10 ms·

Building a rocket engine from scratch

362 points by sr-latch 2y ago

datadrivenangel 2y ago
"At one point, I received a well-deserved earful from our machinist about how tightly spaced the turbine blades were. The program was taking nearly a month to run and required tiny end mills that broke often. We performed a turbine blade count study to see if we could use fewer blades with more space between them. It turned out the performance impact of running fewer blades was minimal, so we cut the number down, allowing our machinist to use larger, less fragile tools. Machine time dropped to less than a day, which was a significant win for turbine cost and machine time. It was also a good lesson in thinking comprehensively about a design’s manufacturability (those passages between the blades looked so big on the computer screen!) in addition to its performance. " Once again people learn the hard way that it's valuable to have tight feedback cycles and embedded knowledge on your team.
- FredPret 2y ago
  And also the value of a completely obsessed engineer. If the mech eng designing the parts is also the type to build things in his spare time, this type of machining problem would stand out right away. Of course, not everything can ever be anticipated, so tight feedback loops are fantastic when you can get them.
  - numpad0 2y ago
    I'd agree with this take more. It's just cheaper and faster to just know whys and plan ahead than waiting for failures, if you could. I've been tinkering with 3D printers long enough that I've trivially halved print times after seeing long print time predictions. Lots of my brackets and parts fits together on first try and needs no supports. Rapid turnaround did help in _acquiring_ those skills, but now I could just waterfall little projects and tight looping is not that important.
  - psunavy03 2y ago
    Valuing "completely obsessed" employees is a great way to build a company with toxic burnout culture. Some people want lives outside of work and are still good at their jobs.
    - quartesixte 2y ago
      Or some people are so good at their jobs that even while being completely obsessed, they still have great work-life balance. I interact with a lot of people like the author in the article from this industry. And while yes, their worklife balance skews "work", they all have interesting hobbies and find time to hang out with friends, go on adventures, and live a pretty full life. Unless you're on the Starship team at SpaceX. Truly, hats off to y'all at Boca Chica some intense hours you are pulling there.
      - nothercastle 2y ago
        Employees who are married to their job are the best employees. Just like in a real relationship you can treat them like shit and they mostly stay married
        greenish_shores 2y ago
        Honestly I did read that as "Employers who are married to their job are the best employers". You probably won't treat them like shit, but they'll still remain married anyway. But such employees will sometimes leave and build their own stuff. Which they'll very likely succeed at, exactly because of the approach they have. Ultimately, I think these are a bit different fields/markets, and a bit different kind of jobs. Engineers creating real unique new stuff, vs more "bland" engineers just being a hand to accelerate someone's efforts. I worked in both roles and can clearly see pros and cons of both. Largely depends also on life phase one's at, which one would a person prefer.
        avmich 2y ago
        > But such employees will sometimes leave and build their own stuff. Which they'll very likely succeed at, exactly because of the approach they have. Microsoft was known at some time to provide good environment for their engineers so they could focus on making the product, instead of learning the business side of things, which both removed distractions and reduced the incentives to leave and do things without well-learned habits.
- cwillu 2y ago
  It was an explicit choice they made to have this manufactured in-house, to _enable_ that tight feedback loop. Far from a hard lesson, it was one they designed the company around.
AndriyKunitsyn 2y ago
Nice read. The fact that it is possible to 3d print metal parts that withstand temperatures and pressures of a rocket engine is so exciting. How expensive is it?
- jfoutz 2y ago
  Stratasys? I’m not sure about pricing, and the website won’t say. However, a bunch of places rent out time on those machines. Draw up your rocket, and get a quote. Price is generally cc^2 volume Metal is not cheap. Make a few out of plastic to verify dimensions.
  - quartesixte 2y ago
    Plastic to verify dimensions is a great approach. But don't forget that the laser sintering of metal powders might result in design constraints not present in plastic printing!
- WJW 2y ago
