DEIMOS Critical Design Review
November 17, 1995

3.1 Optical Changes Since the PDR

The changes in the optical design of DEIMOS since the PDR are minor. These include moving the nominal focus of the focal plane and changes in the glasses used in the camera. Some minor changes also resulted from an optimization study for the shape of the slitmask and the shape of the collimator mirror.

3.2 Movement of the Nominal Focus

During the PDR, the method of inserting and removing the slitmasks from the focal surface was found to involve motions and handoffs that raised serious concerns about reliability. A new approach was developed. However, this required the back focal distance of the Keck telescope to be increased by at least 3.0 inches to obtain adequate room for the slit handler (to clear the bearing into which the DEIMOS instrument is inserted). This 3.0-inch pullback of the instrument causes a degradation of only 0.01" in the rms spotsize diameter. This is discussed further in the Technical Report "DEIMOS Collimator and Slitmask Surface Design" by Brian Sutin, which is included in the Appendix.

3.3 Slitmask Shape

The difficulties associated with making a spherical slitmask surface led to consideration of a cylindrical slitmask. The required shape would be obtained by fitting a flat slitmask into a cylindrical slitmask holder, with the axis of the cylinder perpendicular to the long direction of the DEIMOS field of view. A cylindrical shape was found which fit the overall surface quite well, with a degradation during spectroscopy of about 0.02" in the rms spotsize diameters for the worst parts of the field. This was an acceptable level of degradation, especially given the difficulty of fabricating and handling spherical masks. This is discussed further in the Technical Report "DEIMOS Collimator and Slitmask Surface Design and End-to-End System Performance" in the Appendix.

3.4 Collimator Shape

An exploration of the possible shapes for the collimator mirror showed that the rms image sizes had a strong minimum during imaging for a conic shape with a conic constant of -0.75, an elliptical cross section which is very nearly parabolic. The effect of the conic constant on spectroscopy was much less pronounced, so the collimator shape was chosen for the best images. This is discussed further in the Technical Report "DEIMOS Collimator and Slit Mask Surface Design and End-to-End System Performance" listed in the Appendix.

3.5 Changes in Camera Design

The changes in the camera design are discussed in the technical report "Adopted Preconstruction Optical Design for the DEIMOS 15.0-Inch f/1.29 Camera Lens" listed in the Appendix.

3.6 Overall Performance of the Instrument

In imaging mode, the end-to-end performance of DEIMOS as an imaging device gives worst-case rms spotsize diameters of 0.4" near the edge of the field for both the "V" or "B" bands. For much of the field the rms image diameters are ~ 0.2". The rms image sizes are comparable in its spectroscopic mode. For example, for much of the slit, DEIMOS gives ~ 0.2" - 0.35" spotsize diameters with 0.17" (at the center of the slit), with worst-case spotsize diameters of 0.4" rms at the end of the slit and the end of the spectrum.

3.7 Collimator Mirror

3.7.1 Fabrication and Testing

The DEIMOS collimator mirror, as shown in Figure 3.1, is a 46.3 inch diameter Zerodur concave ellipse. Its radius of curvature is approximately 173 inches, with an edge thickness of about 4 inches and a center thickness of about 2.25 inches. The central 6.5 inch diameter hole is stepped to accept a support flange.

Preliminary work was subcontracted to Eastman Kodak in Rochester, New York because the diameter exceeded our in-house optical machining capacity. This work included diamond generation of the concave surface to a 173 inch radius sphere, machining a flat annulus 0.25 inches wide and perpendicular to the optical axis, and drilling and machining the central hole. The 0.25 inch wide annulus will be used as a reference surface to ease alignment procedures.

