SBIG ST2000XCM CCD Camera / GM8 Mount: Learning Curve

This is a fine camera, and such a pleasant experience compared to the Meade DSI and it's obtuse software. Of course, it cost ~10 times as much, so maybe it's not a fair comparison. The 1600x1200 chip is arrayed as 800x600 elements of 2x2, each with 2 green, a red, and a blue filtered pixel, giving single-shot color pictures. The chip reads out 16 bits in the SBIG compressed format. So far, I've used it strictly on the GM8 mount + Megrez flourite APO refractor at prime focus. The Megrez is another fine piece of workmanship, and gives very sharp stars when looked at visually.

Sept 10, 2005. This is the ST2000xcm's first shot. The Orion nebula from Carrizo Plain under clean skies. With a full moon and dawn beginning in the east, this nebula is so bright it still contrasted well enough against the background sky. 2 minutes, no self-guiding; just free running on whatever polar alignment I managed to achieve.

I focused carefully, but yet I see fat stars. Why were they so fat even though the focus was good and the image scale was large? The stars should be tack sharp. Here's the problem I think, the CCDOPS software will generate single-shot color images, but compresses the 16 bits into 8 bits. That's only 256 levels. That's 6 magnitudes of surface brightness range - not enough to cover a moderately bright star and a faint nebula adequately at the same time. Surely some bright programmer out there will write software to allow outputting 16 bit TIFF single-shot color pictures? I'll have to ask SBIG why they did this. I see this in all ST2000XCM pictures on the web. However, it may be possible to overcome this another way; by stacking short exposures separately from the grand total, using Photoshop 'layers' command. I need to practice with this.

Oct 8, 2005. Next time out... at our Astro 9 photo trip to Mercey Hot Springs. Clouds wrecked most of the night, but I had time for one good set of images on the Horsehead / Flame nebula region. It was shot with the SBIG ST2000XCM: a stack of 4 x 5min images, each dark-subtracted at -30C below ambient. The final 5min image was in noticable twilight. The frames were single-shot color processed, then stacked in Registax v.3. Then a low-pass wavelet filter further smoothed the nebulosity. Finally, the brightness, contrast, and color saturation was tuned up in Photoshop by Shahram. So, this trip I learned how to stack images. I focused with some care, but did not put CCDOPS5 into "planet" mode for focusing.




Nov 16 '05 - At the observatory, I mated the Losmandy dovetail plate to the 8" Meade SN scope. Got it on the GM8 and then put the ST2000XCM in the eyepiece holder and focused. Found the ST20000XCM nosepiece is a bit undersized for a 1.25" eyepiece holder, causing it to rock back and forth even after tightening the thumbscrew(s). And also causing the focal plane to not be parallel to the detector. Got to solve that before more photography. Also found that my idea of attaching a splitter and having both the hand paddle and the autoguider cable both connected to the GM8 does not work - the GM8 computer recognizes only one at a time, and calibration and autoguiding can only be done when the hand paddle is disconnected altogether. It was a full moon night, and the Dumbell Nebula was only 15-20 degrees up in the west, so the image at left is not bad, all things considered. 5 minutes, converted to jpg but otherwise no post-processing or cropping. The stars in the lower left corner are out of focus. The blurry stars do not show radial distribution, but instead seem better explained by the detector not being parallel to the focal plane. The size of the field here is 0.9x1.5 degrees, for this 8" f/4 = 32" focal length. The limiting magnitude I estimate at 17.

Also, I added "RA" and "Dec" labels to the GM8 motor housings. A small thing, but it always seems to take more mental energy than you might think to plug the cords in correctly. And, I found this handy piece of iron stock (left) at the local scrap yard, and tapped it to accept a 1/4-20 knob. It's perfect for a variable lightweight counterweight, and also protects the threads of the solid stainless steel counterweight shaft. Finally, since it was well below freezing that night, I decided to use my watercolors box as insulation against the cold ground for the batteries. I use 12V gel cells - one for running the laptop, one for running the ST2000XCM, and a small 7.5AH for running the GM8, which draws little power (the GM8 manual explicitly recommends not running the GM8 mount with other equipment off the same battery).




