One billion pixels under the stars (11/30/08). Live view on a Canon 50D and some new alignment and stacking software renew my years-long effort to do truly decent astrophotography without having to transport and rebuild an observatory every single time I want to shoot something way out there in the dark.
The usual drill was to set up my Losmandy G11 mount with a 5-inch Astro-Physics refractor on it. Mount an Orion ST80 as a guide telescope on top of the A-P. Put a Meade DSI camera behind the ST80. Connect that to a notebook computer running Maxim DL. Run another connection from the computer to the G11 to let Maxim DL make fine guiding corrections as the guide star imaged by the ST80 drifts on the DSI's CCD. Meanwhile, behind the A-P could be a thermoelectrically cooled SBIG 2000XM CCD camera with a narrow band filter (if I were aiming to go very deep in a relatively small part of the sky) or a Canon 20D (if I wanted a wider, shallower, casually colorful starfield). The CCD would be connected to the notebook by a third connection; the DSLR could keep its own data or send it back to the computer as seemed more apt for any given project.
The notebook computer (and, optionally, the cooled SBIG CCD) required a big battery or an AC line. Focusing the 20D or the CCD requried tons of care and repeated inspections of downloaded images on the notebook's screen or tedious inspection of the point spread function as analyzed by Maxim. Decent deepsky images almost always require stacking of multiple images to build up more signal than noise, and even with Maxim's help, aligning and stacking was enough trouble that I'd rather stack 5 or 6 or 7 than 20 or 30 or 40. So if I wanted an hour's data on, say, M31, there was a reason to keep individual exposures longer rather than shorter. That meant polar alignment and guiding calibration had to be spot on. So, all in all, even with practice I was lucky to get a good image started in less than an hour from deciding to make one.
So see that image up top? That's 62 minutes of data on the bright galaxy M31 in Andromeda. It's sixty-two 1 minute exposures. It's worth looking at the higher res version. The mount required no special control, and a simple Canon remote release operated the camera (set the camera on Manual, Bulb, ISO 1600, RAW; set the release for 62, 1 minute exposures with 1 second in between; take aim; push the button; go away). Why not 60? I wanted an hour's exposure so I made a couple of extras in case I needed to toss some data owing to accidents (cat rubs on telescope blurring the image; airplane flies through the field leaving brilliant skid marks; that sort of thing). It took all of five minutes to set up the outfit; software (freeware!) did all the image alignment and stacking.
Here's what's different: Live view and the hi-res screen on the 50D made focusing casual. (It could have been better; I'll use the diffraction mask I use with the CCD next time.) And the ability to stack effortlessly an arbitrary number of short exposures meant that I could forego setting up and using the guide scope and notebook computer. All I had to do was line up the G11 on the north celestial pole with a modicum of care using its internal alignment telescope, plug in a rechargable 12v battery, turn the drive on, and go away. The G11 can be depended upon to track for 1 minute at a time without guiding corrections at focal lengths of 540 - 760mm, the range at which I use the 5-inch A-P. It can go quite a bit longer, actually, but only with more meticulous setup. It's a trade-off. (What isn't?)
An hour and change later, I went back outside, retrieved the CF card, and copied the files to a hard drive. DeepSky Stacker (v3.2.2) can read the 50D's RAW files and it needs no user input to find and align images. I selected all 62, invited DSS to align them and stack them (I used the "entropy-weighted-averaging" flavor of its stacking algorithm to maximize dynamic range; other options are much faster but don't work quite as well), and then I made some coffee and read The New Yorker for a while.
On a 2004-vintage Pentium IV with 2GB of RAM running at 2.8GHz, DSS took about an hour to align the data and another ninety minutes to add it all up and present it to me as a 16-bit TIFF file. That's very close to a billion pixels (15.1e6 x 62 = 936,000,000) distilled into a single image which I cropped very slightly to retain only that portion common to all 62 frames. I pulled that into Photoshop CS4 and carefully stretched them out to reveal dark lanes in the outer whorls of the galaxy as well as the bright blue starcloud at upper right, the faint (very faint!) trail of stars behind the satellite galaxy M110 and two subtle dust clouds near its core, and retain visible detail down into the bright inner reaches of the galaxy. Not bad. Pull up the 1024-pixel version linked above (but understand that you're looking at a JPEG compressed 8-bit image representing something less than 5% of the full-image's data). I'm still not real pleased with the color rendition, but I've got a lot of options left to explore (and there's a user group for DSS with lots of archived messages and expert assistance if I get completely stuck).
If I'd had them, DeepSky Stacker would also have applied flat fields (to even out the illumination across the chip), darks (to control noise and hot pixels), and bias frames (to compensate for readout noise and other sensor-specific effects). They'd matter. Especially the flat fields and the darks. That's something to make routine next time.
The goal here is to be able to set up quickly, either at various places in the yard or around the neighborhood (to avoid trees and obstructions) or out in the dark on friends' farms, up on the Blue Ridge Parkway and Out West (to avoid clouds and city lights), without having to transport enough electronics to open a small business, alarm the TSA, and make it all seem too much work to enjoy. So far, so good.
:: top ::