Hacker News Reader

> Assuming your figures are correct, you'll get double the data for another 30% cost.

I think 30% is probably a reasonable guess, based on past programs where people have flown 2. You get to reuse design and some fixturing. You get to share some operational costs. But you're not at unit counts where you benefit from mass production techniques and a whole lot of verification and qualification work are still effectively one-offs for each one.

(You save a whole bunch of costs related to making a repeatable program that can turn out hundreds of an item, but more has to be validated/verified for each unit).

The thing is-- what's the marginal value of the additional data (and of the higher priority data arriving earlier)? Would you rather have 2 James Webbs for $13B, or 1 James Webb & some other $3B mission?

(Or, at the outset/original decision making: do you aim for 2 somewhat simpler telescopes or 1 really awesome telescope with the block of money you're given?)

Another poster here remarked that spares would be made. That would indicate the cost of doing so isn't that remarkable.

Lots of engineering models. Some spare parts for parts of assemblies.

Not a lot of complete, flight-capable assemblies.

I'm actually teaching a bunch of high school students to make a cubesat right now, and we are going through costing of:

* Prototypes -> protoflight -> flight, no engineering model

* "Flatsat" electronic engineering model, prototypes -> protoflight -> flight

* Prototypes -> qualification article -> flight article, + "flatsat"

* Prototypes -> qualification article -> flight article + ground engineering model

There's a whole lot of derisk that happens as you go down that list, but the cost and labor increase (the latter is "free" for us but also limited) is substantial.

Double the data may not be worth 30-50% more money to NASA. Going from no data of type X to 1 data of type X is worth a lot more than going from 1 to 2, and so the leap from 1 to 2 might just not be worth it, even if it is 50% cheaper than the leap from 0 to 1.

>Making two adds little cost.

I think you are overextending your experience in a production environment here.

I’ve worked in both, and the type of builds in these aerospace applications still have huge costs in subsequent runs. Hell, even rebuilding an existing component can be prohibitively expensive.

Much of the GSE was likely existing so that’s probably a non-issue.

> huge costs in subsequent runs

Often that's because they have to rebuild and reset the tooling. Within the run, though, the incremental cost should be minimal.

The huge costs may be huge, but not huge compared to the cost of the prototype.

IMO your logic fits with large production runs.

In space applications costs can be exaggerated compared to actual production environments because the risks aren’t mitigated by something like the FAA, meaning they are often mitigated by some downstream process. Besides, a lot of the the designs already include critical spares, so there’s more than a single run, even in a one-off design

So build a copy out of the spares! If they're there, they're already paid for.

Not every part meets the criteria to be a "critical" spare. You don't have a complete set of spares for every component.

Are you suggesting that spares are only produced for the cheap parts?

No. Only for the parts deemed critical.

An example may be a part with a long lead time and no redundancy. If it is found defective during testing, they need a replacement right away and don’t want the schedule to slip while it gets fabricated

A lot of times when you make one of something, there's stuff you would make tooling for if you were going to do more than one, but which instead you do by hand. Sometimes this is even true in software, but not very often, because the advantages of being able to recompile something from source code are so huge.

Even when doing things by hand, making a duplicate alongside the original is not much more time. That happens to me every time I do something by hand.

I think the error in your assumption comes, in part, that there’s is tooling set u for mass production of each part.

A lot of parts are hand-fabricated. The setup may take time but it’s not necessarily like a assembly plant that requires new dies on each design.

This is true for things like motor mounts and pipe flanges, but not for things like optical surfaces and lathe ways. Even in cases where it is true, if you didn't make the duplicate alongside the original, you have to do the setup again to make a duplicate.

Yes, make it alongside the original. I've been saying that all along.

Doesn't help with telescope mirrors, as you likely know from experience.

I think your mindset is stuck in thinking about this as an airplane production environment because that was your experience. And your points make sense in that type of environment. But research aerospace in which JWST operates is a very different scale, operating much more at the margins where the tech is less well understood and takes much longer to develop. Investing billions for redundancy on an unproven system is a big risk (besides the fact that it would likely be designed differently - and better - if started today).

The sunk cost fallacy already drives a lot of these projects...imagine how much worse it would be (and how many other, competing projects wouldn't get funded) if the costs were higher.

> double the data

Yes, but the data has diminishing returns as well.

The real value in a second unit is having a backup in the event of a total failure of the first. But that's a hard sell to the taxpayers who fund these things.

> Yes, but the data has diminishing returns as well.

I find that hard to believe considering there's literally a universe of things to look at. We find surprising things everywhere we look.

$3-5B would go a long way to a new Einstein telescope (ie. gravitational waves detector). It would likey allow for significant sensitivity increases in the existing ones. It could propel the National Ignition Facility or ITER forward. It could be used to build a new Arecibo. It could add an enormous amount of low-frequency telescopes - or other scopes for very large integrated scope arrays.

I'd much rather broaden our view than double it in one narrow band. (Or advance fusion research.)

For reference, the entire NSF budget is about $8B (maaaybe $10B next year!).

