Sloan Digital Sky Survey Telescope Technical Note 19980203
Russell E. Owen, Matthew J. Buffaloe, R. French Leger, Edward J. Mannery, Walter A. Siegmund, Charles L. Hull and Patrick Waddell
The Sloan Digital Sky Survey fiber feed consists of nine interchangeable plug-plate cartridges that will be plugged during the day and exchanged at night. Each cartridge contains 640 science fibers distributed between two spectrograph fiber slits, one coherent fiber bundle for sky brightness measurement and ten coherent fiber bundles for guiding. One plug-plate cartridge has been assembled. We present throughput measurements of 5220 science fibers.
The Sloan Digital Sky Survey is a project with the primary goal to measure the red shifts of approximately one million galaxies and one hundred thousand quasars in the northern galactic cap. The survey consists of two parts: digital imaging in five colors to provide data for object selection and astrometric position determination, and spectroscopy of the chosen objects. The project will utilize a dedicated 2.5 m telescope with a 3° field of view, located at Apache Point Observatory in the Sacramento Mountains of New Mexico.
Spectra will be obtained using a pair of 320-object fiber-fed spectrographs mounted on the instrument rotator at the Cassegrain focus. Each spectrograph has a wavelength range of 390 to 910 nm with a resolution of 2000. This is achieved using a pair of cameras, red and blue, each with a 2048x2048 pixel Tektronix CCD.
The spectrograph fiber feed consists of nine interchangeable plug-plate cartridges. Each plug-plate cartridge contains 640 optical fibers for science targets and sky spectra plus one coherent fiber bundle for broadband sky brightness measurement and ten coherent fiber bundles for telescope guiding and image quality monitoring. These cartridges will be manually plugged during the day and exchanged as needed at night. The integration time for each field will be approximately 45 minutes. For maximum efficiency, the interchange process has been highly optimized.
As of January, 1998, one plug-plate cartridge has been assembled (except for the coherent bundles) and the remaining cartridges are ready to be assembled. All science fibers have been procured except for some spares. The optical fibers were terminated by a commercial vendor, C Technologies, Inc., in New Jersey. The vendor tested each fiber using a highly automated throughput tester we designed and built. We used a second identical tester to measure most of the fibers ourselves. The results of these measurements are presented below.
Each plug-plate cartridge consists of a cast aluminum housing supporting a plug-plate bending ring and two fiber slit assemblies. The cartridge contains 640 optical fibers and will contain 11 fiber optic bundles for guiding and sky background measurement. The plug-plate bending ring positions the plate and bends it to match the curve of the focal plane (Siegmund, et al. 1998). Each slit assembly is spring mounted to the housing and is automatically kinematically aligned to its spectrograph when the plug-plate cartridge is installed in the telescope.
Plug-Plate Cartridge
The optical fibers for science targets are Polymicro FHP all-silica UV-enhanced step-index fiber. The core diameter is 180 µm, or 3 arc seconds on the sky. The fibers are only 1.8 m long, so absorption is minimal from 390 to 910 nm, the bandpass of the spectrographs. The telescope feeds the fibers an f/5 beam, and the spectrograph accepts f/4, resulting in negligible loss due to focal ratio degradation. The fibers are uncoated. Predicted loss due to end reflections, focal ratio degradation and absorption is 7%.
The science fibers are packaged as harnesses of 20 fibers. For ease of manufacturing, testing and installation the individual fibers are permanently mounted in their harness and are not individually replaceable.
Harness of 20 Science Fibers
Each harness is terminated at the fiber slit in an 8 mm wide stainless steel v-groove block. The exit face of this block is polished flat and has slightly fanned grooves to provide the correct alignment of the output beam from each fiber. This v-groove block is glued to a thin aluminum substrate to form the fiber slit. The resulting slit thickness is 3.5 mm, which results in a 3% loss of light due to obstruction of the collimated beam. A high precision steel positioning jig is used during assembly to align the v-groove blocks along the appropriate curve.
