[kragen]

better, lighter-weight computing stacks have been a huge boost; that's what made cubesats possible. but the much bigger deal is the dramatic drop in launch costs driven by spacex, even though so far that's only a factor of 3.4 https://ourworldindata.org/grapher/cost-space-launches-low-e...

with lightweight computers driven by the cellphone industry, it became possible for a small team or even individual to launch a low-power ham radio satellite or weather satellite. but they can't launch a high-resolution space telescope, earth observing satellite, or high-power communications satellite, nor can they do laser communication pointcasting. and lightweight computers are a crucial enabler for starlink-style communications constellations, but there's only one of those, because that's still a big-money kind of project

suppose that, instead, you had 01980s computing power, but the cost of space launch dropped by a factor of 100. if you need to launch a 200-kg satellite to get the sky-observing optical aperture you're looking for so diffraction doesn't cremate you, you don't care if the onboard computers weigh 1 gram or 10 kilograms. (i mean, you do care, because it lets you cut your launch budget 10%, but it's not a dominant determinant of viability.) with saturn v or zenit 2, according to the plot linked above, that launch would cost you a million dollars. today, at falcon heavy's 1500 dollars per kg, it's 300,000 dollars, which is already a radically more feasible project

spacex's 'starship' is supposed to carry 150 tonnes to leo for 10 million dollars. that's 70 dollars a kilogram. our hypothetical 200-kg aditya athalye space telescope satellite would then cost 14000 dollars to launch. it becomes a hobby project comparable in cost to an engine lathe or a camper van. this would change the economics of space in a profound way, far beyond what cellphone chips have done

for comparison, the csis aerospace security project number for https://en.wikipedia.org/wiki/Scout_(rocket_family) on that chart was 118500 dollars per kg in 01961. (but of course you couldn't launch a cubesat on it for that price; it was a military thing.) by 01967 saturn v had brought that down by a factor of 22. after that it remained constant for 43 years until falcon 9 in 02010. starship, if it works, will reduce launch costs by that same factor of 22 over the current falcon heavy number i described above, and by a factor of 73 over saturn v

[HeadsUpHigh]

The decline in price per kg from Falcon 9 has made a big difference and yet it's nowhere near as impactful as Starship is going to be... falcon 9 is either too powerful for LEO missions and limited by fairing volume( so can't really launch space stations on a falcon 9) or too weak to push further into GEO and into the solar system. Starship will enable a lot of LEO applications previously unthinkable.

[adityaathalye]

Oh I don't dispute the fact that reusable vehicles are an incredible leap forward. Anything that manages to launch significantly more payload cheaper and more frequently is awesome.

That said, things like reusability work at all because of precision control which, as far as I can tell, is near-impossible to do in a lightweight package without the cooperation of a compute stack. Same for other kinds of soft landings and autonomous control scenarios.

As an aside: achievements like reusable rockets also became possible because of advances in materials research driven by advances in computer simulation and computer-aided manufacturing. So this is perhaps another under-appreciated layer of the rocketry compute stack.

[kragen]

yes, those are excellent points i strongly agree with

what kinds of new materials are they using in the falcon rockets, do you know?

quite aside from the weight, latency is extremely important to precision control; a factor of 10 in latency may be equivalent to 100× or 1000× more sensor error. so, even if weight was no object and you could mount a dozen 5-tonne 10-megaflops cdc 6600 supercomputers on your reusable rocket as the guidance system, you might still be better off with a 100-megaflops high-end microcontroller, just because it can respond more rapidly to perturbations of homeostasis

[adityaathalye]

> what kinds of new materials are they using in the falcon rockets, do you know?

I was thinking of materials innovation and new materials.

Without looking anything up, I would guess exotic alloys, ceramics, polymers, fuels and fuel mixes etc. at a chemical level, stress/failure modeling at a mechanical level, new manufacturing techniques that exploit properties of existing materials to achieve new capabilities, increased production efficiency for faster turnaround at better tolerances and integrity (because you figured out how to precision weld the thing in a new way) etc.

(edit: clarity)

[idlewords]

The big improvement is not launch costs, but miniaturization and automation. A university team can make a credible lunar rover or cubesat today using off-the-shelf components. Unlike launch vehicles, that technology did not exist in 001960, or even 002000, at any price.

[kragen]

i came here for an argument, but this is just contradiction

it doesn't matter if you can make a credible lunar rover or not if the price to launch it is three times the grant the nsf will give your university team. given that nasa published their lunar roving vehicle documentation in mostly 0001973, down to circuit schematics and some machining dimensions, i'm pretty sure a university team in 00002000 could have duplicated it for less than the several million dollars the launch would cost. it only took boeing 17 months to design and build them in the first place, and they mostly use technology from the 000001940s (aluminum tubing, nylon webbing, wire mesh wheels, silver-zinc plexiglass batteries, brushed dc motors, cable brakes; fiberglass arm rests and fenders were apparently the highest-tech part and, unsurprisingly, the part that failed)

https://www.nasa.gov/history/alsj/alsj-LRVdocs.html

with respect to cubesats, you're in violent agreement with my comment. you obviously can't put a trs-80 in a cubesat. in 0000002000 you could put a basic stamp in it, but you couldn't get mems gyros and accelerometers, and you weren't going to be able to run your star tracker camera off a pic16. computer and imu miniaturization is a big deal for cubesats. that's one of the main points of my comment

the other main point is that it isn't nearly as big a deal for bigger satellites; when we were launching cubesats (before i joined the team) it was really important that we could use tiny cellphone components, but once we were launching 37kg monstrosities, the fact that the gumstix boards only weighed a few grams was just nice, not critical. the optics weighed a lot more

so tell us, what's your experience?

[idlewords]

My big-picture point is simply that electronics are getting small and cheap much faster than launch costs are decreasing, so that is the important trend. We could have a revolution in space exploration and remote sensing even if launch costs flatlined.

Beyond that I'm not sure what we're arguing about, so I'll tap out here.

[kragen]

they aren't commensurable; as ben tilly pointed out, you cannot replace the oxygen you need to breathe with a sufficiently large quantity of granite (without equipment and energy to process it, anyway), and you cannot replace lower launch costs with smaller and cheaper electronics, in particular in the areas you named, space exploration and remote sensing, for the reasons i described

now, bring me a shrubbery!