Level 2: I know how to use a camera, what do I need to know for astrophotography?
Image processing methods used to reduce background and noise become much more important. Also, it helps if you understand some basic optics: f-number, Seidel and chromatic aberrations, angular field of view, airy disc; and have some familiarity with digital sensor technology: Bayer filter, analog or digital signal processing and noise. You don’t need to master everything but knowing more helps you learn faster.
Your camera needs to be set to fully manual operation (including focus), and you need a decent tripod. Because modern image sensors are so good, blur is easier to see, vibrations are easier to see, and lens aberrations are easier to see. The overall goal is to provide your stacking program a series of images taken under identical conditions, so make sure all in-camera processing is turned off: no HDR, no noise reduction (especially no long-exposure noise reduction), no adaptive exposure, no automatic gain, etc. Because your camera is on a tripod, any built-in anti-vibration stabilization should be disabled. If your camera has a ‘mirror up’ setting, make sure you use it. A ‘Live view’ setting, if you have it, can also be used if the shutter does not re-engage between frames.
Mirrorless cameras have an advantage here, as there is no induced ‘camera shake’. However since mirrorless cameras permanently divert a fraction (up to 1 stop) of the incident light to the viewfinder, you also have lower throughput as compared to a DSLR.
You will be spending significant amounts of time outside- up to several hours- so make sure you dress accordingly- temperatures drop when the sun goes down, even in summer. And don’t forget bug spray. You will be working in the dark- setup, replacing batteries, inserting and removing memory cards, camera adjustments, etc. Night vision takes a few minutes to develop and when it does take steps to avoid looking at bright lights. Some essential pieces of equipment include a second camera battery (already charged), a small flashlight to illuminate LCD displays, some lens wipes, and a chair to sit in. A shutter release cable is also needed. Many manufacturers offer ‘intervalometers’ which are programmable shutter release cables, and some cameras already have that functionality built-in. Get one, they don’t cost much more than a nonprogrammable cable. Once you are set up and are happy with the exposure settings, don’t change any of the settings- stacking programs work best if all the images are taken under identical conditions.
Be safe. You are outside, possibly alone, at night, in as dark a location as possible, far away from lit areas. Stay on your own property or get permission from the property owner before setting up- in the US, public spaces close at sundown. Night-time is when animals are looking for food- most animals have poor vision and may not see you until they are very close. I’ve had face-to-face surprise encounters with cats, rabbits, raccoons, possums, skunks (no spraying, fortunately!), mink, and deer. There have been reports of coyotes in my neighborhood, but I have yet to see any- my point is, be aware of your surroundings.
Drift can occur when the temperature drops or as various lens elements settle under action of gravity. Check your focus periodically, every 20 minutes or so- focus error can’t be tolerated! Live-view, if you have it, really helps to focus precisely and rapidly. Similarly, when the temperature fluctuates you may find that your lens becomes fogged with condensation. That’s what lens wipes are for.
Just as your location on Earth can be specified in terms of (latitude, longitude), positions of night sky objects can be specified in terms of (declination, right ascension) or (DEC, RA). During the night, stars appear to move around the poles at constant declination. Converting the celestial DEC/RA coordinates to one centered on you, with “horizon” coordinates (azimuth-elevation) requires knowing the date and time, and converting between the two gives me headaches. I prefer to ‘star hop’ to my desired location: using bright stars as a guide, I aim in the approximate direction of what I want to image, acquire a test image and fine-tune as needed. I prefer google sky’s star maps to get oriented:
I also use SIMBAD and/or ALADIN databases for more specific information:
Log scales vs. linear scales:
Fortunately, if you are already used to log scales (f-numbers and Ev, for example), thinking in terms of stellar magnitudes is easier. Specifically, a unit increment/decrement in apparent magnitude corresponds to Ev changes of -1.3 or +1.3. This matters for dynamic range. If you have struggled to set a correct exposure for a harsh, sunlit scene, then you understand the problem of capturing sufficient dynamic range.
Digital sensors are linear sensors, but your eye is a logarithmic sensor. Consequently, digital image sensors have a much more restricted dynamic range than your vision. This is why your photographs of high dynamic range scenes often don’t ‘look right’. Astrophotography pushes dynamic range requirements to an extreme level.