At the other end of the fiber harness the fibers are terminated in individual 3 mm diameter stainless steel ferrules. These ferrules are off-the-shelf components used in standard connectors, and as such are both very high precision and inexpensive. The high precision allows a very close fit between the ferrule and its hole in the plug-plate (Owen, et al. 1994).
The fibers are protected in the plugging region using 3 mm diameter nylon tubing. The tubing terminates below the plug-plate (above it when the cartridge is turned over for plugging) in an aluminum anchor block. The tubing is permanently attached to this anchor block but the fiber is free to slide inside it, allowing for some expansion and stretching of the tubing without harming the optical fiber.
The fiber harness is attached to the plug-plate cartridge in only two places. At one end the v-groove block is glued to the slit plate using a strong, fast setting acrylate adhesive. At the other end the anchor block is bolted to the plug-plate cartridge using a single bolt. This simple system permits easy installation and replacement of the science fibers. Mounting all 640 science fibers in our first plug-plate cartridge took less than two days.
Guiding and sky background measurements are performed using a set of Sumitomo Corp. coherent fiber optic bundles. At the plug-plate, the guide bundles are terminated using the same ferrules as the science fibers (but with a larger diameter central hole). This allows science fiber holes and guide bundle holes to be drilled in the same operation, for maximum precision. At the other end, the fibers are imaged onto a Photometrics 768 x 512 pixel SenSys thermoelectrically cooled CCD camera with 9 µm pixels.
One coherent bundle, 1 mm (16 arc second) in diameter, will be used for sky brightness measurements. Ten coherent bundles will be used for guiding, one with a diameter of 1 mm and nine with a diameter of 0.5 mm. The larger guide bundle will also be used for field acquisition, if necessary. Guide star data from ten stars distributed across the 3° field of view will enable us to actively correct not only pointing but also the angle of the instrument rotator and the plate scale of the telescope. The plate scale is adjustable (over a fairly small range) by moving the primary and secondary mirrors axially. This allows the aluminum plug-plates to be drilled in advance and used over a range of operating temperatures.
The 11 coherent fiber bundles for one plug-plate cartridge will be assembled as a single harness. The overall design of this harness is similar to that of the science fibers. There will be one coherent bundle per anchor block, allowing the anchor blocks to be distributed across the plug-plate. The output end of the coherent bundles are terminated in a puck attached the slit head; this puck is then reimaged onto the guide camera. The reimaging optics and camera are mounted to the spectrograph
Guide Harness Puck
We developed an instrument to measure throughput of the science fibers. We built two identical copies of this instrument, one for the vendor and one for in-house testing. The fiber throughput tester feeds a uniform broadband f/5 beam into the input end of the fiber under test and detecting all output light that falls within an f/4 cone. Absolute calibration is obtained by bringing the output detector to the input source to measure the intensity of the source. Input intensity is measured both immediately before and immediately after measuring the fibers of each harness, and the higher (more pessimistic) source intensity is used to compute throughput. Large variation of input intensity is flagged as an error.
Fiber Tester
Light is supplied by a Newport model 780 tungsten halogen lamp source, feedback stabilized to within a few tenths of a percent intensity. This light is fed to the source through a 5m long, 400 µm core diameter, UV-enhanced all silica fiber from Oriel. The source reimages the output end of the source fiber onto the fiber under test using a Rolyn 5x 0.2NA microscope objective, creating an f/5 beam (verified by measurement) with a uniformity of approximately 10%. An uncoated pellicle beamsplitter directs some of the light reflected from the end of the fiber under test into a microscope eyepiece, allowing the user to verify that the spot is properly centered and focussed.
The detector consists of a precision aperture, to reject all light outside of an f/4 cone, followed by a pair of Newport PAC070 cemented doublet lens and a silicon photodiode with a 2 mm diameter active area. The two lenses reimage the output end of the fiber under test onto the photodiode. The image is slightly broadened to to reduce the effect of variation in sensitivity over the photodiode surface; this is done by placing the photodiode is 1 mm to the right of focus. The photodiode current is measured using a Keithley 485 picoammeter. In addition to all this, there is a 3 mm thick BG-38 filter between the two lenses to flatten the spectrum of the light source.