  Mostly depends on the volume of the part, or equivalently it's weight. Complexity you get mostly "for free" when it comes to 3d printing. What type of material you need to "withstand temperatures and pressures of a rocket engine" is entirely dependent on which part of the rocket engine we're talking about. A fuel injector has radically different requirements than a supporting strut for example. 3d printed titanium goes for 300-400 USD/kg, steel is a bit cheaper at ~150 USD/kg for most inconel grades.
- spxneo 2y ago
  side question to this: where can i design stuff involving metal parts (presumably in CAD tools) and have it printed en masse? With PCBs? ex) car components
  - numpad0 2y ago
    You can draw up mechanical components in Autodesk Fusion, OnShape, SOLIDWORKS for Makers, or FreeCAD, and send STEP or STL to PCBWay or JLCPCB in China for manufacturing(note that export restrictions may apply if it's literal rockets or otherwise dual-use/controlled in nature). PCB mounting outlines can be exported from above 3D CAD and imported to EDA tools such as Altium and KiCAD; KiCAD is fine unless you're doing DRAM or PCIe. Same PCBWay and JLCPCB takes your design, and optionally assemble PCB with parts for you. That should take you to first 2-3 working units at ~$500 and up to few dozen beta units with zero initial cost and much inflated unit costs, and I guess beyond that involves significant human resourcing and networking problems outside of PoC hardware scope.
spxneo 2y ago
now all that is required is a github for passive radar...for educational purposes
- sr-latch 2y ago
  https://github.com/krakenrf/krakensdr_docs/wiki/ https://github.com/krakenrf/krakensdr_docs/wiki/ The SDR world is incredible
  - spxneo 2y ago
    Is this the codebase that they took down ???! Yes to all of the things you can do (minus the passive radar stuff beyond educational purposes that is ;) edit: looks like they took down all the passive radar doppler videos : https://othernet.is/products/kerberossdr-4x-coherent-rtlsdr https://othernet.is/products/kerberossdr-4x-coherent-rtlsdr
avmich 2y ago
Awesome story, thanks! It seems the design choices are rather conservative, which is entirely justified by the "from scratch" part for the first engine. I'm sure subsequent designs will be more bold and adventurous. Keep up great work!
brcmthrowaway 2y ago
Can you use ChatGPT to build a Saturn V?
- trollerator23 2y ago
  You can try!
  - brcmthrowaway 2y ago
    I would get sent to jail for breaking ITAR
    - trollerator23 2y ago
      I don't see how. Because ChatGPT is not from a US company?
- WalterBright 2y ago
  If it's like some code I asked ChatGPT to generate, it will blow up on the pad!
buescher 2y ago
NASA's report server is a national treasure, especially the material from the 50s and 60s that he references. Some of the most crisp and succinct technical writing you'll find, and you can infer a lot about how they ran projects. Declassified NRO reports are also very good - you can see the Lockheed Skunk Works principles in action. Example: https://www.nro.gov/Portals/135/documents/foia/declass/WS117L_Records/984.PDF https://www.nro.gov/Portals/135/documents/foia/declass/WS117...
- Horffupolde 2y ago
  English is great for technical writing.
  - beambot 2y ago
    It probably helps that you have to be exceedingly thoughful about every keystroke & word less you need to completely retype the document or page!
    - scoot 2y ago
      "lest" perhaps? At least you don't have to retype the whole sentence. ;)
    - mhh__ 2y ago
      I'd have imagined you wrote a draft, corrected with the red pen, and then sent it off to (presumably) an army of women to do it properly.
  - adrianN 2y ago
    I think English has too many ambiguities to be well suited for technical writing. There are efforts like https://en.wikipedia.org/wiki/Simplified_Technical_English https://en.wikipedia.org/wiki/Simplified_Technical_English but they don't have a lot of adoption.
    - trhway 2y ago
Linda231 2y ago
[dead]
mdorazio 2y ago
Great post on building extremely complex hardware from scratch. At the same time I hate to be that business guy, but both this blog post and the abl site are missing a good answer to my first question: Why? Given that SpaceX exists and is quickly approaching feasibility of Starship on top of Falcon, what is the primary goal of this rocket system? How will it compete? Who will its customers be? Is it getting its metric ton payload to orbit faster/cheaper/easier? Is this "from scratch" engine design superior to existing designs in some way? What is its current ISP? Is Jet-A + LOX a better fuel choice given expected mission parameters? I'd love to see a blog post that tackles these kinds of questions.
- avmich 2y ago
  SpaceX is certainly giving a hard time to its competitors, but it doesn't mean they shouldn't exist. Some of them may actually follow the same route, designing reusable hardware and reducing launch costs. SpaceX took 20 years to become relying on robustly reusable system; maybe some other companies reach a similar state sooner.