Adding more bits (e.g. 8-bit vs 14-bit) does not mean the camera is more sensitive to light- quite the opposite. Subdividing an intensity interval (0,1) either into 255 or 16383 equal bins doesn’t change where the endpoints are; more bins means that the noise floor is pushed down, enabling more low-intensity information to be retrieved. Often, you will read or see how to use the RAW format to ‘push the exposure’ and rescue an underexposed or overexposed photo; this just means applying a nonlinear tone map to either preferentially boost the low intensity pixels or conversely, preferentially limit the high intensity pixels, in either case to bringing those extreme tones back into the midtone range. Tone mapping in general is also a method of image compression- a 14-bit RAW will eventually be tone mapped to compress tones initially spread over (say) 16000 bins into an 8-bit format with only 255 bins.
Astrophotography typically deals with scenes that have dynamic ranges of 100,000:1 or more- beyond even a 16-bit sensor’s capability to capture. Understanding how to compress your 16-bit/channel or 32-bit/channel (or more- I’ve used 48-bit/channel formats) high SNR stacked image into a visually balanced 8-bit/channel TIF or JPG is critical. Remember-in the end, your display will only show 8-bits of dynamic range. Supposedly, your vision can only distinguish 8-10 bits of greyscale- display manufacturers long ago standardized to 8-bit/channel, but there are a few 10-bit/channel specialized displays on the market.
Choice of lenses
All lenses can be used in astrophotography. Older, manual focus lenses can be ideal because they are often very inexpensive. Start with your lens having the largest entrance pupil. Wide angle lenses excel at photographing meteors and large-scale starfields like the milky way. The short focal length of wide angle lenses relaxes the time limits for preventing star trails, but wide-angle lenses are also very suited for creating star trails because of the large field of view. Normal lenses are great for medium-sized regions of the sky (single constellations or so), are usually inexpensive (I’ve seen 50/1.4 lenses, which are ideal lenses, on sale for less than $100), and generally well-corrected. Telephoto lenses provide high magnification but also are most susceptible to vibration and have very stringent limits on acquisition time.
Ensuring optimal optical performance
In general, lenses can’t be used at full aperture- off-axis aberrations become objectionable. That said, because fast lenses have noticeable fall-off, off-axis aberrations can sometimes be controlled simply because the intensity is lower, reducing the appearance of aberrated points. You may also find that the off-axis aberrations are not perfectly symmetric around the optical axis. It often helps to stop down the lens between 1/3 and 1 full stop to reduce coma, field curvature, chromatic aberrations, and flare. Alternatively, because stopping down a lens reduces the optical throughput, use the lens wide open and simply crop out the objectionable margins. Stopping down a lens can be used to introduce an artistic element: when you stop down, (after stacking) bright stars will have the characteristic ‘sunstar’ appearance due to diffraction by the aperture.
At some point, ‘seeing conditions’ will limit your image quality. Seeing, also known as clear-air turbulence, occurs due to thermal gradients in the moving atmosphere: the refractive index is proportional to the density, and the density varies with temperature. Seeing is often modeled as a weak lens with a spatially-varying focal length (a ‘slowly moving variable lens’). Seeing is a critical topic in observational astronomy, and the reason big telescopes are constructed high up on mountaintops, in locations with stable local atmospheres. Seeing can vary seasonally. The more air you have to look through (objects closer to the horizon) the greater the image will be affected by seeing. Qualitatively, seeing conditions can be forecast just like cloud cover and transparency (caused by airborne aerosols).
Seeing sets limits on the maximum entrance pupil of a lens- larger entrance pupils no longer provide smaller PSFs- but that limit is not particularly relevant for most camera lenses. In fact, camera lenses are often insensitive to seeing conditions because the entrance pupil is comparatively small.
The night sky is not dark. Light pollution is caused by inefficient terrestrial lighting. Light pollution is more problematic the closer you are to a major population center, but image stacking and additional post processing provides a way to greatly reduce this form of light pollution.
Question: How long can I expose for?