The fabrication strategy for the collimator mirror follows traditional approaches for making large aspheric surfaces at Lick. This includes use of Lick's opto-mechanical profilometry for in-process and final certification of the surface shape and figure. After acceptance testing the machined mirror blank, and performing profilometry on the spherical blanchard surface to establish a starting radius, the concave surface will be ground with a 24 inch diameter tile grinder using a standard loose abrasive grinding regime. A sub-diameter tool was chosen because our in-house curve generator has a capacity of only 24 inches. The tile grinder must be pre-shaped to the desired radius prior to use. Aspheric correction will begin after final grinding the radius to the optimum initial sphere, and verification on the profilometer. Generally, this procedure employs the use of various size tile grinders, either rigid or flexible backed, to lower the surface in the areas of departure from a sphere. The collimator mirror has maximum departure from a sphere at the center and at the edge of about 0.0013 inches (33m). The 48 inch Strasbaugh Tilt Spindle Machine will be used with it's oscillating side arm, moving the correction tools in short strokes axisymmetrically on these areas as the mirror rotates under them. This process will take only a matter of days to grind and blend the surface to the approximate desired shape. When the surface is within about 0.010mm peak-to-valley (P-V) of the prescribed asphere, grinding will be suspended and polishing will begin. Our usual approach has been to use flexible laps to polish out and preserve the ground aspheric surface, with monitoring to ensure that the best strategy is being used. Once the surface is polished out, sub-diameter laps are again employed to lower and smooth out figure errors, followed by further smoothing of higher spatial surface ripple with flexible laps. This iterative process will likely take about three months to complete.

The qualifying tool for this surface will be the profilometer. It will be used both for in-process measuring for radius and figure, and for the final certification of the optical surfaces. We are discussing a full diameter interferometric null test also, but have not yet decided if the effort is necessary.

The mirror blank is currently in the Lick Optical Lab, and work has begun on testing and preliminary grinding operations.

3.8 Camera Optics

3.8.1 Fabrication and Testing

The DEIMOS Camera will be fabricated in the Lick Optical Lab. It consists of nine lenses and one window, ranging from 6 inches to 15 inches in diameter, and is made out of seven different optical materials. The camera has 15 spherical surfaces, 3 aspherical surfaces and 2 flat surfaces. The pre-construction design of the camera is shown in Figure 3.2.

Fabrication of a state-of-the-art camera such as the DEIMOS camera is an interactive process between designer, optician, engineers and the astronomers. This process starts at an early stage when there is a fair amount of interaction between the designer and the optical lab in determining what can be made, as was the case with the DEIMOS Camera. The goal is to lead to a design that is state-of-the-art for the astronomer, yet is manufacturable by the optician. While challenging, we feel that the DEIMOS design meets this goal.

3.8.2 Manufacturing

In thinking through the manufacturing of a lens or system, the optician considers difficulties in fabrication and determines ways to deal with those aspects. A number of issues arise in assessing the ability to manufacture a given optical system. Three key areas, materials, shapes and surface figures are discussed below.

1.) Lens Material: One of the first things to consider is materials. What are they, and what's difficult about working with them?

The chart "Glass Properties of Interest to the Optician" (Figure 3.3), answers that question for the DEIMOS Camera. In particular, one needs to look at the "Stain" grouping of the materials in the camera; such groups rate the amount of "resistance" to chemical attack. The higher the stain group the less chemical resistant it is, and the more difficult the material is to handle, grind, polish and maintain a clean surface. (BK7 is at the top to use as a reference, not because it is in this camera. Opticians know how to work with BK7 and often compare it to other materials when evaluating their ease of handling and processing.) The hardness, or more practically, the "softness" of a material helps to determine the abrasive selection, grinding regime, the pitch hardness, polishing compound and slurry mixture, polishing weight, time and figuring techniques. Abrasion is the measure of the weight loss of a standard size sample being ground with 20 m alumina on a 250mm diameter iron flat, located 80mm from the tool center under a standard weight for five minutes. The higher the value the more material was removed.

Using only the resistance to chemicals as a criteria, we have arranged the materials in rough order of descending "ease", with the higher stain group materials, the most difficult, at the bottom. In addition, coefficient of thermal expansion is also a very important factor when considering difficulty of handling. The higher the coefficient, the more likely the material will fracture when exposed to thermal change. FK01Y is a material which experience shows can fracture with temperature changes. It is in the middle of the chart because of its stain group, but it is a very delicate material to work with. Additionally, high thermal coefficient materials require ample settling time after a polishing run to come to equilibrium to ensure valid test results. Of all the materials in the camera, none is more delicate and prone to thermal shock than calcium fluoride, and there are three large lenses made of it! Crystals aren't rated in the same stain categories as the glass in the chart, so we have put it at the bottom and given some properties that are meaningful to the optician. Pyrex, Fused Silica and Zerodur are also included for comparing and to illustrate the (extreme) range of properties faced by opticians. We have had experience working most of these specific materials, and many similar materials, and feel confident with our planned techniques. Figure 3.4 illustrates two recently completed systems with similar materials.

