From Glass to Silicon: A Short History of Astrophotography
Silver and Glass
John W. Draper made a daguerreotype of the Moon in 1840, the first photograph of a celestial object. Ten years later William Bond and John Adams Whipple turned Harvard's 15-inch refractor on Vega and captured it in a 90-second exposure, the first photograph of any star other than the Sun.
The dry gelatin plates that arrived in the 1870s were more sensitive and far easier to handle than anything before them. Henry Draper used one for a 51-minute exposure of the Orion Nebula in 1880, essentially the first deep-sky astrophoto. The photographic plate was astronomy's memory for the next 100 years, recording positions, brightness, and spectra on glass instead of in sketches and logbooks.
In fact, I remember working with one of the very last observers using photography on Mauna Kea back in the mid-1980s. He actually put an eyepiece in the 88" UH telescope, and we got to stand on the lift and peer down the barrel. It was very cool. Of course, using photography meant he spent all night taking photos and all day developing them. He told us that by the end of his three- or four-day run, he was seeing pink dinosaurs by the side of the road on the way down to Hale Pohaku!
The CCD Arrives
The charge-coupled device (CCD) was invented at Bell Labs in 1969 by Willard Boyle and George Smith. In 1976, Jim Janesick of NASA's Jet Propulsion Laboratory and Brad Smith of the University of Arizona took the first astronomical CCD image, using the 61-inch telescope at Mt. Bigelow. Not too surprisingly, CCDs quickly replaced photographic plates as the professional astronomer's detector of choice.
The leap in image quality was enormous. A photographic emulsion responds unevenly to light, so a bright core quickly saturates and swallows the faint detail around it. A CCD records each pixel on its own, in a near-linear response to the light that lands on it, across a far wider range of brightness. A blazing star and the faint nebulosity beside it could finally live in the same frame without one washing out the other.
The new detectors also needed a common language. Astronomers began shaping the Flexible Image Transport System, or FITS, in 1979, and Wells, Greisen, and Harten published the first standard in 1981. It gave every observatory a single way to package an image together with a header describing how, when, and where it was taken.
Putting a Number on the Seeing
In 1984–85, I spent a year on Mauna Kea with the National New Technology Telescope site survey, using a first-generation CCD camera to make some of the first quantitative FWHM measurements of the mountain's seeing. Before instruments like that, judging the steadiness of the air leaned heavily on the observer's eye. A CCD turned it into a number.
CCDs Come Home
By the mid-1990s, cooled CCD cameras had reached advanced amateurs. Companies like SBIG built cameras such as the ST-4 and ST-6, and for the first time a serious backyard imager could work with the same kind of detector found at professional observatories. FITS came home with them. The standard was formalized as NOST 100-1.0 in 1993 and NOST 100-2.0 in 1999, and it became the default format for amateur capture software just as it was at the big observatories.
The CMOS Revolution
In 2012 and 2013, ZWO's ASI120 helped bring affordable CMOS cameras to amateur astronomers, and CMOS quickly displaced CCDs for planetary imaging. In 2015, Sony announced it would stop producing CCD sensors in March 2017. Today, CMOS sensors dominate amateur deep-sky imaging as well.
FITS kept pace. Version 2.1b added 64-bit integer support in 2005, version 3.0 followed in 2008, and the IAU FITS Working Group approved version 4.0 in 2016, with the final edited text approved in 2018.
Flexible, Maybe Too Flexible
The "F" in FITS stands for Flexible, and the standard lives up to the name. It requires only a handful of header keywords and leaves nearly everything else to whoever writes the file. That freedom let FITS adapt to five decades of new instruments, but it also meant every capture program could record the same information in its own way, under its own keywords, in its own spelling.
Standardizing object names, filter names, and the like across all of them was simply too complex, and as far as I know it was never seriously attempted. That's the chaos Blueshift was built to bring order to.