That's not true. The potential for surprises gets smaller and smaller the more data you collect. The Webb will produce surprises at first because it is able to make new kinds of observations that were not possible before, but after a while that new data will be used to improve our models and subsequent observations in the same regime will (almost certainly) be less surprising as a result.

> The potential for surprises gets smaller and smaller the more data you collect.

Mathematically, you are quite correct.

But do you really believe that with one little ole' telescope pointed at the freakin' universe you're going to reach a significant point of diminishing returns?

It's like saying if you invent the first microscope, and discover bacteria, why bother with another one?

You might see more new things if you spend the next US$4 billion on a different instrument instead of a carbon-copy Lagrange-point infrared telescope. But you're right that you also might not.

I wonder what the optimal number of James Webb space telescopes would be? The cost per unit as a function of number built has been debated, but what about the marginal utility (benefit) ? If you built two instead of one, you would get lots of benefit from the second - an entirely independent set of scientists could decide what to do with it, and maybe explore entirely different subjects. If you made ten thousand of them though, the last one would be nearly useless, because there aren’t enough smart people to make proper use of it. My guess is that after a couple of them, the marginal utility falls off very quickly.

> It's like saying if you invent the first microscope, and discover bacteria, why bother with another one?

Exactly, why bother with discovering millions of bacterias with 20 copies of the same microscope if for the same cost can build 5 different microscopes allowing discoveries of thousands of fungi, viruses, protozoans, algae, plankton etc.

In astronomy there are multiple bands to observe - gamma rays, x-rays, EUV, UV, visible light, near-IR, far-IR, short/medium/long radio waves. There are also gravitational wave observatories and special instruments like spectrographs and coronagraphs for imaging exoplanets. Every of those observations needs highly specialised instrument but brings a lot of new insight about what is happening in the universe.

Most of Hubble's groundbreaking discoveries came in first 10 years of its operation. Less in next 10 years. Even less in third decade. In fact its final service mission had been cancelled once but was reinstated once it was clear that JWST will be massively delayed and risky. For combined cost of Compton (gamma/high x-rays), Chandra (low x-rays) and Spitzer (infrared) we most likely could have built and launched another Hubble, but there was no point as those 3 generated much more valuable science than another Hubble ever would.

> But do you really believe that with one little ole' telescope pointed at the freakin' universe you're going to reach a significant point of diminishing returns?

I believe that James Webb has some unique capabilities, and a whole lot of overlap with other instruments, too. For things that you can only do in infrared, we have Herschel's history of observations, and VISTA. Sure, JWST is bigger and up at L2, and will be better overall, but VISTA's instruments have some advantages.

Not to mention all the other telescopes and ways we have of studying the universe.

There's all kinds of other things the government has spent $3B on in the intervening development time that I'd trade for another JWST. But it's a bit moot, here: NASA didn't have another $3B to spend. If you made JWST $3B more expensive, you'd not get 2 telescopes instead of 1: you'd get 0, because JWST almost died because of cost overruns.

(And even if NASA had $3B more--- there's a lot of other things that might have been better to do with it than JWSTx2).

It's not quite the same thing, but as an analogy: should we have built two CERNs (edit: LHCs, not CERNs) to "find" the Higgs boson faster? Or was it better just to build one, expecting that it had a good chance of finding the particle and other new physics, and then put future funding toward different/more advanced and capable instruments that will give us a window into even higher energies/entirely new observation spaces, rather than just accelerating our search of the same spaces by low integer multiples?

My understanding is that the JWST opens up new observation spaces, specifically very distant and highly redshifted objects that Hubble couldn't capture. So we should have a lot of data on a new class of objects in fairly short order, and thanks to the cosmological principle we can expect to see similar distributions of the same objects and phenomena no matter what direction we look in. As we gather more data, we will converge on an understanding of these new spaces, and eventually the error bars will shrink to the point where further observation is generally not giving us much new information.

Is it better to build more JWSTs, to accelerate that convergence by low integer multiples and similarly increase the chance that we'll happen to point one at something truly new and "surprising"? Or should we spend our money on bigger and more capable instruments that we know will give us access to entirely new observation spaces that are completely out of reach of the JWST and other extant instruments?

I don't think it would be such an obvious question, if we weren't constantly getting better and better at designing and launching large and complex instruments. As things stand, we can let JWST and its ilk blaze the trail, then follow up with cheaper instruments building on lessons learned and our general technological and economic progress.

It's not quite the same thing, but as an analogy: should we have built two LHCs to "find" the Higgs boson faster?

Depends. Is experimental time at the LHC anywhere near as competitively-rationed as it is for a space telescope like Hubble or Webb?

I don't think Hubble or the JWST have much to do with the question of whether we should have built another LHC to collect data twice as fast, except as an imperfect analogy to help frame the question in casual consideration—i.e. the reverse of the analogy I made in my comment.

In fact, answering that question should come down to whether we expect the LHC to realize the majority of important observations "in range" of its instruments within its projected lifetime, and also in time to make useful contributions to the selection and design of future experiments.

