It drives a hell of a lot. It doesn’t matter how big your satellite is; at some point you have to choose, for example, which solar cells to get. Do you get the top-of-the line cherry-picked triple-junction cells with 32% efficiency? Or do you get some more modest cells that are a few ticks less efficient, and accept the additional mass?
Likewise, if you need electric thruster propellant, you can use xenon–extremely expensive, but with high efficiency due to the low ionization needs–or you can use krypton, which is far more abundant but requires more power for the same thrust? Again, this is a simple mass-cost tradeoff where the answer depends on your launch costs.
Or in the case of the JWST, why the beryllium mirrors? That wasn’t the only possible choice. No rational person would use beryllium for anything unless it was an absolute requirement–it’s just absurdly expensive, not to mention toxic. But the JWST was at the absolute limits of its mass and volume constraints and so beryllium it is (plus every other expensive design choice).
The Starlink satellites are showing the right path forward (note that I’m saying nothing about whether it’ll ultimately be profitable; I’m speaking only of the satellite design). These are smallsats, but aren’t cubesats–they’re about 300 kg, and a few meters on a side. They’re designed for maximum cost effectiveness, which means every mass trade they make is dependent on the launch costs. The v2 satellites take the same approach and are no longer even in the category of smallsat.
Obviously a space telescope will never reach the volume of a megaconstellation, but there are many intermediate points between building a single perfect gem of a satellite vs. pumping them out like a car factory. The experience curve is real, not to mention sharing the design costs.
The point is that we shouldn’t have such large, complex projects. Decrease the scope, decrease the complexity, and increase the volume. I certainly don’t think that we should accept a 90% failure rate, but as k9bfriender pointed out, every 9 you add to the success rate makes things vastly more expensive. Of course it would be better to have 10 telescopes with one failure, even if they were individually a bit less capable than the JWST. We might still get unlucky and have a micrometeorite hit that takes out something important on the JWST. More scopes = more redundancy.
The funding model is a problem. Big projects get attention, and having a single point of focus seems to help people concentrate. It is also easier to defend cost overruns when all of your eggs are in one basket, basically using the sunk cost fallacy to your advantage. The JWST was going to happen no matter what; it could have cost $20B and it still would have flown eventually. An array of relatively cheap projects are too easy to cancel, especially if anything else is demanding attention. It sucks that the NASA funding model is designed to encourage this, but there’s no obvious way around it.
It’s a lot better than a billion-dollar spacecraft that’s guaranteed to work. Especially since a large fraction of a spacecraft’s costs are in R&D, so the marginal cost is far lower.
If it’s not ultimately because of launch costs, why do you think that spacecraft are so much more expensive than equivalent ground-based devices? The largest Earth-based telescope has a mirror twice the area of Webb’s, in a less-forgiving wavelength range, and it cost about 1% the price of the Webb. It’s not just the harsh environment of space, because plenty of Earthly environments are far harsher. The stresses of launch? Easy, build the thing more robustly (which is easy when you don’t care about the weight).
Falcon Heavy is already pretty cheap. Of course, it wasn’t available when the JWST was first proposed, so I can’t exactly blame them for not designing to that, but it’s been around for years and we should be seeing proposals for inexpensive Hubble-class scopes based on today’s costs.
But even aside from all that, I consider the JWST approach to be wrongheaded. Why was there no test mission? They could have gained a lot of confidence in their designs in an actual space environment. The JWST had something like 300 deployment steps, and each one had to be tested until it was absolutely perfect (despite being on the ground, where it’s impossible to simulate perfectly). They could also have tried out multiple approaches for some of the more complex parts, like the choice of sunshield design. Even with expensive late 90s launch costs, I can’t believe that an actual test in space would not have saved some of the $10B total costs by retiring risk more effectively.
Here’s some back-of-the-envelope math for our readers. Hubble’s 2.4-meter glass mirror weighs 828 kilograms. At that weight per square meter, an 8-meter diameter – that is, 10 times the surface area – would have clocked in at over 10 tons.