Stars appear to move at an angular rate of 15 arc-seconds per second (360 degrees per day). Assuming you want your image to be populated with colorful round blobs, the maximum acquisition time per frame is limited by the apparent motion of the star field across your sensor during an exposure. The linear velocity a star appears to travel across a particular pixel on your sensor will depend on lens focal length and the DEC that the pixel is viewing (The exposure time itself is then simply velocity*pixel size). A rough estimate can be made with the “500 rule”: 500/(lens focal length (35mm equivalent), in mm) = maximum time in seconds. This is a very crude estimate, you may also see “600 rule”, “400 rule”, etc. It’s possible to get better estimates, but the bottom line is simply to take a few images with different exposure times and see what the limit is. Personally, I end up closer to a “200 rule”. However, once you determine your maximum exposure time, it will remain that exposure time as long as lens focal length, pixel size and viewing direction remain the same. Now, if you are interested in star trails, then by all means leave the shutter open for as long as you like- more information is given below, under ‘special topics’.
PSF: Point spread function
This is essential metric of your images. As a practical matter, the PSF will approximate a Gaussian function (not an Airy disc) due to seeing and dynamic range compression. While the diffraction-limited PSF of a particular lens is easy to calculate, in my experience the actual PSF you will obtain is far, far below this limit. As a specific example, my 400/2.8 lens should generate a PSF FWHM = 1.7 microns (0.88 arcsec), well below a single pixel (5 microns; 3 arcsec). However, my PSF FWHM is, at best, 3.6 pixels- easily 10x the diffraction limit, mostly due to seeing. When I am finished post-processing, my FWHM is slightly larger at 4.3 pixels due to dynamic range compression (discussed below).
Image stacking: more details, please!
Image stacking is more than simply per-pixel averaging, or selecting the brightest value. Certainly, those are settings you can choose in a program, but stacking is used to generate a high SNR image from many low SNR images. This is done by taking advantage of the fact that the background intensity in your images is not due to sources with constant output, but from atmospheric scattering of terrestrial sources, which is stochastic. The contribution to intensity from this scattered light (approximately) follows Poisson statistics. More frames permits reduction of this noise component.
Data! Data! Data!
It is not uncommon (for me) to acquire 50 GB of image data per night. This is a lot of data to transfer and process, so I’ve optimized my ‘workflow’. Images are transferred from the camera card onto my local hard drive. Images are then ‘registered’ (the first step in stacking) and bad images are deleted- I threshold images based on the PSF FWHM. The remaining images are then transferred to my external hard drive for ‘permanent’ storage. The stacking program uses images stored on the external drive.
How do you get the best stacking results?
Definitely use RAW images, you are trying to maximize the dynamic range. Individual images should appear somewhat underexposed- you want to avoid overexposing on bright stars, so the majority of your scene will appear black if you linearly re-scale RAW images to 8-bit TIFs. This is good- you may be able to use lower ISO values than you may expect, increasing the per-frame dynamic range, resulting in better color rendition, lower background noise, and less posterization. In general, it’s important to avoid clipping- either at the high end or at the bottom. Clipping represents a permanent loss of signal.
Here’s some example values using my images (single images, not processed), obtained with 15-second exposures (see the next section covering ‘tracking mounts’): ISO, the background level (avg, s.d.) and the corresponding grey value of a magnitude 14.1 star (at the low end of what I can see on the image), the SNR, and the brightest non-clipped stellar magnitude:
ISO bkg (avg) bkg (s.d.) star avg star s.d. SNR Brightest mag.
160 698 36 1101 158 1.58 7.95
500 3517 170 4542 272 1.29 9.51
800 6088 243 7648 333 1.26 10.87
I obtained the stellar magnitudes from SIMBAD. As you can see, I get the best SNR ratio by using ISO 160, as expected- lower ISOs have higher dynamic ranges. That said, I have had poor results when stacking images with backgrounds close to black- here, I use ISO 500 rather than ISO 160.
If you are familiar with a ‘film characteristic curve’ or ‘H & D curve’, then this idea is easy to understand. Put the background greyscale near the ‘toe’, and the brightest stars near the ‘shoulder’. Set ‘gamma’ to be as high as possible to better separate the background and features of interest. That’s it. Honest.
What’s up with astrophotography cameras?