To answer your question directly: yes, AFAIK the LHC is fully utilized; when it's not running, it's down for maintenance or upgrades, and its observations are (or have been, in its active periods) in high demand. But it's far from clear that building another one would yield a good return on investment, especially given the opportunity cost of diverting funding from future experiments.

I think it's more that this argument can be continued ad infinitum. Yes the JWST had significant cost overruns, but in the design phase these types of projects necessarily are looked at as how to best utilize a fixed size grant of money. At the end of the day the decision got made to make a better single telescope than to make two simpler telescopes.

I wouldn't be surprised though if we start to see clever design proposals coming down the pipeline, like several cheaper telescopes, swarm designs for giant radio telescope arrays, and even amateur designed and operated space telescopes. Remember that the JWST started it's design phase back in 1996, and the economics of space launches have changed considerably since then.

I remember having manuals printed in the 1980s. The first manual cost $1000 to print, the second one $1. This was with camera-ready copy. Never mind all the cost of writing the manual, proofing it, and formatting it.

When I worked at Boeing, the first forging of a part cost $250,000. The next, just a handful of dollars.

At Boeing, the first airplane gets a ton of testing, as the design is being tested. Airplanes #2 and on only get tested to verify it was built according to the design, at a tiny fraction of the cost of testing #1.

You still don't choose forging as a process to make quantity 2 (unless you absolutely need forged parts for strength). Instead, you're milling two parts from billet.

Milling 2 parts from billet is cheaper per unit than milling 1 (some shared setup and programming costs), but it's the same amount of raw material and basically the same amount of operator time.

You're right that a one-off forging is so expensive that a hogout will be used instead. My point was how expensive one-offs can be compared to multiple ones, and I used extreme examples as illustrative.

Even in custom machine work, the cost is in the setup. A machinist can make two identical cuts on two parts for not much more cost than one cut on one part.

>you'll get double the data for another 30% cost.

Sending just another similar telescope that provides more same typed data (wavelength, angular resolution) is probably not worth it. Spending the same sum for different type of telescope would be better use of the money.

Hubble, Hershel and Webb were are made for different wavelengths, they are complementary.

The Extremely Large Telescope (ELT) is ready around 2027 and it will be the next revolution. 0.005 arc-seconds compared to 0.1 arc-seconds of JWST. (978 m2 vs 25.4 m2 collecting area)

Doubling the data will also double the processing requirements for that data. The observatory is just the remote part of this operation. The data doesn't do squat unless stored, analyzed and interpreted by expensive professionals on expensive equipment.

It's like saying if you're gonna build a chip fab, why not build two while you're at it? Well because the building isn't the operation.

Why do I have to do all the thinking?

1. Grad students and computers are cheap.

2. Crowd source it. Make the data (and the programs that process it) freely available. Let anyone who wants to analyze it - if they find something cool, they can be famous. They'll do it for free.

3. Processing doesn't have to be done in real time. There's no problem with taking 5 years to analyze 1 year of data.

All of the above are already utilized, to varying degrees of success, and lack thereof. I believe in this stuff but let's be real, we're exploding fossil fuel to propel a camera into space. Baby steps for global consciousness.

>I do have some experience with this. I worked for 3 years on the design of the 757. Thousands of engineers spending maybe 5 years on it.

Just wanted to say that's really awesome. The 757 is by far my favorite Boeing jet of all time. They are so overpowered it feels like taking off in a fighter jet. Delta still flies them from LAX to HNL, and it's always so much better than cramming into a 737.

I am always pleased to be boarding a 757! I was sad when the front page of the Seattle Times showed one being cut up for scrap.

> Assuming your figures are correct, you'll get double the data for another 30% cost.

The debate over the cost to build another is largely moot. You'd struggle to find the people to do it for many reasons. The talent is even scarcer than the money.

Agree. A replica wouldn’t have to be sent to L2 for that matter. Pretty sure a lot of science could be done with even a slow trajectory leaving the solar system, if orbiting something else is infeasible.

As I understood the engineering of the JWST requires it to be at L2 so it's in permanent shadow of the sun and can keep the very low temperatures necessary for good infrared image quality.

It has solar panels, so it can't be in shadow. IIUC the advantage of earth-sun L2 is it's away from the light and shadow of the earth and moon, while being reachable by radio year round.

I didn't check for sources again earlier, but this NASA article [1] explains it.

Yea shadow was the wrong word and doesn't actually apply because of the relative size of the sun and earth. However the L2 point is by definition in a straight line Sun -> Earth -> L2, so if the sun would be a single point light source that would block the sun and place the JWST in the permanent shadow thrown by the earth. The point is actually that the sunscreen is always facing earth and sun at the same time to block/reflect a maximum of heat.

[1] https://jwst.nasa.gov/content/about/orbit.html

It's orbiting L2, not at it. It will have basically constant sunshine, which is good - solar panels provide all its power. (no RTG)

It has constant sunshine on one side, and constant darkness on the other.