Equipment and tools

Goals: Low cost, off the shelf, user friendly

Some of the traditional barriers to digitization include cost of equipment/tools and operator expertise. Our research focuses on the use of financially accessible digitization hardware and tools that can be easily purchased at big box electronics stores or on the Internet. Additionally, we aim to use applications and interfaces that are easy to use, moving beyond financial accessibility to technical accessibility as well.

Scanner: Epson Expressions 12000XL

The Yerkes team works with an Epson Expressions 12000XL (hereafter 12000XL) graphic arts scanner with a Transparency Unit. The 12000XL is a large format flatbed scanner with the ability to scan opaque material and transparencies when utilizing a transparency hood. The platen accepts material up to 12.2 x 17.2 inches, and scans up to 16-bit depth in grayscale and 48-bit depth in color using a color CCD line sensor and an LED lamp light source. We have scanned plates as large as 12 x 12 inches. The 12000XL can be purchased for ~$3,500 through a variety of outlets. It is the same model scanner that the Digitization Department at The University of Chicago uses for archival digitization. Others have evaluated the characteristics of similar scanner for plate digitization (Simcoe 2009, Yuldoshev 2019, Yuldoshev 2016, Sokolovsky 2014) and we build on their findings to use the scanner to produce scientifically viable outputs.

Off the shelf hardware that is not customized for scanning astronomical glass plates has limitations. We explored the response of the scanner to dark-to-light transitions, an important feature for measuring negative images of the sky where the dark sky appears “white” (light gray) and the stars appear black. Utilizing a reseau for the images, research demonstrates that the 12000XL is capable of measuring features at a scale significantly smaller than 50 microns.

Reseau scanner test

Test of the response of the scanner (raw pixel value) to narrow, closely spaced lines on the réseau. The lines on the left-hand side are spaced by 50 microns, or about 5 samples at the scanner setting of 2400 dpi.

We have also tested to see if scanning in a completely dark environment impacts the output in any significant way. Ambient light around the scanner does not seem to influence scanning results. We continue to explore the impact of ambient light within the scanner itself by blocking the platen around the plate with black poster board and determining if this adjustment impacts the output file in any way.

Additionally, the 12000XL does not produce a 1:1 relationship between input (color on the plate) and output (digital file) density. Utilizing a calibrated Stouffer 21-step density wedge, we determined that the scanner does not accurately capture the highest densities. While the scanner specs claim to measure a maximum density of 3.8, in practice we are able to measure densities up to ~2.94 before seeing degradation in results. However, this work is not dependent on the extremes of the equipment, so we are unconcerned with this discrepancy.

We are able to calibrate scanner output to transmission with two free parameters and a scale factor via the transformation equation T= C * [B/(s-P2)]^P1:

T: transmission
B: scan background level
s: the RGB-summed scanner output value for each pixel in the FITS image
P1: corrects for the relationship between the scanner output and the calibrated transmission increments on the step wedge
P2: correction for light scatter in the scanner
C: Scale factor to utilize the full dynamic range

The result of this transformation equation applied to the step wedge demonstrates fall off near Step 16, possibly due to the light reflecting off the darkest parts of the step wedge. Optimization of this equation for applying the transformation to plates is discussed in Process.

Transformation applied to step wedge

The mean values returned by the Epson 12000XL scanner for each step on a Stouffer 21-step density wedge, transformed in three ways to approximate the true photographic density on the wedge.

Finally, the digital output of the 12000XL is plagued by what are affectionately termed “the wiggles.” Other studies (Vicente 2007, Simcoe 2009, among others) have detailed the inability of off the shelf scanning mechanisms to move smoothly and at a consistent pace, and the impact this has on the positional accuracy of the scan output. Wiggles occur due to the left-to-right motion of the mechanism and this mechanical “slippage” results in inconsistent locations of density outputs in the digital file, which are critical to identifying objects, establishing their right ascension and declination, and creating catalogs for scientific use. We applied a simple solution to solve this problem – each plate is scanned twice, with the second scan a 90º rotation of the first. This means the wiggles affect the RA and dec once per scan. Researchers can determine the more accurate scan direction for RA and the more accurate scan direction for dec, then combine them to match and create catalogs. We continue to study if these slippages have any consistency in our machine, both over the short term (scan to scan) and the long term (session to session).

