Hacker News Reader

Comment by slwvx | Hacker News Reader

If I understand Table 4 [1] correctly, they estimate the cost to be $28.5M, which looks inexpensive for a cutting-edge telescope (BTW, the new html papers on arxiv are nice).

Phys.org [2] has further details.

[1] https://arxiv.org/html/2408.01386v1#S5

[2] https://phys.org/news/2024-08-inexpensive-kilometer-telescop...

View on HN

Has there ever once been a telescope that cost what was initially estimated? I would expect that price to more than double at least

A lot of these estimates seem to be using a projected present price - this is what it costs for the hardware if you build it now - rather than what it would cost if you delay 15 years to slow-roll the program and keep a portion of an academic lab's human resource (and thus a portion of a research university, who take their large cut) alive on shoestring grants in the meantime to run it.

This is one of the many reasons I tend to favor doing big hardware spends on programs to complete them at scale in five years or less. If it's projected at $50M for N science capability, why not built it at $100M for 10N science capability starting yesterday instead of building it at $125M for N science capability starting after fifteen years of being the hobby fixation of a few scientists? There are large economies of scale associated with increasing manufacturing spend in an R&D intensive area, and larger associated with getting the results sooner than later.

The big spend is okay for construction executed by the trades but it doesn’t scale to e.g. instrumentation or software. You can’t integrate in parallel for many of the systems

Building a big camera and an integral field spectrophotometer and a single-objct coronograph+spectrometer so that you have three instruments to hang off of a telescope does not reduce your costs.

Building ten big identical cameras, does reduce your costs. Much of the money is spent designing things, doing engineering overhead, creating novel metrology, documenting all of this, or writing software to control and interpret data off of that novel instrument design. More is spent keeping grad students housed and working; There are at least as many cases where a 10x faster instrument DECREASES their work-hours as cases where it increases their work-hours.

As much I enjoy the quote from Contact about "Why build one when you can build two at twice the price?", and my wife and I have used it with each other as appropriate in some amusingly bizarre situations over the years, it did always bother me a bit that it wasn't true, because building a second one would be cheaper than the first. You only have to decipher the blueprints once, and anything you have to build a thing to build the thing to build the thing, which that structure probably had quite a bit of, you generally can get a second copy of the final output much more cheaply. Not a lot of construction systems involve the total, absolute destruction of everything you used to build the first one. (Although that can certainly be part of a construction process for sufficiently interesting things.)

Also, in astronomy the actual science systems are almost always entirely bespoke hardware, so it can only be bought by spending work-hours in addition to capital.

From a comparable project, Georgia State University's Center for High Angular Resolution Astronomy ([CHARA](https://en.wikipedia.org/wiki/CHARA_array)) interferometric array was constructed for about $15m in 1985. That facility has six one-meter telescopes. That would be about $45M in 2024 dollars.

While I cannot attest for the accuracy of the cost estimate, I can note that the BFT takes several approaches which I think will help control costs. In particular, they are leveraging a multitude of proven technologies, COTS components, and smaller telescopes.

It's also going after bright targets (down to 8 mag). GRAVITY on the VLTI using the UTs (4 × 8.4m) gets down to 17 mag (https://www.eso.org/sci/facilities/paranal/instruments/gravi...).

It wasn’t constructed in 1985, that’s when planning began. It was constructed from 1996 to 2003. At least that’s what your linked Wikipedia article says.

Reply on news.ycombinator.com