- choeger 2y ago
  From the outside: Diversification is always great. Build a whole ecosystem of small-scale rocket manufacturers instead of one big monopoly. That will foster competition and innovation. From the investor: SpaceX might fail. Even if there Falcons are pretty much unbeatable now, you don't know what's going to happen with Starship. And even the Falcons could conceivably be grounded for years after some hypothetical flaw is found. More likely: With the price reductions made by SpaceX, the market will grow and there will be more than enough clients. From the inside: Because it's a fun challenge and literally rocket science, of course.
  - hinkley 2y ago
    SpaceX was looking to reduce prices to orbit by 10x, and they are currently sitting at around 6.7x.
    - avmich 2y ago
z3t4 2y ago
The science of pressure chambers have also advanced, we could just pump whatever material, like liquid air into a pressure tank and then load it into the rocket. No mixing or pumping needed just open the valve and let the pressure out and you will have a very cheap and simple rocket.
- sr-latch 2y ago
  This is absolutely not true - injector design is the most important aspect of designing a thrust chamber. Poor mixing of propellants leads to severe combustion instability, which often leads to explosions. Even the earliest space programs did significant testing on propellant choices and injector designs (see Ignition! by John D. Clark) Also, pressure fed rockets have always been a fairly terrible design. Pressure feeding requires heavy tanks, and incurs a big mass fraction (dry mass / wet mass) penalty. Outside of rare cases, it's only used for ground testing.
  - z3t4 2y ago
    just drop the tank when the weight/thrust ratio is too low. Thanks for the book suggestion! (have not read it yet)
    - quartesixte 2y ago
      The rocket equation would like to have a word with you. The wall thicknesses required to create rockets with enough thrust to get to orbit on pressure-fed would render the rocket physically impossible. There is also a lot of control that goes into flying a rocket, and pressure-fed rockets are kind of hard to control.
cuSetanta 2y ago
Oh wow, I work for a supplier for ABL and am today in the process of putting some of their stuff into our thermal chamber for cycling. Thats neat. We work for a lot of launcher companies, but ABL is the most interesting for me (even though we do relatively little for them). The containerised approach to the entire system is a really clever adaptation of existing methods to create a rapid launch system.
- philipwhiuk 2y ago
  The jury is very much out on 'rapid'.
criddell 2y ago
> You might call me an unlikely candidate to lead an engine program from scratch, but I was hired by ABL in 2018 to do just that. My background was in commercial aircraft interiors, web development, semiconductor fab fluid components, and SpaceX Falcon 9 hydraulic systems. I don't mean this as an insult, but why did they hire you? It's obvious now that you were a great choice, but from your background story I wouldn't have guessed that to be true.
- quartesixte 2y ago
  Seems like from the blogpost that the writer and the founder were part of the same cohort at SpaceX. I would hazard a guess that they became friends, and had planned to do this together, and he joined ("hired") as soon as feasible for him (or the founder gained enough traction to pull him away from SpaceX).
  - quartesixte 2y ago
    Oh and as an additional comment: Propulsion Engineers are a niche, specialized branch of mechanical engineers. They are the ones who usually design engines. The writer probably was referring to the fact that he was just a regular mechanical engineer and not a Propulsion Engineer.
alfor 2y ago
I was wondering what is that space company and what are their edge with spaceX. On the site you can see: launch on demand, simple system that can go anywere, tactical launchs. This is for nukes or similar stuff.
- philipwhiuk 2y ago
  It's not for nukes. We have silos and submarines for those. It's for "responsive launch" and (skeptically) because the DoD has lots of money for space and not much idea what to do with it. It's the Astra business model, hopefully without the Astra failure model. (And practically speaking it's because you can't bootstrap a heavy lift launch company on VC funding or even a SPAC - the small sat launcher is a proof-of-concept for your medium/large launch vehicle).
hi-v-rocknroll 2y ago
Since it's 3D printed, I'm guessing from the embedded ports that the nozzle is hollow in parts and Jet A cooled since LOX latent heat of evaporation is orders of magnitude less. Probably one of those ports is for a temperature sensor.
mrbluecoat 2y ago
> Example of a Clean Sheet lol, engineering humor

