Sloan Digital Sky Survey Telescope Technical Note 19940311
A 640 fiber feed for a pair of moderate resolution spectrographs is being constructed for the Sloan Digital Sky Survey. The spectrograph slits, optical fibers and plug-plates are mounted together in removable cartridges. The cartridges protect the optical fibers and provide quick-change interfaces to the telescope and the spectrographs. Ten cartridges will be made to allow plugging to occur during the day.
The fibers are constructed in harnesses of 20. These are mass-terminated in v-groove blocks at the spectrograph slit. At the input end the fibers are terminated in simple ferrules and manually plugged into aluminum plug-plates. The plug-plates are deformed to match the best focal surface of the telescope, and the holes are drilled to be parallel to the principal ray after bending. We describe a simple technique for drilling accurate holes.
The Sloan Digital Sky Survey (SDSS) is a project to measure the red-shift of approximately 106 galaxies and 105 quasars in an area of pi steradians centered at the north galactic pole. The survey consists of two parts: digital imaging in five colors to provide data for object selection and position, and spectroscopy of the chosen objects. The survey will use a dedicated 2.5m azimuth/altitude optical telescope with a 3° field of view, located at Apache Point Observatory in the Sacramento Mountains of New Mexico. The SDSS is a project of the Astrophysical Research Consortium consisting of the Institute for Advanced Study, Johns Hopkins University, Princeton University, University of Chicago, New Mexico State University, University of Washington and Washington State University, and is currently being funded by the first four of these institutions. Fermi National Accelerator Laboratory and the Japanese Promotion Group are participating in the SDSS via memoranda of understanding.
Red-shift will be measured using twin fiber-fed spectrographs which together take 640 simultaneous spectra. The wavelength range is 390 to 910 nm and resolution is 2000. Each spectrograph has a red and blue camera, each with a 2048x2048 Tektronix CCD. The optical fibers are positioned in the focal plane using a manually plugged aluminum plug-plate. The plug-plate is mounted on an instrument rotator at the Cassegrain focus. The spectrographs are attached to the same rotator, on either side of the plug-plate, so the fibers flex very little during exposures.
The two spectrographs are permanently mounted to the back of the telescope. The plug-plate, fibers and two spectrograph slits are mounted together as a removable cartridge. We are making 10 such cartridges, enough for a night's observing. This permits us to plug the plates during the daytime and simply swap cartridges at night. All aspects of cartridge handling--changing, transportation and storage--are being carefully designed for ease of use, speed of cartridge exchange, and reliability.
Figure 1. The spectrographs and a plug-plate cartridge.
We are using Polymicro FHP 180 µm diameter core optical fiber. This is an all-silica, UV-enhanced fiber with a hard thin polyimide buffer layer. Similar fiber has been used for a number of fiber-fed spectrographs, including Hydra (1), Argus (2) and Optopus 2 (3). Our fibers are 2 m long.
The telescope feeds the fibers with an f/5 beam and the spectrograph accepts f/4. Predicted loss due to focal ratio degradation for unstressed fibers at these focal ratios is approximately 3% (4). Dielectric reflections at the ends claim 7% of the light and approximately 1% is absorbed in the fiber. We expect to lose another 2% due to misalignment at the input and output ends, based on our measurements of light loss versus tilt and our specifications for alignment of the fibers. Hence the best throughput we can expect is approximately 87%, but stress in the terminations will likely reduce this.
The project calls for 6400 terminated fibers plus spares. To simplify termination of the fibers and construction of the fiber slit, the fibers are being produced in harnesses of 20 (see figure 2). At the output end each fiber harness has a small v-groove block with the 20 fibers mounted permanently and polished as a unit. At the input end the fibers are individually terminated in stainless steel ferrules.
Figure 2. Harness of 20 optical fibers.
Our plan is to have the fiber harnesses made by an outside vendor. At this moment three vendors are manufacturing prototype harnesses. We will measure the throughput of these prototypes to determine a throughput specification for the production contract. The chosen vendor will measure each fiber of each harness for acceptable throughput using a test apparatus we provide. We will spot-check the vendor's measurements using our own identical fiber throughput tester.
Each plug-plate cartridge includes two spectrograph fiber slit assemblies; these are attached with a flexible linkage so that the two slits and the plug-plate can be independently aligned. After the plug-plate cartridge is attached to the telescope, an air-activated lever in each spectrograph presses precision tooling balls on the slit assembly against kinematic hard points in the spectrograph to align the slit. The system is automatic and no adjustment is required.