So we looked at silicon carbide, at composite materials with a very thin layer of glass on top, and at beryllium, which ended up being the winner. We aimed for a tenfold reduction in mass density. [And succeeded; Webb’s beryllium mirror weighs about 20 kilograms per segment.] We had a competition that went on for several years. We pumped in probably about $50 million into figuring out the technologies.
So in fact they did have an explicit goal of increasing the mass density, apparently with little regard for cost. But mass density is a nonsense goal. It doesn’t give you sharper pictures or allow gathering more light. Aperture gives you that. Of course, if you’re mass constrained, or if upmass is expensive, then improving mass density might be the only way to improve aperture, but another way is to just have cheap lift.
As noted in the article, a Hubble-style glass mirror would have only weighed 10 tons. Sort of a lot for most rockets. But no big deal for heavy lift.
More evidence that I am not actually insane–a Sky & Telescope blurb from 1998 (BTW, Jonathan McDowell is still doing great stuff on Twitter and elsewhere):
So they did want to fly demo missions. Those would have been a great idea and retired a huge amount of risk. Anyone want to bet that they were cancelled for cost reasons?
A space observatory or other spacecraft with the capabilities of JWST are going to be a “large, complex projects” for the foreseeable future simply because it is pushing the limits of precision and capability in manufacturing. JWST is expensive (arguably far more expensive than it should have been because it was not given funding priority to get the mission launched in a timely fashion). Again, even if there were funding to produce these in serial fashion, there just isn’t the technical throughput to design, build, and deploy these things by the dozens to achieve some hypothetical economy of. scale. They are incredibly complex machines and there are literally millions of hundreds of thousands of person-years of effort that has gone into JWST, and there is neither the trained technical base nor facilities to expand the pipeline to make spacecraft of this class by even an order of magnitude. Nor could you replace a JWST by a hundred smallsats; it does a very specialized thing that requires particular capabilities that can’t just be produced by a bunch of cheaper satellites ganged together.
This discussion is actually kind of ironic because when it comes to crewed exploration of Mars I point out that we could pepper the planet from pole to pole with Perseverance-type rovers for far less than the cost of a single crewed mission I get shouted down by the space enthusiast community claiming that human astronauts could cover far more ground and do more work in a month than these “slow and dumb” machines could do in the years that they would operate, notwithstanding the incredible difficulty and time lost in supporting human explorers both on planet and for the transit there and back. Of course, we could not afford to build hundreds of Mars rovers any more than we could build a continuous production like of large space observatories because of the technical requirements and limitation of available technical labor and facilities but the point remains that practical space science isn’t just about funding but the actual capability to produce and send spacecraft (or people) into space, and the surface-to-orbit transportation aspect is one small segment that isn’t even that much of a cost or capability driver for large programs.
You’ve hit on a salient point here; NASA tends to focus on big projects because they are less likely to get cut even in lean years, and part of the problem with cost growth on JWST was cutting down funding to some minimum and stretching out schedule which increased the overall cost by somewhere around US$ 2-3 B, while smaller projects get cut, often to pay for the crewed space program that produces very little in useful science beyond studies of human physiology in freefall and orbital space that are pretty much telling us that human beings do not do well in the space environment. It would be great to have secure funding along different levels of programs but NASA is ultimately at the mercy of the budget that Congress allocates, and despite the hue and cry that NASA doesn’t manage funding well they’ve often done amazing work on a relatively shoestring budget for uncrewed exploration.
“Big Science” projects get a lot of criticism for eating up the funding from useful smaller programs, and much of that criticism is valid as it stands, but there are certain things that can only be done with “Big Science”. We would not have statistically definitive evidence the Higgs Boson without the LHC, nor produce pentaquark particles with the Tevatron or other lower energy devices. Certain types of scientific work require large systems and massive budgets to be successful, and the answer isn’t to stop supporting “Big Science” but rather to make sure there are also dependable allocations for smaller but still useful scientific research.