362 points by sr-latch 2y ago

"At one point, I received a well-deserved earful from our machinist about how tightly spaced the turbine blades were. The program was taking nearly a month to run and required tiny end mills that broke often. We performed a turbine blade count study to see if we could use fewer blades with more space between them. It turned out the performance impact of running fewer blades was minimal, so we cut the number down, allowing our machinist to use larger, less fragile tools. Machine time dropped to less than a day, which was a significant win for turbine cost and machine time. It was also a good lesson in thinking comprehensively about a design’s manufacturability (those passages between the blades looked so big on the computer screen!) in addition to its performance. " Once again people learn the hard way that it's valuable to have tight feedback cycles and embedded knowledge on your team.

FredPret 2y ago
And also the value of a completely obsessed engineer. If the mech eng designing the parts is also the type to build things in his spare time, this type of machining problem would stand out right away. Of course, not everything can ever be anticipated, so tight feedback loops are fantastic when you can get them.
- numpad0 2y ago
  I'd agree with this take more. It's just cheaper and faster to just know whys and plan ahead than waiting for failures, if you could. I've been tinkering with 3D printers long enough that I've trivially halved print times after seeing long print time predictions. Lots of my brackets and parts fits together on first try and needs no supports. Rapid turnaround did help in _acquiring_ those skills, but now I could just waterfall little projects and tight looping is not that important.
- psunavy03 2y ago
  Valuing "completely obsessed" employees is a great way to build a company with toxic burnout culture. Some people want lives outside of work and are still good at their jobs.
  - quartesixte 2y ago
    Or some people are so good at their jobs that even while being completely obsessed, they still have great work-life balance. I interact with a lot of people like the author in the article from this industry. And while yes, their worklife balance skews "work", they all have interesting hobbies and find time to hang out with friends, go on adventures, and live a pretty full life. Unless you're on the Starship team at SpaceX. Truly, hats off to y'all at Boca Chica some intense hours you are pulling there.
    - nothercastle 2y ago
      Employees who are married to their job are the best employees. Just like in a real relationship you can treat them like shit and they mostly stay married
      - greenish_shores 2y ago
        Honestly I did read that as "Employers who are married to their job are the best employers". You probably won't treat them like shit, but they'll still remain married anyway. But such employees will sometimes leave and build their own stuff. Which they'll very likely succeed at, exactly because of the approach they have. Ultimately, I think these are a bit different fields/markets, and a bit different kind of jobs. Engineers creating real unique new stuff, vs more "bland" engineers just being a hand to accelerate someone's efforts. I worked in both roles and can clearly see pros and cons of both. Largely depends also on life phase one's at, which one would a person prefer.
        avmich 2y ago
        > But such employees will sometimes leave and build their own stuff. Which they'll very likely succeed at, exactly because of the approach they have. Microsoft was known at some time to provide good environment for their engineers so they could focus on making the product, instead of learning the business side of things, which both removed distractions and reduced the incentives to leave and do things without well-learned habits.
cwillu 2y ago
It was an explicit choice they made to have this manufactured in-house, to _enable_ that tight feedback loop. Far from a hard lesson, it was one they designed the company around.

AndriyKunitsyn 2y ago

Nice read. The fact that it is possible to 3d print metal parts that withstand temperatures and pressures of a rocket engine is so exciting. How expensive is it?