The grooves of the v-groove block are fanned out, rather than parallel. Hence each fiber must be independently positioned to properly aim at the aperture. This is handled automatically. For each fiber, the control computer moves the linear stage and records intensity vs. position. After all fibers have been measured the computer figures out which fiber is in which groove of the v-groove block, and hence the proper position of the linear stage for each fiber. It then interpolates the associated intensity vs. position curve to determine the intensity at the correct position. Doing this rather than simply taking the peak intensity assures that the tester can detect errors in alignment of the fiber in the groove.
The system is highly automated. The user need only move the detector in front of the source (for input measurements) and plug ferrules into the source when requested. The result is a table of throughputs and a pass/fail indicator displayed and written to a data file.
The fiber vendor was required to measure the throughput of each science fiber, to verify that it met the following specifications: the 20 fibers in a harness must have a mean throughput of at least 90%, and each fiber must have a throughput of at least 87%. These specifications were originally determined by measuring the throughput of prototype fiber harnesses fabricated by several vendors during the development phase and in cooperation with the vendors.
We also measured the throughput of 5220 of the fibers we received, to accurately characterize their throughput. Our results are shown in graphs of throughput for each fiber and mean throughput for each harness. As can be seen, the fibers greatly exceed the throughput specifications.
Throughput for Each Fiber
Mean Throughput for Each Harness
We also compared throughput measurements made at the University of Washington with those reported by C Technologies. These data were acquired with different throughput testers of the same design. Earlier tests indicated that the system used by the vendor consistently measured lower throughputs by a few tenths of a percent. In addition, the throughput measurements are not directly comparable due to the way the fibers were cleaned. Our intent at the University of Washington was to accurately characterize the fibers, so we tested various cleaning techniques and settled on reagent-grade isopropyl alcohol and cotton swabs. In contrast, C Technologies was only asked to verify that the fibers were acceptable, not to characterize them, so C Technologies used the more efficient cleaning technique of wiping the fibers with tissue.
Subtracting C Technologies throughput measurements from ours we obtain the histogram shown below. The agreement is excellent. The mean difference of 0.4% is consistent with the differences between the two testers and the different cleaning regimens discussed above.
UW and C Technologies throughput measurements are compared in more detail on the scatter plot below. The diagonal line shows the ideal case of equal throughput. The vast majority of the points are clustered at the upper right corner, indicating both excellent agreement and excellent fibers. The horizontal spreading of the main clump of data and the scattered points to the left of the clump are probably due to the different cleaning regimens discussed above. The few points scattered lower along the line of equal throughput appear to be genuinely inferior fibers. The very few points along the right hand side and well below the line appear to be fibers we did not clean well. We tested this conjecture by carefully recleaning some of worst examples in an earlier version of the graph and the corresponding throughputs did move above the line.
Figure 5: UW vs. C Technologies Throughput Measurements
The design and manufacturing process for the Sloan Digital Sky Survey fiber feed has been very successful. The science fiber vendor has delivered 5880 optical fibers of excellent quality. The fibers have high throughput and are very easy to install in the plug-plate cartridge. The plug-plate cartridge is robust, provides safe fiber routing and offers generous room for plugging.
Throughputs of 5220 science fibers were measured by both the fiber vendor and ourselves using two automated fiber testers of the same design. The fibers have a mean throughput of 92.0% and a standard deviation of 0.4%. The two fiber testers agree to within 0.4%. We believe this is by far the largest homogeneous data set of astronomical optical fiber throughput measurements reported to date.
We are grateful to C Technologies, Inc. for their excellent work manufacturing and testing the science fiber harnesses.
A shorter version of this paper is published as "Status of the Fiber Feed for the Sloan Digital Sky Survey", R. E. Owen, M. J. Buffaloe, R. French Leger, E. J. Mannery, W. A. Siegmund, P. Waddell and C. L. Hull, in Fiber Optics in Astronomy III, ed. S. Arribas, E. Mediavilla and F. Watson, A.S.P. Conference Series, 152, (Astronomical Society of the Pacific) 1998, p. 98.
Date created: February 3, 1998 Last modified November 16, 1999 Copyright 1998 Russell E. Owen
