Fermilab Builds 500-Megapixel Camera

heyitsme writes "Fermilab, a U.S. Department of Energy research lab, is part of a collaboration on an experiment to measure the properties of dark energy. The Dark Energy Survey would measure the history of the expansion rate of the universe more precisely than ever before, using the largest camera ever built with Charge Coupled Devices (CCD). The 500 megapixel Dark Energy Camera (DECam) would be placed on an existing 4-meter telescope located in north-central Chile at the National Optical Astronomy Observatory's Cerro Tololo Inter-American Observatory. The DECam together with the CTIO 4-meter telescope will allow for a survey of 15 percent of the sky to light levels faint enough to measure the colors of galaxies at redshift one."

Comments

Filesize?

[FSWKU]

I would hate to see how much space one frame from this thing takes up...

Re:Filesize?

[Anonymous Coward]

The detector actually contains roughly 503,316,480 raw pixels, this amounts to slightly less than 500,000,000 effective pixels after initial processing.

The data will, of course, be stored directly to a large SAN storage system, probably from EMC or Hitachi.

The detector should generate single frame images of roughly 1.7G prior to post-processing, and roughly 700M single-frame image files after processing to TIFF or PNG format.

Re:Filesize?

[hyc]

Interesting, the numbers you show don't appear anywhere in the article. Instead, I see:

At the heart of the DECam are 60 rectangular (2k x 4k) CCDs, each with 8 million 15-micron pixels. The CCDs, developed at Lawrence Berkeley National Laboratory, are over five times more sensitive at near-infrared wavelengths than conventional CCDs currently used for astronomy.

60 CCDs at 8 million pixels each for a total of only 480 million pixels. There's no mention of color filtration so grayscale is a safe assumption. There's also no mention of resolution but 16 bits sounds good as a guess.

So 960 million bytes per frame, which is only 915.5MB (1M = 2^20).

Re:Filesize?

[BWJones]

I have heard that some sites are planning a huge SneakerNet (or, rather, JumboNet? CessnaNet?) and hoping to fly out stacks of DAT tapes of the unreduced data back "home."

So, we here in academia are a bit spoiled in terms of bandwidth. However, companies and some in academics have to pay for lots of bits and bytes and are thus interested in costs to move these sorts of data. I was talking with Jim Gray a couple of weeks ago and he was telling me that a recent study revealed some of the true costs of moving lots of data. For instance: Lets say you are trying to move a terabyte of data from London to Los Angeles. It turns out it is cheaper (and faster) to put it on magnetic storage and fly it from London to Los Angeles than it is to try and move it over the Internet.

I'm glad

[toddhunter]

That they were able to save money by using an existing telescope.
Because the compact flash cards for this thing cannot be cheap.

500 megapixel??

[fodi]

I dont EVER want to be photographed in that much detail !!

why you need 500 Mpx

[sdedeo]

(Not an expert, they're all fast asleep right now.)

One of the things Fermilab is trying to do is get a measurement of the so called weak lensing effect. Matter distorts spacetime, and light is thus bent as it passes nearby a big cluster. This is gravitational lensing.

Famously, it is seen as "strong" lensing -- when the source is very close on the sky to the cluster, and the light gets bent enough that there are multiple images. Nobody really believed it could happen, but then in the last decade or so it's become an accepted and popular thing to play with and observe.

Weak lensing is when there are no multiple images, and instead only a slight distortion. Much harder to see and measure -- you basically look for a whole bunch of galaxies that are slightly distorted.

That means you need a very wide field of view -- to get enough galaxies quickly enough -- but also a very good resolution -- to be able to measure the slight distortions. Hence the need for such an insane[ly cool] device.

Why go through all this trouble? Well, weak lensing is one of the view ways to measure all the matter in the universe on very large scales. Because nearly all the matter is supposed to be invisible, in the past people have used various "tracers" that we can see. But there's a huge amount of debate as to how good the various tracers are, and, of course, you need a direct measurement to be sure you're not off in la-la land.

Weak lensing measures it all because all matter, regardless of how bright it is, bends spacetime in the same fashion. So, if you can get a good weak lensing measurement, you can theoretically create an unbiased map of the matter distribution. No need to cross your fingers and hope that some tracer is behaving properly.

It all fits into dark energy because dark energy is supposed to alter the extent to which matter can cluster (roughly speaking, dark energy behaves like antigravity, and pushes things apart, stopping them from falling together.)

Of course, weak lensing is just one of the things this guy is meant to do -- there are lots of other neat things that hopefully someone more awake than I can describe.

Re:Just Think...

[Zog The Undeniable]

Probably not. Digital zoom is useless because it just magnifies part of the picture without increasing resolution. It may be useful for video (especially if the CCD has a higher resolution than TV) but it's essentially a "marketing feature" for still photography.

Anyway, with 500MP you're going to be severely limited by the resolving power of the lens. It's quite difficult to get even 100 line pairs per mm with the best 35mm photographic lenses (lenses for larger formats tend to be much worse because it's harder to maintain accuracy over a large glass area, plus it's not as necessary with lower enlargement factors). A 500MP sensor needs a pretty exceptional telescope in front of it.

Finally!!!

[Eberlin]

Finally something that has higher system requirements than Longhorn!!!

I bet..

[platypibri]

Even at this level of digital imaging, it has a weak, useless flash, intolerable low light noise, and sucks batteries in no time. Actually, I'm really looking forward to seeing the images this thing captures

Nice piece of kit!

[Paul Townend]

The article mentions "The five-year DES hopes to generate about 100 terabytes of data" that will be released to the public at regular intervals....

This kit is probably one example of why the world needs more 92 Tbs routers; sharing the data generated by this baby will probably be a task not unlike that faced by the Large Hadron Collider at CERN. You're going to have to have a really nice architecture and set of protocols to be able to efficiently pass around these images - possibly this is where Grid Technology comes in to play....

Of course, then you'll need something to actually process the images on! I guess Intel and AMD still have a rosy future ahead of them...

Nice, but...

[Bill_Royle]

Is there Linux support for it?

Re:but

[supergiovane]

I dont't know that, but I heard that it has quite good lenses and an amazing zoom. You really don't need to go to Japan with it: you can take detailed photos of Japan directly from home!

Taking a rule of thumb too far...

[supercytro]

[Person A]I told you before, megapixels don't matter... [Person B]But it's 500 Megapixels... [Person A]500 megapixels, 5 megapixels... it doesn't matter. Everyone knows that. It's common knowledge that megapixels is just a marketing trick. [Person B]But... [Person A]Look I read slashdot and everyone says the same. [Person B]ok...

Re:CCD?

[CyberBill]

CMOS is cheaper and can transfer the image faster off of the chip, but CCDs offer lower dark noise and lower reading noise, which means that your pictures are clearer and more scientifically usuable. CCDs are also INCREDIBLY more sensitive than CMOS, with the newer chips able to get upwards of 85% efficiency.

-Bill

Search for Dark Energy

[Chelloveck]

How can they tell a picture of dark energy apart from a picture where they just forgot to take the lens cap off?

they should wait

[lordmoose]

I heard that the 502 mega-pixel camera is coming out in six months.

pixel size < cell size

[morcheeba]

If you're 6 feet tall and we use the long dimension of the image (240k pixels), that's 7.62 microns per pixel. A typical cell is 10 microns, so we've got a pretty detailed picture of you.