Except no one is ever going to fund ten missions with the expectation that nine of them are likely to fail even if it were cumulatively cheaper than a single mission; not only is managerial psychology not going to accept failure after failure, but again there is the opportunity cost that comes with repeated failures vice having a successful mission and observatory that can be relied upon with high confidence.
The problem isn’t just “the harsh environment of space” (although it is quite challenging, and in particular the thermal issues that arise from only being able to cool through radiation) or “the stresses of launch” (again, a more significant challenge than you would imagine even with a supposedly “robust” design because a spacecraft with adverse modal dynamics can cause a launch vehicle to become unstable or break regardless of how much payload capacity it has) but the fact that a space observatory can’t really be serviced or maintained like a ground-based telescope, has to be able to operate with virtual autonomy save for communications, and doesn’t have any kind of stable foundation which means that it relies on extremely precise inertial systems and flywheels to maintain precise orientation versus the many tons of concrete and anchors into bedrock that keep terrestrial observatories from wantonly floating away.
It is certainly not the case that you could take a ground-based observatory design and just launch it into space on a sufficiently capable rocket with any expectation that it would work; space-based systems have entirely different design requirements and drivers, and again, the cost of failure and impact upon program schedules is far more of a driver than the cost of launch such that even if the cost of a replacement launch is negligible the loss of observing time, the opportunity cost of assigning technical labor and facilities to build spacecraft and run missions with low probability of success, and the human lack of tolerance for repeated (even if expected) failures just don’t make it viable to accept a mission with a 10% success rate, regardless of how much it might reduce cost.
I thought a big part of the choice to go with beryllium was its thermal properties. It does not change shape (much) as temperatures fluctuate and temperature was a big deal with the JWST. It is also pretty strong.
That beryllium is relatively light was a bonus.
(And even though beryllium does not change shape much with temperature it still changes shape so the mirrors were intentionally made with errors that would correct themselves as the mirrors cooled in space…but doing that required the relatively small changes beryllium afforded versus some other material).
Only because it wasn’t built. We know the general expertise is there, because one of them was built. It’s virtually axiomatic that tooling, setup, and other one-time costs dominate when manufacturing a singleton. Building a second is almost free, which is one reason why NASA almost always builds spares (which are useful, but not so useful that they’d warrant doubling the price). So… build a dozen spares instead.
Obviously there’s little point in building a dozen JWSTs as such. But, given something like Starship, what could be done is to build an 8m-class telescope platform, produced in moderate quantities, with a modular sensor system. JWST has four main instruments. It would be more useful to have four telescopes, each with one of those instruments. The complexity of each one would be lower, it would be more resilient to failure (not least because one of them would be launched first, and any lessons could be folded into subsequent units), and by producing things on a more regular schedule it’s likely that the costs would be lower.
I want humans on Mars because I want humans on Mars. I don’t want NASA to spend a trillion dollars on it; I want it to be done cheaply or not at all. And if it can be done cheaply, it implies low costs to landing things on Mars, which implies cheap and abundant robotic missions. I don’t see any contradiction here, though I couldn’t speak for anyone else.
I agree, but I disagree that JWST had to be one of them. There’s no obvious way to make particle accelerators cheaper. Kilometers of tunnel filled with superconductors and hard vacuum pipe will always be expensive. But an 8m class telescope is not obviously expensive, even if it has to live in space.
And hell, the JWST was supposed to be part of NASA’s “faster, better, cheaper” strategy! It was supposed to be a 9-figure project, not an 11-figure project. It’s just that they made the wrong cost decision at every turn (though possibly the right decision when it comes to preserving funding).
I’ve seen it phrased this way, but as best I can tell it always comes down to “it has great thermal properties for the mass”. It may well have been that to reach their mass targets, beryllium was the optimal choice. Ultra-low-expansion glass actually has a much lower expansion rate than beryllium, but much worse thermal conductivity. So if you’re going in with the idea of ultra-thin mirrors, you need the ability to spread heat fluctuations across the surface with a low cross-section of material, and that means decent thermal conductivity. Metals are great for that.