2.) Lens Shapes: Another aspect of a proposed design that catches the optician's eye are lenses that are difficult to fabricate. These include lenses with thin centers, and lenses with thin edges and thick centers. These may be difficult because of support issues during fabrication, or their support later in the cell. A thin-edged lens with a lot of mass in the center may deform at the edge in the test fixture or in its cell. A lens that is thin in the center may require extra attention being given to support during processing and figuring, not merely to make it easier to figure, but possibly to prevent it from breaking while being worked.

3.) Surface Figure: The aspect that receives the most attention when assessing if it is possible to manufacture a lens is the surface shape. For example, steep spherical surfaces may be very hard to hold, figure and test. Means of dealing with these fabrication issues must be addressed. However, the single most influential factor in determining the difficulty and cost of a lens system is the existence of aspheric surfaces. An aspheric surface in a system cuts the fabrication playing field down to only a few dozen optical shops capable of doing the work, as opposed to thousands who can do spherical work. When only one or two such lenses are needed, and the figure on the asphere has to be highly accurate, the choice dwindles to only a few who are either interested in doing short run jobs or are capable of producing the lens. Lick has chosen to specialize in producing difficult, one-of-a-kind aspheric optical surfaces. Aspheric surfaces at Lick also receive the most attention when judging how difficult it will be to make a lens system.

Now that we know what determines difficulty and concern to the optician in reviewing a proposal, we need to know how to handle these different issues, specifically with regard to the DEIMOS camera and its fabrication in the Lick Optical Lab.

3.8.3 Manufacturing Equipment

Optical manufacturing can be divided into six areas of operation. Progressively they are Shaping, Grinding, Polishing, Figuring, Testing, and, if the scope of the work includes full system delivery, Assembly and System Testing.

The "Lick Optical Lab Equipment List" (Figure 3.5) illustrates that we are well equipped to handle most species of optical components. We generally sub-contract machining operations larger than 24 inches in diameter, and usually start our fabrication procedures with loose abrasive grinding. Such is the case with the 46.3 inch collimator mirror. We also sub-contract larger flat surfaces, particularly ones that are not round. There are optical labs equipped with planetary polishers that can do that work at quite reasonable prices. We are well equipped to handle spherical and aspherical surfaces up to 24 inches in diameter from the raw glass blank through final figuring and testing. For DEIMOS we expect to do all the optical fabrication for the camera, plus the remainder of the fabrication on the collimator.

3.8.4 Optical Metrology

Even before unwrapping the raw glass blank from the supplier, we prepare a worksheet of parameters that we must meet and work out the details on how we are going to achieve them. Figure 3.6 titled "Typical Optical Fabrication Tolerances Of Spherical Lenses At Lick Optical Lab" is a list of the parameters that the optician must control to produce a quality lens, and typical tolerances. These can be called "standard optical shop tolerances" because they can be met with a reasonable amount of effort. An amount smaller than those listed can be met, but usually with extra effort, so they are more costly to achieve.

The mechanical dimensions of diameter, axial thickness and wedge are measured with dial gauge fixtures as illustrated in Figure 3.7. We do not have a centering-edging machine to final edge our lenses to ensure that the optical axis is coincident with the mechanical axis, so we set and maintain the wedge from the generation process on through polishing. We do this with a three-ball and post fixture that indicates the edge thickness variations on a lens as it rotates on the balls while being held against the posts on it's mechanical axis. Any thickness difference indicates optical axis displacement. We mark the high of the wedge between the two surfaces and hand grind to correct it to typically < 0.010mm edge run out.

Radius of curvature can be measured in three ways. The first is mechanically with a spherometer. We have a Trioptics hand held digital spherometer with several ring diameters up to 225mm. Figure 3.8 shows this instrument and the measuring principle on a concave surface. The sagitta can be resolved to 0.001mm and a quick calculation or output software gives the radius of curvature. In the worst case (longest radius of 1508.20mm) on the DEIMOS camera lenses the radius can be measured to an accuracy of 0.023% of the radius specified, or a radius difference of 0.358mm. We sometimes need to reach that accuracy or better. We can use the profilometer, which measures to 0.000001" or 25 nm (in sagitta), as a spherometer to certify or cross check any spherical or aspherical radius. The third method is limited to lenses 10 inches in diameter or smaller and uses the Zygo interferometer as a lens bench as illustrated in Figure 3.9, showing interferometric set-ups. This method measures to a resolution of 0.010mm in radius.

