LSST Camera: largest camera for astronomy | Hacker News Reader

LSST Camera: largest camera for astronomy

www6.slac.stanford.edu

75 pointsdefrost22 comments

A project I worked on donated compute cycles to LSST- it was a full scientific-grade ray tracer that takes star catalogs and uses them to render the image on the camera. It worked well enough, and early enough, that it caught a design error in the physical telescope. That work was done 10 years ago, I think they are wrapping up LSST's assembly now and they will have imaging going next year.

It's very exciting because it's a survey which produces statistical quantities of information about a wide range of objects, and also it's great the US and other governments were willing to fund something at this scope for so long.

Do you recall the name of the software? I am aware only of PhoSim (https://www.phosim.org/)

3200 megapixels custom CCD array cooled to about -100 °C sounds amazing.

The 3.5 degree FoV is around 700 mm focal length equivalent on a full frame camera which makes this relatively wide field by telescope standards, allowing it to capture more of the sky per shot. But the array is 64 cm wide so the actual focal length is around 10 meters.

It has the greatest light-gathering ability (product of field of view * aperture) of any telescope yet. That's the key figure of merit for a surveying telescope! (As I understand—I'm not a domain expert).

- "Combined with its large aperture (and thus light-collecting ability), this will give it a spectacularly large etendue of 319 m²⋅degree².[6] This is more than three times the etendue of the largest-view existing telescopes, the Subaru Telescope with its Hyper Suprime Camera[36] and Pan-STARRS, and more than an order of magnitude better than most large telescopes.[37]"

https://en.wikipedia.org/wiki/Vera_C._Rubin_Observatory

Wow yes, 10.31 m (f/1.23) overall (optical diameter of 8.360 m). That's insane.

Thanks. It boggles my mind that an article like this about a scientific camera can fail to list the focal length and aperture, which are basically the two most important numbers for any camera.

Maybe before you get too boggled, consider that the original article is about a scientific instrument designed to be attached to a telescope. The aperture and focal length given in the other source are properties of the telescope, not the instrument.

It's not uncommon to discuss CCD cameras for large telescopes without making any mention of the focal length or aperture of the telescope. See, for example: https://www.eso.org/public/teles-instr/lasilla/ntt/susi2/

Here's a page for another imaging device which lists the aperture and f-number of the telescope—and then gives a separate f-number for the instrument, without stating the instrument's aperture: https://www.eso.org/sci/facilities/lasilla/instruments/wfi/o...

So I don't think the LSST media department dropped the ball on this one. The numbers (focal length and aperture), which you understandably think of as being essential info for any camera, just aren't as relevant here, partly because comparison with other options isn't in the forefront of the scientists minds (the camera is totally bespoke and there are many ways in which it is tailored to the telescope, which constrain its design, including in those two respects).

Focal length and aperture of the LSST camera aren't mentioned here, either https://www.lsst.org/about/camera

Yeah haha I mean even when you buy a consumer camera like, say, a Sony a7 or whatever, it doesn't come with the lens.

Focal length and aperture literally mean nothing for this instrument (camera). Telescopes are multi instrument and it’s the telescope plus instrument specific mounting configs that determine these. (Focal length is actually pointless except to make sure you design the light path correctly).

Aperture is talked about in the light collecting area of the mirror, and often summarized to its diameter. 8m class telescope, is a telescope with an approximately 8m diameter mirror.

Hubble is a little over 1m.

Other things matter way more for a telescope and are much more interesting.

I remember the original Google announcement where it was announced that Google would help with the huge storage requirements for the project, I just dug up the YouTube announcement. Pretty amazed that this video is 17 years old now, a testament to how long these projects are in development for... this 2007 video wasn't even when it began.

https://www.youtube.com/watch?v=thPlpDcaewo

so cool

I wonder if there's more information on the logistics of transporting an instrument of that size and complexity from California to a mountaintop in Chile. They say it's the size of a small car, but the support structures certainly look bigger than that!

Packing, cleanliness, dealing with drops/shocks, loading, unloading.... are they flying it down?

