4 ms·

I take your point, but it does seem to have actually worked.

WalterBright 4y ago
Obviously a great deal of the cost also came from making sure it would work, because there wouldn't be another. What if 10 were built, each with only a 10% chance of success? What would that have cost? After all, it doesn't need the expense of being man-rated.
- zabzonk 4y ago
  > What would that have cost? No idea. Do you? Genuinely interested, particularly regarding methodology. Not intending at all to be insulting.
  - WalterBright 4y ago
    Look at SpaceX's rockets. They cut costs by what, 10x? over NASA's? A part of that was not trying to build 100% the first time, but to accept failure and iterate. My dad told me that Pratt&Whitney made the most reliable aircraft engines by putting them on a test stand and running them at full power until they broke. The engineers would figure out why it failed, redesign the part, stick that on the engine, and continued running them at full power. It's straightforward, inexpensive, and it works.
    - noizejoy 4y ago
      Could it be, that iterating fast through additional failures may be harder to finance with public funds, because the general public hasn’t caught up with the recognition of failing early and often being a good thing?
      - WalterBright 4y ago
        Government programs fail all the time, and they just get a budget increase. Hire the salesmen who accomplish that.
        noizejoy 4y ago
        lol - you do have a point there!
    - dekhn 4y ago
      Surely you can see that the P&W approach is only going to discover a subset of problems that will affect the engine in the wild. For example, engines don't run at 100% power during the full flight, and they go through heating and cooling cycles. It also doesn't address failure modes due to manufacturing problems specific to a single engine- for example, turbine blades are manufactured with serial numbers and you can go back and get detailed process and QA for that specific item to understand what went wrong and why it failed and how to not do that during manufacture again. Then there's all sorts of complex failures that aren't addressed by single engines: " Prior to this crash, the probability of a simultaneous failure of all three hydraulic systems was considered as low as one in a billion. However, statistical models did not account for the position of the number-two engine, mounted at the tail close to hydraulic lines, nor the results of fragments released in many directions. Since then, aircraft engine designs have focused on keeping shrapnel from puncturing the cowling or ductwork, increasingly utilizing high-strength composite materials to achieve penetration resistance while keeping the weight low" (I'm not disrespecting P&W- I'm sure they have more tests than just "100% power until it breaks, fix and repeat")
      - WalterBright 4y ago
        > Surely you can see that the P&W approach is only going to discover a subset of problems that will affect the engine in the wild. Keep in mind that my former job was designing gearbox parts for the Boeing 757. This included doing the math, and devising a test plan. I've spent a lot of time on "what could go wrong" scenarios in the real world. Also, when I prepare slides for a coding presentation, the implementation code for a concept gets trimmed way down to what will fit on a slide.
- mmazing 4y ago
  What is the measure of success though? Successful launch and arrival at the observation point? The 6 month setup period before observations can be made? I contend that success here is a full mission that yields science data over decades of observation. If we cut down the acceptable error rate so we can launch early, how does that impact longevity?
  - zmgsabst 4y ago
    I think the question is, imagine we built 10 — and got an outcome like: 2x — failed to deploy, mission lost 4x — early operation error, 6mo lifetime 3x — lower performance, 3 year lifetime 1x — mission successful, 10 year lifetime …do we get a better deal building them with larger faults, at a lower cost? A lot of expense is in developing technologies, assembly tooling, and test rigs — all of which is easier if we don’t need them to be as assured because only 1 in 10 needs to succeed.
    - WalterBright 4y ago
      Right. And if they are launched serially, you've got a chance to correct mistakes in the previous launches. Odds of success increase with every launch.
      - mmazing 4y ago
        Yeah, makes more sense when I think about it like that. Thanks!
- prpl 4y ago
  What is the timescale you are thinking of for 10% chance of success? 10% first light success is different than 10% 10-year survey success. There’s physically not enough infrastructure (clean rooms, testing facilities, vacuum chambers, etc…) nor skilled manpower available to NASA to be able to build more than one in parallel, so that would add to the cost significantly.
  - WalterBright 4y ago
    What's being done with that infrastructure (clean room, testing facilities, vacuum chambers, etc..) right now? Nothing? Of course, the sensible thing to do is to build the test equipment, procedures, train the testers. Then test #1. Then test #2. Then test #3. That's what everybody does. The cost of the Saturn V rocket was amortized over many launch vehicles, despite a host of new technologies that had to be developed for it, despite much of it being hand-built.
    - zabzonk 4y ago
      But the cost of grinding the mirrors?
      - WalterBright 4y ago
        Consider the cost of designing the mirrors, designing a machine to grind the mirrors, testing the machine, and verifying the grind of the mirrors. All that adds to the cost of the first mirror, and adds $0 to the second.
        zabzonk 4y ago
        But the grinding machine costs (which is significant if you have 10 of them) and the grinding process takes a lot of time and cost (if you have 1 machine x 10 mirrors). I don't think these NASA/JPL guys are quite as dim as you seem to suggest (but you have more info) and they may have done a little cost/benefit analysis before embarking on the project.