For surface figure testing we will use two test methods. The most common, traditional means of determining surface figure error is with the use of an interferometer. We have a static fringe, Mark II Zygo Interferometer fitted with a phase shifting adapter that runs a fringe analysis software package called Zymod. An example of some of the output information possible can be seen in Figure 3.10. We will use this for most of the spherical surface testing. Convex surfaces can pose a problem when being tested interferometrically. The radius that is practical to test is determined by the selection of transmission spheres, and the portion of the surface that can be measured is limited by the design of these transmission spheres. In five cases in the DEIMOS camera lenses, the convex radius combined with the available sphere limits the diameter being tested to about 2 inches. In these cases we will take sub-diameter tests of the surface, overlapping the areas as we skew off center, and then cross check the figure with profilometry tests to ensure test validity. Full aperture tests are possible for many of the concave surfaces since they will be overfilled with the spherical wavefront beam emanating from the Zygo.

3.8.5 Optical Processing

The first shaping operation is usually accomplished on a diamond curve generator using diamond-embedded cup wheels on a tilted spindle head. As the wheel turns on the spindle, the glass blank is fed up into it so that its upward surface is cut into the required convex or concave curve. This machine can also be used as a flat blanchard grinder, and as a lens edging machine. Once a lens is generated into the lens shape, the tools must be made.

Any optical element needs processing tooling to hold it while it is being worked, and to provide a base for grinders and pitch laps that work the surfaces. The tooling print in Figure 3.11 illustrates the form and function of this tooling. The optician specifies the tooling which is then fabricated in the Instrument Lab, usually out of aluminum. There will be over 80 processing tools needed to make the 9 camera lenses. The tool function is included in its nomenclature. In order is the lens element number, followed by the functions of holder, grinder, grinder corrector, lap, or flexible lap and finally the surface number it will be used for. There are eight tools for each element, not including the aspheric plunge tools for the aspheric surfaces.

Typically, we use ceramic tile cemented or pitched onto the aluminum radius tools as the grinding surface for the lens surfaces. These grinders, normally made in matching sets, are generated and ground together to achieve the desired radius. Then the appropriate sign tool is used on one of the lens surfaces in a standard grinding regime to achieve target thickness, wedge and radius values. The other side of the lens is treated in the same manner to yield a lens ready for polishing. A pitch lap is made for each side of the lens, and it is polished to achieve radius, figure and cosmetic surface specification.

When processing a lens with one spherical surface and one aspherical surface, the spherical surface is polished first and protected with a strippable coating. The lens is then placed on its holding tool, with the potential aspheric surface face-up, centered by indicating the edge and held in place peripherally with tape or edge arcs fastened to the tool. The surface is fine ground to the optimum initial sphere radius. We have developed, to a high degree of accuracy, a technique we call "plunge grinding" for making the aspheric shape in surfaces up to 24 inches in diameter. This involves the use of a CNC lathe by the Instrument Lab to cut the matching aspheric correction of the desired surface into a cast iron tool the same diameter as the lens. The tool is grooved and slurry feed holes are drilled. The tool shape is verified by profilometry, and the accuracy that is typically achieved is shown in the plot of the tool profile in Figure 3.12. The tool shape is adjusted to compensate for the anticipated wear of the tool as it grinds in the shape on the glass, so that the final completely-ground-out surface is that desired. A graph of the glass wear vs. tool wear in a recent application is seen in Figure 3.13.

The plunging process is accomplished by centering the tool over the lens surface, feeding the interface with micro-grit, and holding the tool stationary as the lens rotates under it. A maximum aspheric departure (MAD) of 0.020 inches can be ground into a surface in about four hours to a peak-to-valley deviation from the desired shape of around 0.004mm, or within 10 or 15 fringes. Then we either go directly to flexible lap polishing, or touch up the residual figure error with small flexible grinders before polishing.

Our polishing approach is either to use full size flexible laps, or to use rigid laps to "plunge polish" the surface until it is polished out. We sometimes use short strokes to blend in the surface ripple that resulted from the grinding tool. The figuring process may incorporate small flexible laps, or rigid laps, or full size, area-compensated lap designs, as illustrated in the pedal lap diagram in Figure 3.14. The measuring tool we use in figuring aspheric surfaces is our profilometer. It is capable of profile measurements of surfaces up to 60 inches in diameter to a resolution of 0.000001 inch (25nm) in surface height. We have used this instrument as a qualifying tool in figuring both the Keck I and Keck II f/15 secondary mirrors for the Keck telescope, and in making several aspherics for Lick and other observatories. We are confident it will perform as well in qualifying the DEIMOS optics.