jfoutz 2y ago
Stratasys? I’m not sure about pricing, and the website won’t say. However, a bunch of places rent out time on those machines. Draw up your rocket, and get a quote. Price is generally cc^2 volume Metal is not cheap. Make a few out of plastic to verify dimensions.
- quartesixte 2y ago
  Plastic to verify dimensions is a great approach. But don't forget that the laser sintering of metal powders might result in design constraints not present in plastic printing!
WJW 2y ago
Mostly depends on the volume of the part, or equivalently it's weight. Complexity you get mostly "for free" when it comes to 3d printing. What type of material you need to "withstand temperatures and pressures of a rocket engine" is entirely dependent on which part of the rocket engine we're talking about. A fuel injector has radically different requirements than a supporting strut for example. 3d printed titanium goes for 300-400 USD/kg, steel is a bit cheaper at ~150 USD/kg for most inconel grades.
spxneo 2y ago
side question to this: where can i design stuff involving metal parts (presumably in CAD tools) and have it printed en masse? With PCBs? ex) car components
- numpad0 2y ago
  You can draw up mechanical components in Autodesk Fusion, OnShape, SOLIDWORKS for Makers, or FreeCAD, and send STEP or STL to PCBWay or JLCPCB in China for manufacturing(note that export restrictions may apply if it's literal rockets or otherwise dual-use/controlled in nature). PCB mounting outlines can be exported from above 3D CAD and imported to EDA tools such as Altium and KiCAD; KiCAD is fine unless you're doing DRAM or PCIe. Same PCBWay and JLCPCB takes your design, and optionally assemble PCB with parts for you. That should take you to first 2-3 working units at ~$500 and up to few dozen beta units with zero initial cost and much inflated unit costs, and I guess beyond that involves significant human resourcing and networking problems outside of PoC hardware scope.

spxneo 2y ago

now all that is required is a github for passive radar...for educational purposes

sr-latch 2y ago
https://github.com/krakenrf/krakensdr_docs/wiki/ https://github.com/krakenrf/krakensdr_docs/wiki/ The SDR world is incredible
- spxneo 2y ago
  Is this the codebase that they took down ???! Yes to all of the things you can do (minus the passive radar stuff beyond educational purposes that is ;) edit: looks like they took down all the passive radar doppler videos : https://othernet.is/products/kerberossdr-4x-coherent-rtlsdr https://othernet.is/products/kerberossdr-4x-coherent-rtlsdr

avmich 2y ago

Awesome story, thanks! It seems the design choices are rather conservative, which is entirely justified by the "from scratch" part for the first engine. I'm sure subsequent designs will be more bold and adventurous. Keep up great work!

brcmthrowaway 2y ago

Can you use ChatGPT to build a Saturn V?

trollerator23 2y ago
You can try!
- brcmthrowaway 2y ago
  I would get sent to jail for breaking ITAR
  - trollerator23 2y ago
    I don't see how. Because ChatGPT is not from a US company?
WalterBright 2y ago
If it's like some code I asked ChatGPT to generate, it will blow up on the pad!

buescher 2y ago

NASA's report server is a national treasure, especially the material from the 50s and 60s that he references. Some of the most crisp and succinct technical writing you'll find, and you can infer a lot about how they ran projects. Declassified NRO reports are also very good - you can see the Lockheed Skunk Works principles in action. Example: https://www.nro.gov/Portals/135/documents/foia/declass/WS117L_Records/984.PDF https://www.nro.gov/Portals/135/documents/foia/declass/WS117...

Horffupolde 2y ago
English is great for technical writing.
- beambot 2y ago
  It probably helps that you have to be exceedingly thoughful about every keystroke & word less you need to completely retype the document or page!
  - scoot 2y ago
    "lest" perhaps? At least you don't have to retype the whole sentence. ;)
  - mhh__ 2y ago
    I'd have imagined you wrote a draft, corrected with the red pen, and then sent it off to (presumably) an army of women to do it properly.
- adrianN 2y ago
  I think English has too many ambiguities to be well suited for technical writing. There are efforts like https://en.wikipedia.org/wiki/Simplified_Technical_English https://en.wikipedia.org/wiki/Simplified_Technical_English but they don't have a lot of adoption.
  - trhway 2y ago

Linda231 2y ago

[dead]

mdorazio 2y ago

Great post on building extremely complex hardware from scratch. At the same time I hate to be that business guy, but both this blog post and the abl site are missing a good answer to my first question: Why? Given that SpaceX exists and is quickly approaching feasibility of Starship on top of Falcon, what is the primary goal of this rocket system? How will it compete? Who will its customers be? Is it getting its metric ton payload to orbit faster/cheaper/easier? Is this "from scratch" engine design superior to existing designs in some way? What is its current ISP? Is Jet-A + LOX a better fuel choice given expected mission parameters? I'd love to see a blog post that tackles these kinds of questions.