Obviously a great deal of the cost also came from making sure it would work, because there wouldn't be another. What if 10 were built, each with only a 10% chance of success? What would that have cost? After all, it doesn't need the expense of being man-rated.

zabzonk 4y ago
> What would that have cost? No idea. Do you? Genuinely interested, particularly regarding methodology. Not intending at all to be insulting.
- WalterBright 4y ago
  Look at SpaceX's rockets. They cut costs by what, 10x? over NASA's? A part of that was not trying to build 100% the first time, but to accept failure and iterate. My dad told me that Pratt&Whitney made the most reliable aircraft engines by putting them on a test stand and running them at full power until they broke. The engineers would figure out why it failed, redesign the part, stick that on the engine, and continued running them at full power. It's straightforward, inexpensive, and it works.
  - noizejoy 4y ago
    Could it be, that iterating fast through additional failures may be harder to finance with public funds, because the general public hasn’t caught up with the recognition of failing early and often being a good thing?
    - WalterBright 4y ago
      Government programs fail all the time, and they just get a budget increase. Hire the salesmen who accomplish that.
      - noizejoy 4y ago
        lol - you do have a point there!
  - dekhn 4y ago
    Surely you can see that the P&W approach is only going to discover a subset of problems that will affect the engine in the wild. For example, engines don't run at 100% power during the full flight, and they go through heating and cooling cycles. It also doesn't address failure modes due to manufacturing problems specific to a single engine- for example, turbine blades are manufactured with serial numbers and you can go back and get detailed process and QA for that specific item to understand what went wrong and why it failed and how to not do that during manufacture again. Then there's all sorts of complex failures that aren't addressed by single engines: " Prior to this crash, the probability of a simultaneous failure of all three hydraulic systems was considered as low as one in a billion. However, statistical models did not account for the position of the number-two engine, mounted at the tail close to hydraulic lines, nor the results of fragments released in many directions. Since then, aircraft engine designs have focused on keeping shrapnel from puncturing the cowling or ductwork, increasingly utilizing high-strength composite materials to achieve penetration resistance while keeping the weight low" (I'm not disrespecting P&W- I'm sure they have more tests than just "100% power until it breaks, fix and repeat")
    - WalterBright 4y ago
      > Surely you can see that the P&W approach is only going to discover a subset of problems that will affect the engine in the wild. Keep in mind that my former job was designing gearbox parts for the Boeing 757. This included doing the math, and devising a test plan. I've spent a lot of time on "what could go wrong" scenarios in the real world. Also, when I prepare slides for a coding presentation, the implementation code for a concept gets trimmed way down to what will fit on a slide.
mmazing 4y ago
What is the measure of success though? Successful launch and arrival at the observation point? The 6 month setup period before observations can be made? I contend that success here is a full mission that yields science data over decades of observation. If we cut down the acceptable error rate so we can launch early, how does that impact longevity?
- zmgsabst 4y ago
  I think the question is, imagine we built 10 — and got an outcome like: 2x — failed to deploy, mission lost 4x — early operation error, 6mo lifetime 3x — lower performance, 3 year lifetime 1x — mission successful, 10 year lifetime …do we get a better deal building them with larger faults, at a lower cost? A lot of expense is in developing technologies, assembly tooling, and test rigs — all of which is easier if we don’t need them to be as assured because only 1 in 10 needs to succeed.
  - WalterBright 4y ago
    Right. And if they are launched serially, you've got a chance to correct mistakes in the previous launches. Odds of success increase with every launch.
    - mmazing 4y ago
      Yeah, makes more sense when I think about it like that. Thanks!
prpl 4y ago
What is the timescale you are thinking of for 10% chance of success? 10% first light success is different than 10% 10-year survey success. There’s physically not enough infrastructure (clean rooms, testing facilities, vacuum chambers, etc…) nor skilled manpower available to NASA to be able to build more than one in parallel, so that would add to the cost significantly.
- WalterBright 4y ago
  What's being done with that infrastructure (clean room, testing facilities, vacuum chambers, etc..) right now? Nothing? Of course, the sensible thing to do is to build the test equipment, procedures, train the testers. Then test #1. Then test #2. Then test #3. That's what everybody does. The cost of the Saturn V rocket was amortized over many launch vehicles, despite a host of new technologies that had to be developed for it, despite much of it being hand-built.
  - zabzonk 4y ago
    But the cost of grinding the mirrors?
    - WalterBright 4y ago
      Consider the cost of designing the mirrors, designing a machine to grind the mirrors, testing the machine, and verifying the grind of the mirrors. All that adds to the cost of the first mirror, and adds $0 to the second.
      - zabzonk 4y ago
        But the grinding machine costs (which is significant if you have 10 of them) and the grinding process takes a lot of time and cost (if you have 1 machine x 10 mirrors). I don't think these NASA/JPL guys are quite as dim as you seem to suggest (but you have more info) and they may have done a little cost/benefit analysis before embarking on the project.