The issue is that the spectral response of commercial camera sensors do not well match the emitted spectrum from celestial objects- Hydrogen-alpha emission nebulas are a prime example. There are two basic approaches. One approach is to use a ‘regular’ camera and remove various coatings present on the CCD to allow greater throughout in the far red and ultraviolet. Just remember- your lenses are poorly corrected in the far-red and UV (far blue?), so you will probably also need spectral filters to control chromatic aberration. If you are using a reflective telescope or catadioptric lenses then chromatic aberration is less of an issue. The second approach is to purchase a specialized ‘astronomical camera’. These are scientific-grade imagers, often cooled and monochromatic.
Special topics: star trails, timelapse, imaging the sun, imaging meteors, imaging satellites.
Star trails: You can have a lot of fun with this.
Unfortunately I have to use the term ‘image stacking’ in an ambiguous way here. You will not be stacking these images in the sense we have previously discussed, so use whatever image format you are comfortable with. Leave the shutter open for as long as possible while maintaining the desired tone of the night sky. You can push this quite a bit and then do frame-by-frame background subtraction later. You will need to acquire images for at least 2-3 hours in order to generate dramatic trails. Wide angle lenses are ideal for star trails, because the large field of view provides the image with a large declination range. Wide angle lenses allow you to include some terrestrial scenery in the image, which can help the overall composition. You can have fun with this- adjusting the ISO either at the beginning or end of your session will produce trails with a bright ‘head’ or ‘tail’. You could adjust the ISO as you like to accentuate the effect. On occasion, I will interrupt the star trail acquisition partway through and acquire 50-60 stationary star field images for ‘normal’ image stacking, those images are removed from the overall stack and separately processed with ‘image stacking’. The processed image is then inserted into the original stack to provide a ‘snapshot’ of the star field superimposed over star trails. After you acquire a series of images, use an image processing program to assemble these into a ‘stack’, then simply project the brightest pixel values through the stack onto your final image.
similar to star trails, but don’t leave the shutter open for as long: 2 or 3 times the no-star trail limit is a good estimate- I use 1 minute exposures with my 15/2.8 lens. Acquire images with no intervening time delay (using an intervalometer is key) for 5 or 6 hours, then process and assemble the image series into an AVI or whatever format you like.
Depending on where you live, your images may capture ‘airplane streaks’. They look cool. Lots of things go on at night!
Imaging the sun
Unless you have a reflective ND filter >5, don’t try this. If you have one (mylar survival blankets are supposedly pretty good for this), use the filter and you should be fine. Spectral imaging (the Hydrogen alpha line, Calcium K line, etc) requires extremely narrow spectral filters (angstrom widths), these are usually special temperature-controlled etalons.
A wide angle lens is best. Here, you will need to open the lens as much as possible and increase the ISO to the tolerable limit in order to capture the (relatively) faint meteors. Adjust your exposure time as needed.
Short shutter times (1/1000 s or 1/2000 s) are essential to eliminate motion blur. Advance preparation is key:
Learn the path and find out what the maximum magnitude will be. In the field, use a star of similar magnitude to set the ISO. Don’t bother with mirror-up, since the short exposure time eliminates camera shake. You will need to slew your camera up to 30 degrees/minute over a 120 degree range, make sure your tripod can handle that.
Useful exercise #1: photographing airplanes
Imaging high-altitude aircraft (during the day) will give you practice focusing and exposure settings. The need to track will especially pay off when you try and photograph satellites.
Useful exercise #2: photographing Jupiter’s moons
Jupiter’s 4 inner moons (the Galilean moons) are bright enough to be seen with binoculars, will within the capability of a moderately fast lens. Finding Jupiter (when it is viewable from your location) is easy, and if you can photograph Jupiter at the same time on two consecutive nights (or even several hours apart on the same night) you should be able to resolve the orbital motion of the moons.
Useful exercise #3: crescent moon
The crescent moon (waxing or waning) provides a special challenge: there is only a short window of time that it is visible, and the moon appears very low in the sky, maximizing atmospheric effects. Try various combinations of shutter speeds and ISO settings to see what minimizes ‘seeing’. Also, try intentionally overexposing to capture the un-illuminated portion of the moon- where does that light come from?
Up Next: Part 3: Advanced Astrophotography