The drive for smaller pixels comes from wanting more resolution. But in astrophotography, bigger pixels capture more light.
Pixel size is a big consideration when selecting a camera for astrophotography. Smaller pixels have both some inherent advantages and disadvantages over larger pixels, but the truth is that in most things that matter, larger pixels are generally better.
However, CMOS-based sensors for astrophotography are becoming increasingly popular (see my recent article in Sky & Telescope’s May issue on the CCD to CMOS transition). As a result, it’s getting harder and harder to find cameras with larger pixels. Why does this matter?
While doing some volunteer work at a professional observatory some time ago, I overheard one of the astronomers talking about his new camera. “I just love those tiny little 9- micron pixels,” he said. I laughed out loud because in the amateur market, 9-micron pixels are huge by most standards. The bread-and-butter CCD sensors, such as the KAF-8300 or the Sony 694, are floating 4- to 6-micron pixels, and some popular CMOS choices are starting to dive below 3-micron pixels.
If you recall when I wrote about pixel scale and sampling before, it is important to have proper sampling to obtain good data to work with. All those gigantic telescopes you see at a professional observatory have to follow the same physical rules we do (and my astronomer friend was surely taking this into account with that 9-micron pixel camera).
Sure, professional telescopes are located where the seeing conditions are better than you probably experience. But without the use of some sort of adaptive optics, they have the same limitations as you do when you're shooting from your backyard. Big telescopes with long-focal-lengths need really large pixels in order to work effectively, and the same goes for you.
Signal to Noise
All other things being equal, cameras with larger pixels are generally more sensitive to light and have better signal-to-noise characteristics. I’ve talked about this relationship before.
When Sony came out with its A7s mirrorless camera that had such tremendous capability for capturing the Milky Way, people would speculate about what it’s sensitivity secret was. It wasn’t a secret: It had big pixels. Nine microns in fact, whereas most DSLRs or point-and-shoot cameras at the time had pixels at least half that size.
Twice the pixel size is actually four times the surface area for collecting light. The drive to “more megapixels” is driven by a desire for more resolution. Unfortunately for us, there is a big difference between a 5-minute exposure at night through miles of atmosphere and a 1⁄2000-second image of a sunlit hummingbird that is four feet away.
Unfortunately for us, the sensor market cares more about hummingbird photography than it does for imaging galaxies far, far away. For longer focal-length telescopes, you really need larger pixels to achieve proper sampling and a good signal-to-noise-ratio per pixel.
Small Pixels Rule
Small pixels aren't always bad. In fact, there is a place or two in astrophotography to which smaller pixels are very well suited and even preferred. In deep-sky work, a small-aperture, short-focal-length optic is sampled properly with smaller pixels. That fast focal ratio will deliver a great deal of light to those smaller pixels, which mitigates the signal-to-noise benefit that typically comes from larger pixels. (It overcomes some of the inherent signal loss from using a color camera as well.)
Subjectively, I’d say if you are shooting around f/4, give or take, I wouldn't let the signal-to-noise issues with small pixels bother me; I’d focus only on proper sampling.
Smaller pixels in the past also suffered from poor well depth, that is, a low number of the electron storage capacity that's used to count photons. This disadvantage is disappearing as semiconductor technologies improve. The new Sony IMX 455 CMOS sensor, for example, has pixels smaller than 4-microns square but has a full-well capacity of more than 50,000 electrons.
Sampling rules also go out the window when it comes to lucky imaging. Small pixels rule here, as the targets are very bright, and the exposures are very short. Pixel scales as low as 1⁄10 of an arcsecond are not uncommon for these targets (the Moon, Sun, or planets). The slow focal ratio, due to the extreme magnification of the target, is less of an issue because of the target is also bright.
Over-sampling just makes large “mushy” images with a poor signal-to-noise ratio. Under-sampling, however, will generally not ruin an image. What's more, there are dither and drizzle techniques pioneered by Hubble Space Telescope scientists that can help recover some lost resolution in post-processing when this is the case. (For more information on over- and undersampling, see my previous blogs on how to estimate your ideal sampling and compute your pixel scale.)
So, when shopping for your next great CMOS or CCD camera, don’t just focus on read noise and quantum efficiency. Make sure you take the sensor's pixel size into account as well.