But you can also achieve that with lots of mass. The mass constraint drove the design, and the design drove the choice of materials. I’d say it’s kinda backwards reasoning to say that it’s about the thermal properties instead of the mass.
There’s a lot of good information in this document:
Beryllium has excellent stiffness, for instance. But again, that’s for the weight (ratio of elastic modulus to density).
I don’t know what portion of the $10B cost of JWST went to the mirrors. Most likely, you could easily argue that the additional costs for beryllium were only a tiny fraction of the total. But it’s a form of begging the question. Given that the telescope was super expensive, it doesn’t matter that some individual component was also super expensive. But if you were trying to produce a low-cost scope, you’d never make that choice.
If mass limit was the most important reason JWST was massively over budget and late, they could have simply decided to tell ESA, thanks but no thanks, and launched on something bigger, say a Delta IV Heavy. Roughly twice the mass to orbit of Ariane 5. Cost impost of the launch, $350 million, plus the political damage.
The difficulty is being able to see far enough ahead. The JWST was a management disaster. One might sheet home getting on for half the cost overrun to that rather than technological issues. When the telescope was being planned Elon was just out of short pants, and even now Starship still hasn’t flown. When someone suggested that SpaceX should fly Webb on a Falcon 9 Heavy even Elon was pretty clear he would not go near it. Just too much risk.
Mass budget was a driver for some of the systems design management. This paper makes for interesting reading. But even then it wasn’t extraordinary. Just drove the design regime into a particular interesting structure. Note that it was published in 2010.
Of note, there were a lot of scale prototypes made and tested. Just not launched into orbit. It isn’t clear that you get value for money from testing stuff on orbit. If you can validate the risky tech on the ground you will get vastly more value for your money. Minimally you save the launch costs, which are not exactly small. Also you can test a lot more iterations. Your in space demonstrator has almost the same risk as the final article. Especially risk for delays in the programme if it fails.
The trouble with any situation where you loosen the constraints is that Parkinson’s Law takes over. The JWST was designed to fit into the then available constraints. If they had decided to launch on a Delta IV Heavy the design would have just got more ambitious and fitted just as many bleeding edge ideas into the new box. When design started everyone though it could be done for $1B. Tell them that they have 100 tons and the design will not be a battleship robust version of the JWST, but remain a bleeding edge pipe-dream that costs as much money as they think they can liberate. Looking back on the JWST with 20:20 hindsight is easy. But starting out with what they knew then, a very different matter.
The difficulty with big ticket science is that nothing is clear. By the nature of it, you don’t know at least half of the science that might be done. When Hubble was in the offing there was a great hue and cry from the astronomers bemoaning that all that money could have funded a lot of ground based telescopes and research grants. A lot of that opposition dried up when the science results started to come in. As it turned out, ground based research took off in a manner they could not have predicted, with funding probably spurred on by public interest in astronomy as a result of HST results. Selling the JWST as the new Hubble was always a good bet.
The book The Hubble Wars is a very interesting fly on the wall account of the tensions in the early days of the HST. Back then the idea that pretty pictures for public consumption was even worth considering was quite foreign. The kickback, and realisation of just how important public engagement is, was a scales falling from the eyes moment. You can be sure that the lesson learnt has not been forgotten, and we can be assured of jaw dropping eye candy images from the JWST in the future. If nothing else, this invests in the future, and support for the next big ticket telescope.
The problem of picking winners in science, who gets funding and who gets privileged access to facilities is impossible to do in any sensible manner. Track record helps, but even that is to a large extent a self fulfilling prediction. The JWST clearly has a lot of good science to do, and it is very hard to dispute the goals. Science for the dollar is of course a bit hard to quantify. But how do you measure that? How many papers does it have to enable publication of? That is of course one of the worst metrics possible. But it also how a large fraction of science funding is done. That is a whole different conversation.
