All of these high-resolution planetary images were made with DSLR cameras by recording video at a high-framing rate with "Lucky Imaging."

High-resolution work planetary work requires a lot of magnification. Small details on the Sun, Moon or planets must be enlarged so that they can be optimally recorded by the size of the pixels in our sensor.

High-resolution photography also requires cooperation of the Earth's atmosphere in terms of "seeing". Seeing describes the steadiness of the atmosphere. Bad seeing can make planets shimmer and bounce around and totally destroy the fine details in an image that you want to record.

Sampling

To record critical detail at the highest resolution, you need to match the focal length of the telescope to the size of the detail you want to record based on the size of the pixel in the camera's sensor, for a given set of seeing conditions.

For example, say you want to shoot 1 arc second details on Mars on a night when the seeing permits it. You have a telescope capable of resolving this kind of detail. You will need an absolute minimum of 2x smaller than this in "sample" size. That means you need a pixel size that records 0.5 arcseconds of sky per pixel at the focal plane of the telescope.

This criteria for sample size is based on the Nyquist sampling theorem. Sampling is measurement in discrete, regular intervals. Spatial sampling in a digital camera is done by the number of pixels in a given sized sensor area and their size. In reality, for high-resolution planetary imaging, you need a bit more than this 2x criteria, perhaps as much as 3x to 3.5x the size of the resolution detail you want to image.

Since the size of the pixels are fixed, you can't change those to make them match the size of the detail you want to record. The size of the detail on the planet is fixed also. So the only thing you can change is the magnification of the telescope to make it enlarge the details in the image so that they match the Nyquist criteria for the size of your pixels.

Think of it like this... the sample size of the detector is the individual photosite, or pixel, in the CCD or CMOS array. For the Canon 60Da, it is 4.3 microns square. This is fixed. At a particular focal length, for example, at 1,000 millimeters, the field of view for the detector size will be fixed. It will cover just about 50 x 75 arcminutes. The sensor in a 60Da has 5200 x 3462 pixels. Divide the number of 4.3 micron pixels on the long side of the frame (5200) into the field of view on the long side (75 arcminutes, or 4500 arcseconds), and you get 0.86 arcseconds per pixel. If you are trying to image 1 arc second details on the planet with 1,000mm of focal length, we need to sample at 2x this size at a minimum, and 3x to 3.5x optimum, so we need to change something. We can't change the size of the pixel in the sensor, so we have to increase the focal length of the scope and make the image larger so that we can correctly sample it.

In this example, we want to image 1 arc second detail on the planet. So we need 0.5 to 0.28 arc second per pixel angular coverage for 2x Nyquist and 3.5x Nyquist respectively. We have 0.86 arcseconds per pixel angular coverage now at f/8 with our 5 inch telescope. So we need to increase the focal length. We can do this by using a Barlow. A 2x Barlow will double the focal length. This will give us about 0.43 arcseconds per pixel. That's close to the 0.5 arcseconds that we need. If we use a 3x Barlow, that will give us 0.286 arcseconds per pixel, exactly what we need for 3.5x Nyquist sampling.

Use Formula 12 to determine image scale per pixel.

The difference in image size that magnification makes is seen here. At the top left is a full-size image of Saturn shot at prime focus with a 5 inch refractor at f/8 with 1,000mm of focal length. You can see that Saturn is tiny at this image scale. On the bottom right is Saturn shot with the same telescope, but with eyepiece projection working at about f/40, giving an image size about 5 times larger. This image was shot with a Canon 20Da using the Live View analog video-out signal recorded as an AVI video file on a computer. The AVI file was then run through RegiStax and the best frames selected for stacking and sharpening.

The other consideration here is that as we increase the focal length of a telescope, we also increase the focal ratio, making the telescope slower photographically, and requiring longer exposures. A 2x Barlow makes the scope f/16 and quadruples the exposure time. Going from f/8 to f/11 is one stop, and to f/16 is another stop. Two stops = 4x the exposure time.

Magnification vs. Exposure

In our example, 3x the sampling would require almost a 4x Barlow, and would make the f/stop go up to f/32. If the original exposure was 1 second, it would now need to be 16 seconds. Luckily, the major planets are brighter than this and don't require exposures this long.

This decision on the magnification needed to reach the optimum sampling size turns into a compromise because of the longer exposures that a longer focal length require, the accuracy of the telescope's drive, and quality of the seeing during those longer exposures.

Smaller pixels come into play here because they do not require as much magnification to reach the proper sampling rate. With smaller pixels, faster focal ratios, and faster exposures can be used.

It is the increase in exposure that constrains how much you can magnify an image to increase the sampling. Depending on the particular seeing on a given night, the moments of great seeing my last only a second or two. Your brain can integrate these moments of great seeing and remember detail, but a camera has to sample it during an exposure. The camera's shutter has to be open during these moments of great seeing, but only for the length of the best seeing. If the great seeing lasts 1 second, but you expose for 10 seconds, the detail in that 1 great second will be blurred by the bad seeing in the other 9 seconds of the exposure.

Lucky Imaging

We want to try to match the exposure to the length of the good seeing. The trouble is, we don't know when these brief moments of great seeing will come along and we don't know how long they will last. They may be only once every ten seconds. They may be only once every 30 seconds. They may be more frequent. They may come along randomly. They may only last for a second or less under average seeing.