Dec 3, 2005 - Another trip to Carrizo Plain. Clouds zapped the first 2/3 of the night. It was cold, COLD! The TE cooler hardly had to work to get to -30C; it was 17F when I started photographing at 1:30am. I got over an hour of shooting on the Cone Nebula, and planned to do a side-by-side comparison of the Orion Nebula with ST2000XCM on the Megrez, and then with the 8" LXD75, which has over twice the light-gathering at the same image scale. But I'd forgotten about the poor mating of the nosepiece to the 8" and so that comparison will have to wait. At dawn, I spent some time getting a library of twilight flats. However, even though the flats look basically good, I am disappointed at the flat fielded versions. Compare these versions of the first 5-minute exposure of the Cone Nebula region, zoomed in to show just the Cone itself and contrast enhanced...

No flat field

Flat fielded with a single flat

Flat fielded with average of all 10 flats

Flat fielding should not only remove vignetting and other large-scale patterns, it should also reduce the pixel-to-pixel noise due to differing pixel sensitivities. However, the close-ups above seem to show more noise using a single flat, and even more (perhaps) when using an average of all 10 flats. To get the above images, all from the first 5 minute image, I zoomed in on a small area around the Cone Nebula, single-shot color processed with sRGB+gamma, imported to Photoshop 7 and adjusted the levels and contrast to look as close as possible. See below for more insight into flat fields...

Jan 30, '06. A quick trip up to Coast Rd reservoir and got some images of the Orion Nebula. Walking away briefly, the laptop seems to have run out of power, despite having said it had 4 hours of time left. Strange. I ran it off 120VAC using the inverter. Meanwhile, the CCD camera was still powered continously from the other battery. When I tried re-starting CCDOps5, it gave me a bogus error - "error in USB device. Try reconnecting. If this does not work, replace the device". Replace the device?! That's a $3,000 camera! After more fiddling, I learned I have to unplug the power to the camera before communication can resume.

2006 discoveries.

Lost Exposures: First, I've found that it's pretty predictable to lose ~1 shot per night. WinBlows comes on with a message about having to close the application yada, yada... Of course, this happens right as it's beginning to download the time exposure, so you lose the entire exposure. I don't know why it happens, just typical crap from Microsoft. It's happens randomly, uncorrelated with anything I've yet discovered. Our Dell laptop has 2BG of memory and a Pentium P4 with plenty of cache and with no other applications running so seems to me it shouldn't be happening. Anyway, just learn to expect it. Because of this, I've decided that it doesn't pay to take single long exposures. The frustration of losing a 10 minute or 15 minute exposure is many times higher than losing a 5 minute exposure. The additional advantage of building a long exposure by stacking many shorter ones is that it leave open the ability to post-process and bring out both the faint and bright areas of the photo. Yet another advantage is that sometimes you find you need to take more dark frames because you can't get down to the temperature desired, and it's easier to set aside time for a 5 minute dark frame or two than for longer darks. So my rule now is - All exposures are 5-minute exposures. Disk space is cheap these days.

SBIG says this problem happens because of some sort of periodic checking of Windows and can be fixed by changing it to 'classic'.

