NYT article.
Stable flight in the very thin atmosphere of Mars is an engineering challenge. From the article:
The engineering involved a series of compromises needed to fit within the constraints of size, weight and battery power.
And a lot of iteration. Starting thirty years ago. Perseverance, without a doubt.
I remember watching a video several years back talking about the challenges of making a Mars Drone, both in terms of making it functional, as well as creating/maintaining the right conditions for testing it while still on Earth.
Seriously fascinating Engineering required on both fronts.
Also cool that the drone will use a Snapdragon processor (basically a modern smartphone processor), while the rover itself will use a lightning fast, 200 MHz processor from 2001.
According to Wikipedia, that SBC (the BAE RAD750) also costs at least $200K.
It’s a good example of real-life engineering tradeoffs: if the main processor fails, you lose the mission (they’re typically set up in a dual-redundant configuration, but it’s always possible for both to fail). So it’s worth trading cost and performance for reliability, especially if you don’t need more performance to meet the mission requirements.
This (at a much smaller scale) is true for the RoboRIO as well: compared to something like a Raspberry Pi or Jetson, it’s slower and more expensive, but it makes up for it in reliability (if the Rio goes down, the entire robot stops working).
Absolutely. The RAD750 is radiation hardened and flight proven. You’ll find most processors used in spacecraft are a few generations behind cutting edge consumer tech. When your device costs a billion or two USD, you can’t afford to have a previously unknown error mode or radiation upset ruin the mission.
Ingenuity, on the other hand, is just a technology demonstration. Since it’s not mission critical, NASA can use it to test a new processor that’s (I think?) never flow before.
Here’s a good one for y’all. One of the key team members of Ingenuity is an FRC alum/volunteer. He hasn’t been able to volunteer the last couple of years because he’s been having fun with that pesky helicopter.
Yup. Closer to $400k, once you throw in some rad-hard SRAM and ROM.
The good news is, the architecture is the same as a PowerPC 750, so for software development and testbed, you can just throw a PPC750 in an ATX motherboard.
Better yet, the GameCube uses a similar processor. So you could also underclock a GameCube to develop software for NASA.
It’s kind of crazy how far the PPC architecture went in the 2000s before ARM got popular. For something relatively new, it managed to make its way into basically everything that wasn’t a Windows PC, including every FRC robot from 2009-2014.
I love software, too.
However, there are some other cool technologies enabling Ingenuity. Check out the insanely light rotor blades. Four foot wingspan, four pounds total weight.
Engineering conservatism and flight heritage are definitely part of the reason that spacecraft processors tend to be antiques, compared to mainstream commercial hardware. However, there are also solid technical reasons.
Digital processors fundamentally work by turning transistors on and off. These days, the transistors are MOSFETs. One of the fundamental limits to how fast MOSFETs can switch on and off is gate capacitance: the higher the gate capacitance, the longer it takes the gate to “charge up” and switch on. There are a couple ways to improve this. First, you can make the transistor physically smaller, directly decreasing the gate capacitance. Second, you can design the transistor to operate at lower voltages, so that fewer coulombs have to be shoved into the gate to turn it on.
Altogether, this means that as time goes by, processors get smaller and lower-voltage. Shrinking the size (e.g. “14 nm process”) allows you to fit more transistors on the same area of silicon, decreases the gate capacitance, and decreases speed-of-light delay (although that’s a much less significant factor). Similarly, CPU core voltages are decreasing: 5 V and 3.3 V used to be common, then went to 2.5 V, and are now down to 1.6 V or so.
Unfortunately, all of these things negatively affect radiation tolerance, one of the key requirements for space electronics. Radiation primarily affects electronics by ionizing an atom, and thus depositing some charge. If this happens near a MOSFET gate, the charge can change the state of the MOSFET, leading to effects like SEUs, or worse, latchups. Any changes to a MOSFET which decrease the amount of charge needed to change the state – such as shrinking it, or designing it for a lower voltage – mean that it will be more sensitive to ionizing radiation.
Thus, in addition to sticking with designs like the PPC750 because they’re tried and true, older processors are also quantitatively more resistant to radiation than newer ones.
Here’s a somewhat behind the scenes video from Veritasium talking to some of the engineers behind the Ingenuity drone for anyone curious as to some of the design decisions and trade offs:
This topic was automatically closed 365 days after the last reply. New replies are no longer allowed.