This is why recording video makes great planetary images. We can shoot lots of frames with video in a short amount of time before planetary rotation smears fine detail. At some time during these exposures, by chance, we get lucky and the seeing happens to improve. All of the frames are then analyzed and graded, and only the sharpest ones taken during the best seeing are used. They are composited together to increase the signal-to-noise ratio in the image. This is called "Lucky Imaging."

The same strategy can be employed with planetary imaging with a DSLR camera shooting single frames, although you usually can't shoot nearly as many frames. Match the focal length of the scope to the sample size needed to record the finest detail that the seeing allows. Then shoot lots and lots of single frames and examine each one, and only keep the very best.

In-Camera HD Video Recording

Some of the latest DSLR cameras are now also capable of recording high-definition (HD) 1080p and 720p video directly to the memory card in the camera. The problem with this for planetary work is that the sensor's native optical resolution must be downsampled to the high-definition video resolution. This is bad for planetary detail.

The shadow of Ganymede (lower right) transits across the face of Jupiter. The North Equatorial Belt (NEB) is prominent, but the South Equatorial Belt (SEB) has faded. This image was shot with eyepiece projection with an 18mm eyepiece at f/34 at 4,420mm of focal length with a 130mm f/8 triplet apochromatic refractor. Focusing was done with Live View on Ganymede. The image was shot as video recorded in the camera to the memory card with Canon's 550D (Digital Rebel T2i) DSLR camera with 640x480 movie crop mode at 60 frames per second. The ISO was adjusted in the camera to change the brightness of the image and achieve the correct exposure. One minute of video was shot containing about 3,600 frames. The best 100 frames were selected in RegiStax and stacked and sharpened.

However, Canon's 550D (Digital Rebel T2i), 60D, and 60Da have the ability to crop down to the central 640 x 480 pixels and shoot video at a very high framing rate of 60 frames per second. This is called "640 x 480 Movie Crop Mode." This uses the sensor's native optical resolution and does not downsample, so it is a very effective tool for high-resolution planetary imaging.

Not all cameras have 640 x 480 Movie Crop Mode however. Luckily there is another way to capture video, even on those cameras that don't shoot high-definition video, if your camera has Live View. We simply use a special program to record the Live View images at 5x magnification as a video to a computer. Capturing Live View as video can be done with the following programs:

• Astro Photography Tool (APT)
• Backyard EOS (BYE)
• EOS Camera Movie Record (ECMR)
• Images Plus (IP)

These programs only work with Canon DSLR cameras on a Windows computer.

Note that capturing Live View only gives you an image that is usually in the range from about 484 x 568 pixels to 1024 x 680 pixels. This is not what is considered high-definition video, but because it captures the full resolution of the sensor at 1:1 pixel resolution, it is definitely high resolution for planetary imaging because most of the planets are so small.

Jupiter and Saturn are only about 50 arcseconds at their largest apparent size. If we want to record 1 arc second detail with pixels and a focal length that sample 0.3 arcseconds, then Jupiter is only going to occupy 150 pixels in the image. So the 640 x 480 pixel frame in Movie Crop Mode, or 484 x 568 pixels frame in Live View captured at 5x magnification are more than enough to fit the planet in the frame and capture high-resolution details.

Planetary Image Processing

After the video is captured, a special planetary imaging processing program like RegiStax or AutoStakkert! is used to grade the frames and pick out only the best ones. These really good frames are then automatically stacked by the program to improve the signal-to-noise ratio in the image. Special sharpening techniques are then used to bring out the fine high-resolution detail that is hidden in the image.

Example 1 - Sampling 1 Arc Second of Detail with 4.3 Micron Pixels

For example, say we want to sample 1 arc second of detail. That means we would need each pixel to cover 0.33 arcseconds on the sky to sufficiently sample at 3x. And lets say we have 4.3 micron pixels in our camera. Plugging these numbers into the formula, we get:

With a 5-inch f/8 refractor with 40 inches of focal length, we would need about a 2.5x Barlow to give us 100 inches of focal length to correctly sample 1 arc second of detail on a planet, for a camera with 4.3 micron pixels.

Example 2 - Sampling 1/4 Arc Second of Detail with 4.3 Micron Pixels

Let's look at another example. Let's say your goal was to do the finest planetary imaging in the world, and you had access to a site with some of the world's best seeing, say, 1/4 arc second.

First thing you would need would be an excellent telescope that could resolve 1/4 arc second, so you would need a minimum of 14 inches of aperture. And let's say your camera has 4.3 micron pixels.

To correctly sample 1/4 arc second (0.25") planetary detail, you would need 3.5x this in sampling rate, so 0.25 / 3.5 = 0.071. Your pixels would need to cover 0.071 arcseconds.

Plugging in the numbers, we get:

If you were using a Celestron C14 scope, the native focal length at f/11 would be 154 inches, so you would need to use about a 3x Barlow. 154 x 3 = 462 inches. Close enough. A 3x Barlow would make the system f/33 also, which is fast enough for planetary imaging.

See the appendix section on formulas for field of view and angular coverage per pixel for calculating these, and other, parameters.