avmich 2y ago
SpaceX is certainly giving a hard time to its competitors, but it doesn't mean they shouldn't exist. Some of them may actually follow the same route, designing reusable hardware and reducing launch costs. SpaceX took 20 years to become relying on robustly reusable system; maybe some other companies reach a similar state sooner.
choeger 2y ago
From the outside: Diversification is always great. Build a whole ecosystem of small-scale rocket manufacturers instead of one big monopoly. That will foster competition and innovation. From the investor: SpaceX might fail. Even if there Falcons are pretty much unbeatable now, you don't know what's going to happen with Starship. And even the Falcons could conceivably be grounded for years after some hypothetical flaw is found. More likely: With the price reductions made by SpaceX, the market will grow and there will be more than enough clients. From the inside: Because it's a fun challenge and literally rocket science, of course.
- hinkley 2y ago
  SpaceX was looking to reduce prices to orbit by 10x, and they are currently sitting at around 6.7x.
  - avmich 2y ago

z3t4 2y ago

The science of pressure chambers have also advanced, we could just pump whatever material, like liquid air into a pressure tank and then load it into the rocket. No mixing or pumping needed just open the valve and let the pressure out and you will have a very cheap and simple rocket.

sr-latch 2y ago
This is absolutely not true - injector design is the most important aspect of designing a thrust chamber. Poor mixing of propellants leads to severe combustion instability, which often leads to explosions. Even the earliest space programs did significant testing on propellant choices and injector designs (see Ignition! by John D. Clark) Also, pressure fed rockets have always been a fairly terrible design. Pressure feeding requires heavy tanks, and incurs a big mass fraction (dry mass / wet mass) penalty. Outside of rare cases, it's only used for ground testing.
- z3t4 2y ago
  just drop the tank when the weight/thrust ratio is too low. Thanks for the book suggestion! (have not read it yet)
  - quartesixte 2y ago
    The rocket equation would like to have a word with you. The wall thicknesses required to create rockets with enough thrust to get to orbit on pressure-fed would render the rocket physically impossible. There is also a lot of control that goes into flying a rocket, and pressure-fed rockets are kind of hard to control.

cuSetanta 2y ago

Oh wow, I work for a supplier for ABL and am today in the process of putting some of their stuff into our thermal chamber for cycling. Thats neat. We work for a lot of launcher companies, but ABL is the most interesting for me (even though we do relatively little for them). The containerised approach to the entire system is a really clever adaptation of existing methods to create a rapid launch system.

philipwhiuk 2y ago
The jury is very much out on 'rapid'.

criddell 2y ago

> You might call me an unlikely candidate to lead an engine program from scratch, but I was hired by ABL in 2018 to do just that. My background was in commercial aircraft interiors, web development, semiconductor fab fluid components, and SpaceX Falcon 9 hydraulic systems. I don't mean this as an insult, but why did they hire you? It's obvious now that you were a great choice, but from your background story I wouldn't have guessed that to be true.

quartesixte 2y ago
Seems like from the blogpost that the writer and the founder were part of the same cohort at SpaceX. I would hazard a guess that they became friends, and had planned to do this together, and he joined ("hired") as soon as feasible for him (or the founder gained enough traction to pull him away from SpaceX).
- quartesixte 2y ago
  Oh and as an additional comment: Propulsion Engineers are a niche, specialized branch of mechanical engineers. They are the ones who usually design engines. The writer probably was referring to the fact that he was just a regular mechanical engineer and not a Propulsion Engineer.

alfor 2y ago

I was wondering what is that space company and what are their edge with spaceX. On the site you can see: launch on demand, simple system that can go anywere, tactical launchs. This is for nukes or similar stuff.

philipwhiuk 2y ago
It's not for nukes. We have silos and submarines for those. It's for "responsive launch" and (skeptically) because the DoD has lots of money for space and not much idea what to do with it. It's the Astra business model, hopefully without the Astra failure model. (And practically speaking it's because you can't bootstrap a heavy lift launch company on VC funding or even a SPAC - the small sat launcher is a proof-of-concept for your medium/large launch vehicle).

hi-v-rocknroll 2y ago

Since it's 3D printed, I'm guessing from the embedded ports that the nozzle is hollow in parts and Jet A cooled since LOX latent heat of evaporation is orders of magnitude less. Probably one of those ports is for a temperature sensor.

mrbluecoat 2y ago

> Example of a Clean Sheet lol, engineering humor