linux, but if you find your project needs reliability or safety requirements, you may find you are spending person-years of software engineering time if it’s even possible at all. You might be required for the system to be deterministic and avoid dynamic memory allocation, for example. Interestingly, students making cubesats don’t “know better” yet. And while the overall failure rate is probably bad on a per mission basis, it’s probably good on a per kg or per dollar basis. There’s a ton of cubesats and cubesat payloads that have raspberry pi’s inside for Leo missions.

kanwisher 5y ago
Yeah but part of why they used a high speed arm, is so they could do realtime AI fo determining where to land and what not. If you use a RAD hardened chip you are limited to what 400-500mhz chips from over a decade ago. Fine if your doing a rover that doesn't need to do realtime detection and avoidance of stuff.

pkaye 5y ago

However they still use a rad-hardened FPGA to run some of the critical aspects. This paper discusses a lot of the internals. https://rotorcraft.arc.nasa.gov/Publications/files/Balaram_AIAA2018_0023.pdf https://rotorcraft.arc.nasa.gov/Publications/files/Balaram_A...

jessriedel 5y ago

A bigger effect is that, by the time the next helicopter is ready to launch, SpaceX Starship is likely to drop the cost of delivering cargo to Mars by an order of magnitude or more. That will massively disrupt the mass vs. capability trade-offs we see in all Martian rovers and aircraft.