Just because you say it is “virtually axiomatic that tooling, setup, and other one-time costs dominate” doesn’t make it so. I don’t know what experience—if any—you have on building up, testing, and processing large spacecraft but there is an enormous amount of highly skilled ‘touch labor’ that goes into the process. I have a friend and former coworker who worked on JWST and who spent months at a time working EWW—averaging 60+ hours a week—just to try to maintain schedule, and they were perennially understaffed because the people who were skilled and temperamentally suited to this kind of work are always in short supply.
The notion that the building up a “second is almost free” is so absurd as to taken as satire. The only thing that is really reduced for subsequent units is the non-recurring engineering (NRE), and perhaps a reduction on speciality components that can be built serially, but there is not going to be sufficient number of units to get any kind of volume discount, and even “tooling and setup” aren’t really saved because in order to build up units in a timely fashion there would need to be parallel build lines. These aren’t CubeSats that you can build up in a few weeks; just installing and testing the optical sensors and their power, adjustment, and thermal management system is the work of many months of dedicated labor that is not amenable to the tolerances of a production line process, and there are inevitably engineering changes and rework during any build that make each unit somewhat unique.
This is just absolutely and objectively not true. “Faster, Better, Cheaper” was a mantra from former NASA Director Dan Goldin associated with the Discovery Program within the Planetary Missions Program Offices managed out of Marshall Flight Center. The JWST is part of the Great Observatories Program out of Goddard Space Flight Center in partnership with the European Space Agency and Canadian Space Agency;’ it always intended as a flagship-class program that even in the conceptual stage had a billion dollar price tag that grew to US$6.5B at CDR to achieve the desired capability. Much of the the post-CDR cost growth had to do with funding delays and budget limitations that increased total cost and was outside of NASA purview to control. I invite you to detail how NASA “made the wrong cost decision at every turn” with regard to the program, because while their are definitely criticisms to be made about the management of any large scale program, I’m curious whether you know enough about this program to even make such an assessment.
From Space Telescopes: by Neil English (sorry, I can 't seem to do a proper copy & paste of the text):
No, it wasn’t one of the super-cheap almost disposable missions. But it was pushed as a low-cost telescope of <$1B.
Like I said, there’s arguably a certain value in a ballooning price in that it makes a flagship program harder to cancel. But in terms of cost management, I find it hard to argue that a >10x increase isn’t nuts.
Was there massive scope creep along the way, partly justifying the increased price with additional functionality? Sure. That’s still a sign of bad management.
In 1996 the then Next Generation Space Telescope was just a concept that required development, not a built-from-off-the-shelf components. On such development programs it is accepted that a program cost estimate is only guesswork until it gets to the PDR stage because of all of the unknowns in how much development is even needed in achieving requirements for a system of novel capability. This is just the reality of developing complex systems with technologies that are at or beyond the state of the art, but you seem to want to make it seem like some act of perfidy by NASA in gaming the system for some…purpose.
I swear, interacting with space enthusiasts about the practicalities of spaceflight and spacecraft development is like talking about cardiology with someone whose medical experience is playing Operation!.
You’ve yet to explain how drastically cheaper launch prices wouldn’t make everything else cheaper, or why, other than launch costs, space hardware is more expensive than equivalent Earthbound hardware.
You’re just arguing that NASA is bad at estimating costs. Well, yes. Being bad at estimating costs is a distinct component of being bad at project management.
TRW was awarded the prime JWST contract at $825M. That was well past those early design studies. The PDR for the sun shield was completed in 2008. It seems distinctly odd that a preliminary design review would only be completed 6 years after the initial contract is awarded, but there you go.
At any rate, much of my point here is that we would be better served by not designing these beyond-state-of-the-art systems, and instead building systems that are, if not quite off-the-shelf, at least at a slightly higher TRL. And if the goal is to raise the TRL for useful technologies, use demonstrator missions instead of making your flagship dependent on them.
So when you argue that these beyond-state-of-the-art missions inevitably have cost inflation, and that’s just the way things are, then you’re simply reinforcing my point. Stop doing it that way.