Profilometry plots of two typical aspheric surfaces can be seen in Figures 3.15 and 3.16, along with a description of the part and aspheric departure. Figures 3.17 and 3.18 show profilometry plots of a 12.3 inch diameter aspheric lens made here by Gerard Pardeilhan for the Lick sodium guide star laser launch telescope for the Lick 3-meter Shane telescope. The individual scans are seen in Figure 3.17, while Figure 3.18 shows 15 diametral scans on top of one another to illustrate repeatability. Figure 3.19 shows a double-pass, interferometric transmissive test of the completed lens system through the aspheric lens, and Figure 3.20 is the analysis of that test.

Figure 3.21 is a graph of some of the aspherics that have been fabricated in the Lick Optical Lab, showing the aspheric departure vs. the slope tolerance needed (and achieved). The DEIMOS aspherics have been placed only nominally on the graph since the finished design specifications have yet to be finalized. they are comparable in difficulty to those already fabricated in the Optical Lab.

4. MECHANICAL ERROR BUDGET

4.1 This section will be available at the CDR (not on web or in printed report).

5. MECHANICAL DETAILS

5.1 Structural Design

5.1.1 Drive System

DEIMOS will use a friction drive system. The large, 8 foot diameter disk is driven by an 8 inch diameter wheel powered by a servomotor. There is an encoder on the motor for positional feedback, which is occasionally zeroed against a fiducial. The contact loads at the supports are designed so that the contact stresses do not exceed 40 ksi.

5.1.2 Drive Disk

The 8 foot diameter drive disk is a 2 inch thick 4130 steel disk. It is the primary structural element. Its outside edge serves as the frictional surface for the spectrograph rotation drive, as well as the bearing for one end of the cylinder. It is also the primary mounting surface for many of the major components, with several major load groups being attached directly to it. For example, the slitmask handler, the grating slide and the front enclosure are attached to the front side. The tent mirror is mounted in the center, and the camera and the cylindrical shell are bolted to the back surface. The conceptual approach used here is to counter the moments caused by the front-mounted loads with similar, but opposite, moments generated by the back-mounted loads. These moments will have to be balanced carefully to minimize displacements in the drive disk.

5.1.3 Shell

The structure of DEIMOS has evolved into a 72 inch diameter steel cylinder supported on the front end by the drive disk and at the rear by a large bearing. The cylinder is 1/8 inch thick, of welded steel construction, and is modeled as a thin shell. The concept is of a monocoque, stressed skin structure. Loads are distributed into the shell via bulkheads to eliminate point loads. Penetrations are reinforced to carry loads around the opening. Temporary penetrations such as hatches and access panels have connections with removable panels that are capable of transmitting shear loads across the openings.

5.1.4 Carriage

The carriage is the framework that sits on the Nasmyth platform rails and holds the rotating part of the instrument. It contains the rotational drive system, the rear bearing, the movable parts of the kinematic mounts, and the rolling part of the rail system, including the drive motor and gearing.

5.1.5 Principal Loads

The principal loads that have been identified thus far are shown in Figures 5.1 and 5.1.A. Although the diagram shows the magnitudes and locations of the loads, a better idea of the loading can be seen from Figure 5.2 which shows how the structure reacts to the loads. Most of the loads are actually balanced moments attached to the drive disk.

5.2 Storage and Platform Mounting Systems

5.2.1 Rail System

DEIMOS will always remain on the right Nasmyth platform, but it will share that focus with at least one other instrument (NIRSPEC). The NIRSPEC spectrograph may eventually use either Nasmyth platform, but at least for the near future it will share just the right platform with DEIMOS. Since DEIMOS will weigh nearly four tons it will require some sort of heavy handling system. The Nasmyth deck already has a rail system for handling the various Nasmyth and Cassegrain instruments and equipment. The system shown in Figure 5.3 is designed to work with the existing rails. By turning the telescope to the Nasmyth deck and lining up the rails, which are of the same design and gauge, NIRSPEC can be moved on and off the platform as needed. The turntable is provided as a switch, allowing the same track system to be used for both instruments. This system will allow us to stow DEIMOS by rolling it back onto the turntable and turning it onto its stowage siding. Keck has already agreed to provide this track system, including the turntable.