Example of "wiggle" in scanner output

This plot of the data on Plate 10B-161 matches the right ascension on a scan both rotated and non-rotated to the catalog of the entire plate in Gaia. Each dot is a star scanned and matched to its "true" location as determined by the high-precision astrometry of the Gaia catalog adjusted for proper motion. The scanner introduces quasi-periodic residuals in the coordinate oriented in the scanning direction (in this case, RA). We correct for these errors with a sine function, apply them to the original data, and rematch with corrected coordinates with the least residuals.

The 12000XL can be used with several types of software. We have worked with Epson Scan 2 and SilverFast 8, both of which were included with the 12000XL and are both compatible with Windows and Mac OS. The Digitization Department at the University of Chicago Library primarily utilizes Silverfast for their archival scanning purposes so we initially utilized Silverfast as well. As we continued to work with the software, it became apparent that the many options for image manipulation were temperamental and did not always “stick” when we were scanning multiple times over several days. It was also unclear how many of the features impacted the final output and in what ways. To address some of these issues, we transitioned to the Epson Scan 2 software. The interface is simpler than Silverfast 8 but still allows for the tasks plate digitization requires, including focus and turning off all image sharpening, dust removal, and color enhancement. Epson Scan 2 is more aligned to our goals of accessibility, and we have transitioned to using this software exclusively.

Custom built scanning systems such as those at Harvard’s DASCH, the Shanghai Astronomical Observatory, the NAROO Digitization Center at the Paris Observatory, and the D4A at the Royal Observatory of Belgium address many of these issues with commercial scanners in their construction and software, but these setups are also cost-prohibitive for most plate repositories. One must balance the questions of preservation, accessibility, and usefulness with the limitations of off the shelf scanners. When using appropriate scanning methods and processes, the Epson Expression 12000XL graphic arts scanner and Epson Scan 2 software are both financially and technically accessible tools suited for glass plate digitization for archival and scientific purposes.

Camera: Canon EoS 7D Mark II with a 100mm f/2.8 Macro Canon lens

There are several specialty camera systems designed for archival digitization projects, including the Phase One. The University of Chicago Library Digitization Department utilizes a Phase One for some of its projects, and we have begun to test its output against both the Epson Expression 12000XL scanner and an off the shelf DSLR camera. However, while Phase One and comparable systems are excellent for a range of digitization projects, they are also prohibitively expensive for most repositories without special funding. As we continue to refine our processes, we also continue to compare Phase One results against the results of more accessible equipment.

Research on DSLR camera capabilities is ongoing. We utilize a Canon EoS 7D Mark II with a 100mm f/2.8 Macro Canon lens. Unfortunately, this lens is no longer available for purchase, but the key characteristics include the 1:1 maximum magnification and a focal length greater than 50mm. Longer focal lengths allow the camera to be farther away from the plate, which creates less geometrical distortion across the full plate. When selecting a focal length, also take into account the height of your camera stand as well. We would also strongly recommend an auto focus feature, simply to make focus as sharp as possible before turning it off prior to shooting. The camera has a 22.4 x 15.0mm CMOS (Complementary Metal Oxide Semiconductor) sensor, which because of its pixel density, allows for sharper images than a full frame sensor. We shoot with all menu adjustments at neutral or off except for the edge sharpening setting. This is set to “Faithful,” or as minimal as possible. Without this, the images demonstrate volcanism – artificially bright edges on bright stars – due to the internal sharpening which is an excellent feature for most photography, but not for plate digitization.

