I'm not a satellite expert, but I keep reading the debate over whether datacenters in space are viable. So I made a quick hand sketch and turned it into a live simulation with DrawSolve (drawsolve.com), my neurosymbolic engineering app, to estimate the size and mass of the radiators and solar array.
The rough answer I got is a 1 MW datacenter's solar array, radiators and coolant could come in under 10 tonnes (about 8.4 t in the baseline case). That holds if the radiators point away from the sun, the Earth and the ship, so they absorb very little and radiate from both faces. That's tricky, but probably a lot easier than landing a rocket.
I'm sure there are many ways to build them, but rather than scaffolding I suspect you'd roll them up and unroll them in space. Sandwiching two aluminum sheets with the coolant channels between them would allow the channels flatten when rolled and open back up when deployed, without yielding the metal. The side benefit is you’d get direct contact of coolant through the radiator core.
This is a first-order estimate that leaves out the structure, wiring, pumps and power electronics. You don't have to take my assumptions: the study is a live simulation, I’ve made this one so anyone without an account. You can adjust sun exposure, chip temperature margin, solar array weight and radiator thickness, etc and watch the areas and masses update instantly.
All the equations behind the numbers are also listed on the page, along with what each variable means (many have pop up descriptions with hover over). If you think one is wrong, or something important is missing, tell me in the comments. I'd love to know because as I said I’m not an expert in this area.
DrawSolve is a neurosymbolic hybrid LLM and deterministic compute engine. The LLM is used to interpret the sketch and choose equations and a traditional engineering computation stack solves and checks everything and orchestrates between stages. The cool thing is you can see all the assumptions clearly provided and it's impossible for an LLM to sneak in an imagined assumption because all the compute and presentation are done by traditional deterministic code.
Before DrawSolve I spent 23 years in product design; it's the tool I with I had. I've love any feedback anyone has on this particular sim or the general app! It's still just a beta...
I'm not a satellite expert, but I keep reading the debate over whether datacenters in space are viable. So I made a quick hand sketch and turned it into a live simulation with DrawSolve (drawsolve.com), my neurosymbolic engineering app, to estimate the size and mass of the radiators and solar array.
The rough answer I got is a 1 MW datacenter's solar array, radiators and coolant could come in under 10 tonnes (about 8.4 t in the baseline case). That holds if the radiators point away from the sun, the Earth and the ship, so they absorb very little and radiate from both faces. That's tricky, but probably a lot easier than landing a rocket.
I'm sure there are many ways to build them, but rather than scaffolding I suspect you'd roll them up and unroll them in space. Sandwiching two aluminum sheets with the coolant channels between them would allow the channels flatten when rolled and open back up when deployed, without yielding the metal. The side benefit is you’d get direct contact of coolant through the radiator core.
This is a first-order estimate that leaves out the structure, wiring, pumps and power electronics. You don't have to take my assumptions: the study is a live simulation, I’ve made this one so anyone without an account. You can adjust sun exposure, chip temperature margin, solar array weight and radiator thickness, etc and watch the areas and masses update instantly.
All the equations behind the numbers are also listed on the page, along with what each variable means (many have pop up descriptions with hover over). If you think one is wrong, or something important is missing, tell me in the comments. I'd love to know because as I said I’m not an expert in this area.
DrawSolve is a neurosymbolic hybrid LLM and deterministic compute engine. The LLM is used to interpret the sketch and choose equations and a traditional engineering computation stack solves and checks everything and orchestrates between stages. The cool thing is you can see all the assumptions clearly provided and it's impossible for an LLM to sneak in an imagined assumption because all the compute and presentation are done by traditional deterministic code. Before DrawSolve I spent 23 years in product design; it's the tool I with I had. I've love any feedback anyone has on this particular sim or the general app! It's still just a beta...