5.2.2 Storage

DEIMOS will be stowed on the rail system wheels in the stowage siding. The rail system has a rack between the rails to which the drive can be locked to prevent unintentional movement. The cable umbilical will remain connected during movement of the instrument. We intend that the DEIMOS will be operational while stowed, allowing maintenance and calibration as well as reducing power up-down cycles. The continued power-up state helps keep the instrument stable, particularly for calibration. In both the stowed and deployed position there will be hydraulic decelerators to prevent jarring at the stops. Restraint clamps will be provided to ensure that DEMOS stays on the platform during any accelerations imposed by earthquakes.

5.2.3 Kinematic Mounts

There is an array of hard points on the platform for positioning DEIMOS in its deployed position. These points are shown as the rectangular array in Figure 5.3. We will attach a set of kinematic mounts either directly or indirectly onto these points. When on its kinematic mounts, DEIMOS will be lifted slightly off its rail wheels.

5.2.4 NIRSPEC Interface

NIRSPEC will have a carriage that will use the same rail system as DEIMOS. There is a stowage siding for NIRSPEC on the opposite side of the platform from DEIMOS. The turntable will act as a switch for both instruments to go from stowed to deployed position. The two storage positions also allow access to the Nasmyth focus for a third (e.g., user) instrument.

5.3 Front Enclosure Hatch & Window

5.3.1 Enclosure and Attachments

This part of DEIMOS differs from most of the rest of the instrument in that there are no parts that require a high degree of precision. Unlike the cylindrical shell, the enclosure is not structural and supports only its own weight, the hatch and the window, none of which must be positioned accurately. The slitmask changer, the filter slide and the TV camera all mount to the drive disk. The enclosure needs to be sturdy and light-tight, but flexure is not a concern. We don't have final designs for the items in this area, so we have only a preliminary outline of what the enclosure will look like, but we expect that it will be made from sheet metal and fastened to the front of the drive disk. It will probably be made in several demountable pieces to provide access.

5.3.2 Hatch

The front hatch will have a swinging door, or perhaps two opposing doors, as shown in Figure 5.4. The hatch covers, in addition to acting as dust covers when DEIMOS is not in service, can act as illumination screens for calibration purposes. The covers will be actuated with pneumatic cylinders.

5.3.3 Lights

We will need lamps to illuminate the hatch screen for instrument setup and testing calibrations. Both continuum sources for flat fields, and arc lamps for wavelength setup and calibration will be needed. The types and numbers of these lamps, as well as their mounting positions, have not yet been worked out.

5.3.4 Window

At the front of DEIMOS, just inside the hatch, we will have a window. The purpose is to keep dust out of the instrument when the hatch is open, and also to contain an inert atmosphere (e.g., from N2 blowoff). This will minimize aging effects on (silver) coatings. The window will possibly be made of coated mylar film. Lawrence Livermore Lab has coated mylar with sol-gel anti-reflective coatings successfully, and so losses at this window will be minimal.

5.4 Slitmask Handling

The plan discussed in the PDR for the slitmask handler has been changed. There will still be ten masks per side. As before they will be stored on the rotating instrument, and can still be inserted in the focal plane remotely. The new storage device has been named the "caterpillar", after a suggestion by Frank Melsheimer. Each side will have one caterpillar. The main block diagram for DEIMOS, Figure 5.1, shows the slitmask handler. Figure 5.5 shows the front view (as light enters the spectrograph) of the two caterpillar devices. Figure 5.6 shows the side view in more detail. The packaging of this mechanism became a problem when it was discovered that the image quality degraded appreciably if the instrument's focus was moved further than 3 inches from the nominal telescope focus. The caterpillar drive lies close to the drive disk, and so its space requirements result in the drive disk needing to be moved towards the rear of the spectrograph. On the other hand, the tent mirror support requires that the drive disk be close to the front of the spectrograph. The center of gravity of most of the elements is also close to the front, reinforcing the need for a forward location of the drive disk.