We capture images in RAW (Canon proprietary file type .CR2, 5472 x 3648 px) and Large with Normal jpegs (also 5742 x 3648 px), the largest file sizes available.To shoot RAW/.CR2, you must also capture jpegs and we use these as quick view images, not working files. We continue to research the specifications of .CR2 files and the specifications of the images when .CR2 files open in ImageJ/FIJI and Photoshop. We also continue to research transformation equations for photography comparable to the transformation equation derived for scanner outputs.

The camera is mounted on a Bencher CopyMate II copy stand with a lightbox similar to the reflecta A3 LED Light Pad placed on the baseboard. We selected this lightbox because it is reasonably priced, highly available, and currently produces workable results. Other lightboxes such as the Kaiser 202492 are built specifically for transparency photography with low vibration technology and more consistently distributed and flicker-free lighting. We continue to study if there is a significant difference in output based on the different lightboxes. As with the scanner, we continue to explore the impact of ambient light from the lightbox by blocking the area around the plate with black poster board and determining if this adjustment impacts the output file in any way.

There are several physical limitations to camera digitization. As a measure of practicality, ambient light is difficult for us to control. We make our environment as dark as possible, but in our particular set up, it is impossible to photograph in a very dark space. While this is not an insurmountable problem, we suspect others will also grapple with this challenge. However, we are confident that if we can obtain reasonable results in this environment, a darker environment would produce better results. Another challenge is sensor size. DSLR cameras have a set sensor size, and in order to achieve scientifically viable results and results comparable to the scanner, this results in a very small field in the output. For sky survey plates, where most of our work has concentrated, this is limiting either in the area of study or the need to take multiple images to capture the full plate. We continue to research the camera as a tool to digitize spectra plates, which have much smaller data regions than sky survey plates, and large plates that are too big to fit onto a scanner platen.

The positioning of the camera is also challenging. Copy stand baseboards are generally smaller than standard lightbox sizes, so we must make sure everything is balanced and leveled. Copy stands work by screwing the camera into the vertical mount, which means that leveling the camera both horizontally and front to back is inconsistent across sessions. It is possible that these physical limitations could be minimized with a camera with a built in level, or a more specialized copy stand.

We continue to research the ideal exposure/aperture settings for photography across different plates and the possibilities of utilizing HDR (high dynamic range) settings on the camera to capture the largest possible range of densities. There is also much work to do on the digitization of spectra plates with the camera. Many of the challenges posed by the sky survey plates (size, wide range of densities, distortion) are less problematic with the spectra plates, and the camera may be a more efficient solution than the scanner for spectra digitization.

Tools and access

For this work we utilize several different types of software. The Scanner section discusses Epson Scan 2 and SilverFast 8. We also utilize the following:

Box: Online file management. This is our de facto server, and is part of a subscription through the University. Google Drive or Dropbox could be other online alternatives, and of course independent hard drives could also be used. Of note, produced files are quite large (300+MB), so a significant amount of storage is necessary.

LUNA: Public access. We currently utilize LUNA for public access to scans. LUNA is built to support art history instruction, and we are well aware of its limitations for this type of project. However, we adhere to the philosophy of giving access to as much as possible as soon as possible, and LUNA is an existing tool in the UChicago system that allows us to do so.

Knowledge@UChicago: Research access. The Knowledge database is where we store publicly accessible research data. Also an existing tool in the UChicago system, we use this to share data that does not suit LUNA and needs to be more publicly accessible than Box.

Astronomy tools. We utilize a suite of tools for astronomy work and analysis including ImageJ/FIJI, SAOImageDS9, TOPCAT, Aperture Photometry Tool, Python, and There are a variety of tools available in this realm. We have selected these particular tools because they are freely accessible and both are appropriate and flexible enough for our research goals but understand that each have limitations. We do not make any claims that these are either the only or the best tools available, but our processes and results utilize these specific applications.

Support for this project comes from the National Science Foundation (Grant AST-2101781), University of Chicago College Innovation Fund, John Crerar Foundation, Kathleen and Howard Zar Science Library Fund, Institute on the Formation of Knowledge, and Yerkes Future Foundation.