CHAPTER 13: Digital Astrophotography — What’s NEW?Chapter_13__Digital_Astrophotography.htmlChapter_13__Digital_Astrophotography.htmlshapeimage_1_link_0
 
A-B Comparisons

In Chapter 13 we show a number of sets of images, showing “before and after,” or “with and without” 
A-B comparisons. Their small size in the book may make it difficult to see what we’re trying to convey. 
So here are a number of those same comparison images, depicted in full sized versions. The page may take some time to load, but the comparisons are enlightening — we have not seen similar tests elsewhere.
Conclusion: The cooled CCD camera, as expected, has less noise than the DSLR, with a smoother more uniform background. The DSLR shows a mottling from chroma noise (both are relatively unprocessed images). However, the DSLR records similar detail and limiting magnitude. (The ST402 is a fairly sensitive little camera.) Claims that cooled CCD cameras are 4 to 10 times more sensitive than DSLRs are suspect — don’t expect to record images in much shorter exposure times with a cooled CCD camera; if anything exposures will be much longer due to the need, at least with monochrome cameras, for red, green, blue and luminance exposures.
Page 277 — Cooled CCD Camera vs. DSLR
Cooled CCD Camera

This was taken with a Santa Barbara Instruments ST402, monochrome camera, cooled to 20° C below ambient temperature

It is a trio of 8 minute exposures, through each of red, green, and blue filters, taken with 5-inch apo refractor at f/6.

Note the flaws from cosmic ray hits and hot pixels.
DSLR Camera

This is with a Canon 20Da at ambient air temperature.

It is a single 8 minute exposure at ISO800, taken the same night as above, through the same telescope, with the DSLR image cropped (but not resampled) to match the much smaller field of view of the ST402. 

Note the colored mottling from chroma noise.
Conclusion: While mileage may vary with different cameras, with Canons we’ve used this performance is consistent: Noise at ISO400 and 800 is quite low but starts to climb rapidly at ISO1600 and higher. However, our key point is that the prime contributor to noise is boosting ISO sensitivity not increasing exposure time. These images all have equivalent exposures (the ISO1600 shot has 4 times the sensitivity but 1/4 the exposure time of the ISO400 shot). The lowest noise comes from keeping ISO speed low and increasing exposure time.
Canon 20Da DSLR
12 Minutes @ ISO400

200mm telephoto at f/5.6

All are extreme blowups of a segment of the original image.

At ISO400 noise is quite low and the background sky is fairly smooth.
Canon 20Da DSLR
6 Minutes @ ISO800

200mm telephoto at f/5.6

At ISO800 luminance noise is higher giving the image a grittier, grainier quality. But noise is still well under control.
Page 280 — Long exposures at slow ISOs vs. Short exposures at fast ISOs
Canon 20Da DSLR
3 Minutes @ ISO1600

200mm telephoto at f/5.6

At ISO1600, noise is now noticeable, with more luminance and chrominance noise. This setting is fine if you really need the speed, but is best avoided if possible (at least with the generation of cameras used here).

Note that, perhaps due to temperature shift, the image got sharper by this frame, though the shift in focus does not affect noise levels and visibility.
Canon 20Da DSLR
Stack of 10 x 1 Minute Exposures

Here we took a series of short 1 minute exposures, each of which was severely underexposed. But stacking and adding the images together (in this case with ImagesPlus software) creates an image equivalent to a single 10-minute exposure. Or does it? 

This is a blow-up of a section of the original frame.
Canon 20Da DSLR
Single 10 Minute Exposure

This is a single 10 minute exposure, taken the same night as the above image with the same telescope optics. 

Note the smoother background, lower luminance noise, less pattern noise (vertical banding) and more detail in the faint areas of nebulosity. However, detail in the bright areas is washed out. That can be dealt with through other techniques, like blending in a masked short exposure.
Page 297 — Stacked short exposures vs. Single long exposure
Canon 20Da DSLR
Without Dark Frame

Here we took a single 12 minute exposure at ISO400, with Long Exposure Noise Reduction turned OFF. So there is no dark frame applied, either in the camera or afterwards in processing.

The noise here — the green pox — is excessive because this was a very warm summer night, the worst conditions for good imaging with an uncooled DSLR.
Canon 20Da DSLR
With Dark Frame

Now we turned ON Canon’s Long Exposure Noise Reduction option (under Custom Functions). The 12-minute exposure was followed by the camera taking a 12-minute dark frame and subtracting it internally to eliminate the noise. 

It worked! Most of the pox is gone, leaving a lower level of luminance noise. These are extreme blowups of a section of the original frame.
Page 297 — Without dark frame vs. With dark frame
Modified Canon 5D
Original JPG

Here is a blow up of a section of a JPG image, as it looked out of the camera with default settings applied. The camera was set to take the highest quality JPG with nominal sharpening and contrast applied by the camera. 