The slit is 120 mm long, with fibers spaced on 360 µm centers. The fiber outside diameter is 220 µm, leaving insufficient room to individually terminate the fibers. We did not wish to handle bare fibers because they are fragile and difficult to polish flat and normal to their axis. So we chose to mass terminate the fibers in v-groove blocks, 20 fibers to a block. The end of each block is polished flat, and the grooves are fanned slightly--the proper amount so that (taking into account refraction) light is emitted normal to the curve of the slit. Each slit will have 16 blocks, or 320 fibers.
The v-groove blocks are attached to a stainless steel sheet called the slit substrate (see figure 3). Each block is aligned by pressing it against an alignment jig, a piece of metal accurately machined to the desired curve of the slit. The jig is removed once all the blocks are mounted. The block to block spacing is controlled using a calibrated wire as a spacer. The blocks are attached to the slit substrate using cyanoacrilate. A small hole in the slit under each block facilitates replacement: a custom tool clamps down the two neighboring blocks while forcing the desired block off the slit substrate.
Figure 3. V-groove blocks mounted to the slit substrate.
Behind the v-groove block the fibers are bundled in thin-wall tubing, 10 fibers to a bundle. Molded RTV grooves route the bundles along the slit. Once across the flexible linkage and inside the plug plate cartridge assembly the fibers run bare (unjacketed except for the polyimide buffer) until they near the plug-plate. At the plug-plate the fibers are handled during plugging, and the fibers must be sturdily jacketed for protection. The jacket also provides a torque which cocks the ferrule, holding it into the plug-plate against gravity. Our current plan is to use nylon tubing as our jacketing material. Nylon is stiff, strong and resistant to crushing and kinking. It is also available in bright colors, which aides manual plugging.
The transition between bare and jacketed fiber occurs at the anchor block. This is a heavy piece of metal which positions the fibers below the plug-plate and absorbs the stresses of plugging. The jacketing, not the fibers, are attached to the anchor block. Each fiber is free to move in or out of its jacketing as tension on the jacket or temperature dictates, and there is generous slack in the bare fiber beyond the anchor block to accommodate this motion.
At the input end each fiber is terminated in a stainless steel ferrule. These ferrules are manually inserted in holes in the aluminum plug-plate. The plug-plates are 652 mm in diameter (active area) by 3 mm thick and made of aluminum. They are mounted in a bending jig to fit them to the curved focal plane. The center of the plate is translated by only 3 mm, so the bending is purely elastic. The maximum angle of the principal ray with respect to the normal of the focal plane is approximately 3°. To maximize throughput the holes are drilled at the angle of the principal ray. The ferrules have a tip diameter tolerance of ±5 µm, and the holes will be drilled with a diameter error of 4 µm RMS (see below). This permits a very close fit, yielding good alignment of the ferrule and high throughput.
Each ferrule must last for approximately 300 plugging operations. We conducted a wear test (5) and found that our stainless steel ferrules wear less than 2 µm in diameter in 300 plugging operations into aluminum, which is perfectly acceptable.
Drilling costs for the plug-plates are a significant cost of the project so it is important to drill the holes as simply as possible. On the other hand accurate holes are necessary for good throughput. We conducted a series of tests to determine how accurately we could drill holes in the plug-plates using various basic techniques (6). The results are summarized in table 1. The best holes were obtained using a spade drill; this yielded holes with a diameter standard deviation of 3.9 µm. The holes were 2.2 mm in diameter, drilled in 3 mm thick aluminum plates, and had entry angles ranging from 0 to 1°. The full diameter bits were made to a tolerance of +0/-1 µm and held by a collet machined in place in the milling machine to reduce runout. The undersized bits used before reaming were of conventional accuracy and held in a conventional collet. All bits were carbide steel. Our final optical design requires the plates to be drilled at angles of up to 3°, so we are currently measuring a new set of test plates drilled at steeper entry angles. The new tests also evaluate the possibility of starting the hole with a center drill before drilling.
Method
Twist Drill
7.5
1.3
5.6
5.7
10.1
3.2
Spade Drill
9.1
5
3.9
6.3
7
2
End Mill
74.7
55.9
93.2
7.8
1.6
Drill & Ream
6.9
4.7
7.3
8.7
8.4
3
Mill & Ream
9.7
11.1
12.2
10.2
6.1
A paper similar in content to this note is published as "Fiber feed for the SDSS spectrograph", R. E. Owen, W. A. Siegmund, S. Limmongkol and C. L. Hull, Proc. of S.P.I.E., 2198, 1994, p.110.
Date created: 3/11/94 Last modified: 4/16/99 Copyright © 1999, Russell E. Owen Russell E. Owen
owen@astro.washington.edu