Flat Fields: The problem with the flat field tests above is just inadequate signal to noise. Flats require much more care and effort than dark fields. It pays to take great care and invest the time to make quality flat fields. If you can't, then don't apply them at all. I bought some gray matte finish art paper and easel and mounted it inside the new storage building. I then can set up the ST2000xcm with either the Megrez or the camera lenses and take many images of this gray field. For illumination I use our flourescent camp light placed at least 20 feet away and close to the perpendicular axis off the paper, for even illumination. All the other flat field devices I've found on the web (and made myself) are inferior to this simple solution, in my opinion. Illuminating from a distance is the cheapest and most effective way to guarantee even illumination. Then I've taken 3n images, where n is typically 2. Then use CCDOPS ability to median-combine triplets of images until you have a single final median'd image. The level of the individual frames should be in the neighborhood of 20k or so. Far enough below the 65k full well capacity to still be in the linear range, but large enough to minimize shot noise. When I do this, I get much smoother frames than the darks used in the Cone Nebula tests last year shown above. When using the ST2000xcm with camera lenses, it may make a difference what f/ ratio the lens is set to, but the orientation does not matter since there's no flexure. But when using with the Megrez, the focal point is so far away that we need not one but two 2" extension tubes on top of the far end of travel of the draw tube. This produces detectable flexure, especially when extending the dew shield as I recommend. So, I've assembled a set of flat fields for the ST2000xcm + Megrez, labelled according to the rotation angle that the camera is set to at the end of the telescope. If I really wanted to do it right, I'd expand across a second dimension; the altitude of the target. But that's very hard to arrange with our flat field setup so I have not done it. The flat fields in the image files range from an orientation of 90 degrees to 270 degrees, where 90 degrees has the flat side of the camera facing due left, and 270 due right, and 180 is straight down. For most subjects not compositionally requiring strange tilts, the 180 degree setting is the recommended one as it lets the cables hang straight down. I've tried using the flat fields from the Megrez on photos using the 8" LXD75. Not a good idea; there's much less vignetting in the LXD75. Take honest flats with the LXD75.

Bad Connections to RA and/or Dec: This is an increasing problem with the GM8 mount. The control unit has 6-pin phone-plug connectors going to the RA and Dec motors. I've noticed that sometimes one or more of the 4 directions (-RA, +RA, -Dec, +Dec) will not respond to either the hand paddle or to the autoguider cable. One design flaw (in my opinion) in the GM8 mount is the requirement that the speed of the motors can only be adjusted at the control unit when the hand paddle is connected into its phone plug. This requires a lot of swapping in/out of the hand paddle and the autoguider cable during the night. These don't go in and out as easily as they should. Result; a lot of strain. Are the connectors now flaky in their connections? Nov '06 - no, the problem is actually that the GM8 speed wasn't set high enough. You have to use the 2x speed, not the 0.5x speed for the Megrez or LXD75. The scope doesn't move far enough to pass whatever internal limits the software has, and it defaults to "0" movement on calibration.

Error message: "Shutter Error". Sounds like a mechanical problem,. I got this error on a fair number of my exposure attempts while at the observatory. Tried rebooting the computer. Didn't help. Got home and emailed SBIG - told me it was actually a software issue. I wondered if it was repeated interrupts from pesky adware, or maybe being connected to the internet (I forgot to disable the wireless internet and it was connected (weakly?) to the node off the 12" dome. Once I ran adaware, and disabled the wireless internet, no more problem.

Dark Subtractions when the exposure lengths don't match. CCDOPS says it can still do a good dark subtract by just clicking the 'adaptive' button. However, that is not my experience. Here's an example... each frame below is a 15 minute image cropped to a small part of M78, single-shot color processed with the default settings of the sRGB+gamma option, then jpg'd at 100%, with no Photoshop adjustments. The only differences are the dark subtractions. The chip temperature was set at -13C. I made a fake 15 minute version of the 10 minute dark frame dk10-13.st4k by using the 'pixel math' command and multiplying every pixel value by 1.5.

No dark subtraction.

Dark subtracted with dk10-13 with 'standard' option. Clearly better

Dark subtracted with dk10-13 with 'adaptive' option. No better than 'standard' option

Dark subtracted with dk15-30 with 'standard' option. Much better than using the un-scaled dark.

Dark subtracted with dk15-30 with 'adaptive' option. Not any better than 'standard' option

Conclusion - don't trust the 'adaptive' method to correct for different exposure lengths. You must use the pixel math command 'multiply' to scale the dark to the same time length as the image, if you can't actually take a dark of the proper exposure length. S/N arguments suggest take a series of long darks and then cut them back to shorter times as needed.