Image quality is good and there’s little evidence for blocky artifacts from the JPG compression. But ...
Modified Canon 5D
Original RAW

Here is the same image shot as a RAW frame, and opened with Adobe Camera Raw with only mild sharpening applied and default settings for contrast and other enhancements. This is as a “raw” a RAW image as you can get and still see a sensible image to match the JPG above.

There is a greater range of tonal values recorded. Nebulosity has more structure and is not blocked up. Stars are sharper.
Page 297 — JPG vs. RAW
Modified Canon 5D
Single Image

This is a blow-up of a single 5-minute exposure with a 135mm telephoto lens at ISO400. As Photoshop’s Layers box shows, there are other layers stacked on, but they are all turned off for this demo. 

With one exposure, random luminance noise is apparent.
Modified Canon 5D
Stack of 4 Images

Now we have turned on three other layers in Photoshop. Each layer is another exposure, stacked and registered with the original bottom layer. For this to work each layer must be a separate and different exposure. 

So this is the effect of averaging four exposures as per page 297. Layer 1 is set to 50% opacity. Layer 2 to 33%, Layer 3 to 25%, all at Normal blend mode. This averages the random noise and smooths out the sky.
Page 297 — Single image vs. Stacked images (four)
Canon 20Da
8 Dark Frames Averaged and Subtracted Later

Here we followed the classic advice of CCD imagers and repeated by most DSLR books: we took 8 dark frames (on a very warm night, so noise was high). Then using ImagesPlus we automatically averaged those 8 darks and subtracted the “master dark” from the “light” frame. The result still shows a lot of spotty noise artifacts. (For an image with no dark frame see the Dark/No Dark comparison above.)
Canon 20Da
1 Dark Frame Subtracted in the Camera

Here’s the result, taken the same night, with the DSLR camera taking its own internal dark frame with the Long Exposure Noise Reduction mode turned ON. The result: most of the noise is gone. The camera did a better job eliminating noise than could be accomplished later through the typical process advised by CCD imagers.
Page 305 — Dark frame subtracted in-processing vs. Dark frame subtracted in-camera
Conclusion: 
For long exposure deep-sky images, always let the DSLR do the dark frames. Yes, it takes twice as long to get an image recorded (exposure time + equal dark-frame time) and the camera is ready to take another image, but the results are better. 

The dark frame subtraction process recommended for cooled CCD cameras is not necessarily the best method for DSLRs, which are uncooled, unregulated and that apply a lot of internal processing. This goes against the advice you’ll read elsewhere, but we’ve found it true in actual tests — the camera does a better job doing dark frame subtraction than you can. 

“In-camera darks” also make later processing much easier, with no fussing over libraries of dark and bias frames, taken at different times and temperatures which may never match the noise you are trying to eliminate in the actual image. In-camera darks ensures an exact match of dark to light frame, for the best elimination of noise.
Conclusion: Stacking and co-adding (note that’s adding not averaging) short exposures, while it does increase image density, also increases noise visibility. It does not smooth out noise — to do that you must stack and average exposures, as we instruct on page 301. The lowest noise comes from shooting exposures as long as the equipment and sky will allow. Additive-stacking lots of short, underexposed images works, but does not yield the best results. Lots of DSLR images we see exhibit the pattern noise characteristic of quick-and-dirty techniques.
Conclusion: The in-camera dark frame subtraction (called Long Exposure Noise Reduction in Canon DSLRs) does a superb job eliminating the worst effects of thermal noise, and is essential on warm nights. Only on cold winter nights might you be able to get away with no dark frames in long exposures (over 30 seconds or so). But for the best results, always apply a dark frame. Should you let the camera do it? Or take the dark frame(s) yourself and subtract them later? Scroll farther down for that test.
Conclusion: For critical shots, especially demanding long exposure deep-sky images, always shoot RAW. Processing takes more steps but those steps (available in convertor programs like Adobe Camera Raw) allow much more control of the image and factors like noise, contrast, field illumination, chromatic aberration correction, and highlight recovery. The processing is applied to the RAW data to extract and retain maximum detail. Images can be converted to 16-TIFFs for maximum tonal values. Unless you shoot RAW you are throwing away data.
Conclusion: For demanding deep-sky images, always shoot multiple exposures (typically four of each object) so they can be averaged together later to further reduce noise. This works only for exposures where the camera is tracking the sky. Images then have to be registered precisely as per instructions on page 301 of the book. Photoshop CS3 and CS4 and Photoshop Elements 6 and higher can do the registration automatically, as can specialty astro-processing programs such as ImagesPlus, Maxim, Registar and many others. A stack of 4 works well.