...significantly hampered by low-earth-orbit microsat constellations like Starlink. It boggles the mind that there are half a million microsats planned for launch in the next ten years. Starlink is already causing issues for launches, cutting down launch windows and even blocking ones that had been open because Starlink moves a satellite without clearing it with anyone first.

https://www.lsst.org/content/lsst-statement-regarding-increa...

TLDR: the telescope's field of view means that it will be nearly impossible for them to find a "clear" patch of sky, the brightness off satellites saturates the CCD, which in turn causes crosstalk during readout of the CCDs, and even just masking the primary streak is difficult. I'm sure there are optics issues as well from such bright objects, in the field of view and not.

> During the nominal 30-second visit to a sky patch, satellites in 400-600km LEO orbits typically move about 15 degrees across the sky (about four times the diameter of Rubin Observatory’s field of view), and are visible a few hours after sunset and before sunrise. With 400,000 satellites orbiting Earth, tens of thousands of satellites would be visible above the horizon and it would be difficult to find a circle of 9.6 square degrees anywhere on the sky that does not contain satellite streaks. Simulations of the LSST observing cadence and the full SpaceX satellite constellation show that as many as 30% of all LSST images would contain at least one Starlink satellite trail. With the planned constellations of 400,000 satellites at 400-600 km, all images in twilight will contain streaks. The OneWeb constellation at 1200 km will be visible all night long in Chilean summer. Measurements of the brightness of the current LEO satellites in their final orbits indicate that these trails would cause residual artifacts in the reduced data. If these LEO satellites can be darkened to 7th magnitude, then a new instrument signature removal algorithm can remove some of the residual artifacts. This is challenging due to apparent non-linear crosstalk between the 16 channels on each of the 189 CCDs, the cause of which is still under study. The bright main satellite trail would still be present, potentially creating bogus alerts and systematics at low surface brightness. Masking of these trails is not 100% perfect. This is a challenge for science data analysis, adding potentially significant effort.

> If these LEO satellites can be darkened to 7th magnitude, then a new instrument signature removal algorithm can remove some of the residual artifacts.

The current generation of Starlink satellites is already above the 7th magnitude, so it no longer saturates the CCD. [1] Of course, darkening them further would always be good.

[1] http://arxiv.org/abs/2306.06657

There's however more considerations than just saturated pixels,

https://www.lsst.org/content/lsst-statement-regarding-increa... ("Vera C. Rubin Observatory – Impact of Satellite Constellations")

That's true, but this article was also from before the Starlink Generation 2 Mini satellites with magnitude > 7 came out.

> Darkening satellites to 7th magnitude would simplify removal of some artifacts in LSST images, but there is no guarantee most of the satellites will be limited in brightness to fainter than 7th magnitude.

I'm curious if they were indeed able to implement the artifact removal or if it remains challenging even then.

“Removal” here means “masking”—the pixels under the bright streaks will always be scientifically unusable because the streaks inflate the noise irrevocably. (You can’t see stars in the daytime by “removing” the sun.)

The 7th magnitude limit just minimizes the streak signal cross-talk through the rest of the camera. It also means the satellites are invisible to the naked eye.

Looks like there was concern about this all the way back in 2016: https://twitter.com/lsstepo/status/804412225926463488

>It boggles the mind that there are half a million microsats planned for launch in the next ten years.

Great for Space X, they're creating a need for more space telescopes..

It is fortunate that storing 15TB per night is now only $300, well before supporting hardware and mirroring.

Imagine having to store that much every day a decade ago.

They've got fast fiber to get the data off the mountain and into multiple data processing centers in near real-time including alerts when something interesting happens.

"The nightly pipelines are based on image subtraction, a process that highlights differences between two exposures of the same field, and are designed to rapidly detect interesting transient events in the image stream and send out alerts to the community within 60 seconds of completing the image readout. "

https://www.lsst.org/about/dm

So $600 a decade ago?

[1] https://ourworldindata.org/grapher/historical-cost-of-comput...

well we don’t have to imagine, there are meeting notes!

Reply on news.ycombinator.com