Astrophotography is a hobby rapidly gaining popularity thanks to the fast advancing CMOS sensor technology. Over a decade ago, the light recording material employed in astrophotography was primarily chemical emulsion. Its low sensitivity makes it very hard to record the weak signal from deep space. In addition, the lack of real-time feedback is a huge source of frustration for beginners. Operational errors such as out-of-focus can only be realized after several nights of hard work after the film is developed. In the mid 90s, the advent of cooled CCD cameras provided solutions to both the sensitivity and real-time feedback problems. However, their high prices and miserably small sensor areas limited their uses to only a few kinds of astrophotography and to very enthusiastic astrophotographers. While CCDs revolutionized astronomical research, this technology has never really changed the landscape of amateur astrophotography. The true turning point took place in 2002. After Fujifilm announced its FinePix S2Pro DSLR and showcased amazing astronomical pictures taken by this camera, people started to seriously explore DSLRs for astrophotography. DSLRs can provide real-time feedback, which is very important for beginners. They have sensitivities not much worse than CCDs, and DSLRs with large sensors (APS-C) are quite affordable nowadays. Today’s landscape in astrophotography is shaped by a series of CMOS-based DSLRs from Canon, but DSLRs and mirrorless cameras based on Sony sensors are gaining popularity very quickly.
Because of my job, I have opportunities to use a broad range of imaging instruments, from multi-million dollar CCD cameras on large professional telescopes to amateur CCD cameras and DSLRs. My training in astronomical research also provides me toolsets to quantitatively evaluate the performance of sensors and to know their true limits. This helps not only my research, but also my lifetime hobby, astrophotography. On the hobby side, I mostly use DSLRs (Canon 5D Mark II and Nikon D800) for their high performance and affordable prices. To get the best astrophoto results, the DSLRs’ internal filters are modified to have higher throughput in the deep red, so they can be more efficient in recording the red light from ionized hydrogen gas in the universe. Other than this filter modification, DSLRs used for astrophotography are no different from DSLRs we use daily.
One very common worry about using DSLRs on astrophoto is the thermal noise generated by the sensors. CCD cameras cooled to -20 or even -40 degrees C do not have such problems. However, CMOS sensors produced in recent five years all have very low thermal noise. Under the same sensor temperature, their thermal noise is actually much lower than common CCDs in astronomical cameras. Another important factor that many people overlook is noise sources other than heat in the sensor, one of which is the photon noise generated by the sky itself. With latest DLSRs under many circumstances, the sky photon noise often overwhelms the thermal noise, making cooling unnecessary. Only in places that are both hot and dark (such as the deserts in the south-west US), cooling is needed to fully exploit the dark sky.
The workflow in astrophotography is quite different from that in daylight photography. Because our targets are very faint, we need to expose for a few minutes or even a few hours, to collect enough photo-signal from our targets. However, the sky background is usually so high that it will saturate the image when exposure is longer than 10 minutes or so (this is especially true under a light-polluted sky). Therefore, what we do is to break the long exposure into many shorter (a few to 10 minutes) ones to avoid saturation, and then stack (average) the short-exposure images in post-processing to combine their signal. This gives a result that is equivalent to a very long exposure.
On the telescope, once the equatorial mount is set up and aligned to the Polaris, what we usually do is to first use a bright star to focus. This used to be a very challenging task, but now it is very easy with DSLR’s live view function. Then we move our telescope/lens to point at our target. We can usually very easily see our target constellation through the camera’s viewfinder if we use a wide-angle or short telephoto lens. On the other hand, if we use a long telephoto lens or a telescope to shoot deep-sky objects, the targets are usually too faint to be seen directly. Some test short exposures with very high ISO can help to verify our framing. Once this is done, we just fire away many long bulb exposures through a computer or a timer shutter release. As mentioned above, typical exposure times range from a few to 10 minutes, depending on how fast our lens is and how dark the sky is. A very commonly used ISO is 1600. However, with recent DSLRs with Sony sensors, it is possible to use ISO 800 or even 400 and still get very good results after post processing. The advantage of lower ISOs is of course their higher dynamical range. It goes without saying that we always shoot RAW.
In addition to the on-sky exposures, we also take many “calibration” images to remove the unwanted signal from the sky, the optics, and the camera. For example, we take exposures on objects with uniform brightness (such as a cloudless day-time or twilight sky, or a large LED panel) afterward. Such images (called “flat field”) can be used to correct for the vignetting caused by the lens/telescope in the on-sky images, to restore the uniform background brightness. In the beginning or the end of the night, we fully cover the lens/telescope and take “dark” exposures when the camera is under the same temperature as the on-sky shots. Such dark images can be used to remove the thermal signal in the on-sky images. This is essentially the same as most DSLRs’ in-camera long-exposure noise reduction, but we do this manually to avoid wasting the precious night time. We also take extremely short (1/8000 sec) exposures (called “bias”) when the lens is fully covered, to account for whatever signal the camera generates when there is no light and also no time for thermal signal to accumulate. Like the on-sky exposures, we take multiple (from a few to several tens) flat, dark, and bias exposures and average them to beat down any random noise in the images to improve the signal quality. There are many software packages (such as DeepSkyStacker, which is free) that can process the on-sky, flat-field, dark, and bias images, and stack the calibrated on-sky images to form a very deep, clean, and high dynamical range image. All these have to be done from RAW files, as JPEG images are not linear and do not allow for accurate removal of those unwanted signal.
After the basic calibration and image stacking, we use software such as Photoshop to further process the stacked images. It usually takes very strong curve and saturation stretch to bring up the faint details in a stacked astronomical image. It also requires a lot of skills and experience to achieve this while still maintaining accurate color and a natural look of an image. It is essentially like manually processing a RAW image from scratch, without relying on any raw processing engines. It is not uncommon for us to spend more time on processing an image than its exposure time, and post-processing is often what separates a top-notch astrophotophotographers from average ones.
This guest post was contributed by Wei-Hao Wang, an astronomer working in a national research institute of Taiwan, and is currently visiting the Canada-France-Hawaii Telescope on the Big Island of Hawaii. He is also an astrophotographer and started this hobby in 1990. A collection of his recent astrophotos can be found right here.