Fortunately, the caterpillar mechanism could be packaged within the 3 inch maximum setback constraint imposed by the need to minimize the optical image degradation. The design of the chain drive for the slitmask holsters has been tested in a full size model. The chain-to-holster joint was stiffened as a result of this test. The articulation detail is shown in Figure 5.7. The details of the motor drive system remain to be finished. The circulation will be in one direction. This means that the time to go from position 1 to position 10 will be longer than for a bi-directional device. The current estimate is that a maximum of 90 seconds would be the longest wait for any slitmask to be retrieved. This estimate is based on a move interval of 10 seconds between adjacent positions. It is our goal to speed up this operation, if practical. The preliminary design of the actual extraction mechanism for the mask frames has not begun, but it will be a simple linear actuator with latches for grabbing a mask.

The photo in Figure 5.8 shows a slitmask frame (without a skin) retracted halfway into one of the holsters in the full-size model. The caterpillar was judged to be the hardest part of the slitmask handling to design, and so it was built and tested first. After extraction from the holster, each slitmask will be held in a kinematic mount in the focal plane. This can be done at any spectrograph position angle. After a careful study of our error budget, it was decided that the mask mounts would not have to be adjusted after installing a mask frame. This is a simplification of the design presented at the PDR which had three motors to align mask A with mask B, after inserting both masks into the focal plane.

It will not be possible to remotely move one mask from one handler into the opposite location in the focal plane. Therefore care must be given to the manual loading of masks before an observing run. We are planning to use a bar code reader to verify the mask location and to check that the manual loading has been carried out correctly.

5.5 Slitmask Frames

The basic frame shape is shown in Figure 5.9. The cross bars used to support the skin align with the gaps in the CCD mosaic. Figure 5.10 shows a typical slitmask skin with slitlets. Figure 5.11 shows an alignment mask. At the time of the PDR (November 15, 1994) the frame had a spherical shape which would match the telescope focal plane (83.6 inches in radius). The difficulty of fabricating and handling spherical masks led us to re-appraise the use of flat masks that would be held to a cylindrical shape in the frame. After studying the error budget that was found to be practical, and so we re-designed the mask frame to be an optimized cylindrical shape. Brian Sutin optimized the radius and angle of tilt for best image quality. The slitmasks themselves are discussed in more detail in Section 7.

5.6 Grating Slide

The concept for the grating slide has been established: there will be four positions in a linear slide. There will be two slides, one on side A and one on side B. Figure 5.12 shows the layout in the front view of the spectrograph. Each of the positions is different, as shown in Figure 5.12.

• Position 1 (Figure 5.12.A) is for a large 8 inch by 12 inch grating at a large tilt angle (up to 62.5). This position has a motor and on-axis encoder. This rotation stage must rotate to the "home" or grating-installation angle of 44. This is normally the position for the 1200mm-1 red grating.
• Position 2 (Figure 5.12.B) is for the flat mirror which is not motorized. The mirror is about 7 inches by 8 inches. This mirror may be removed and replaced with one at a slightly different tilt which moves the FOV center for drift-scanning.
• Position 3 (Figure 5.13.C) is for large 8 inch by 12 inch gratings at smaller tilt angles (up to 45.5), and 6 inch by 8 inch gratings with similar tilt angles.
• Position 4 (Figure 5.13.D) is for the 6 inch by 8 inch gratings only. They have tilts up to 38.

All grating positions can also accept small flat mirrors for narrow band imaging. These have small tilts within the range from the flat mirror tilt to 29.

There will be counterweights which move on a linear slide in the opposite direction to the grating slide. The approximate weight will be 180 pounds. These will reduce the motor loads, as well as maintaining the instrument balance (although this turns out to be a small factor).

The gratings can be manually changed, and so the rulings on each grating will need a protective cover that is inserted before change-out. Finally, upon completion of a grating change, the manual locks must be removed so that the computer may resume control of the mechanism.

The steps in grating-changing are:

1. Rotate the instrument to the position angle for grating changes. Both side A and side B will be done from the same side.
2. Engage the manual lock for the instrument rotator. Concurrent with the engagement of the rotator lock (and for all manual locks) a switch is actuated that notifies the computer that the lock has been applied to ensure that accidental motor motions cannot occur.
3. Position the grating service cart next to the instrument. Since the first step will be to remove an existing grating, an empty slot must be chosen and aligned with the access slot.
4. Drive the grating slide to the proper position.
5. Engage the manual lock for the grating slide.
6. Rotate the grating of interest to its "home" angle using its rotator motor. Now the grating sub-cell will line up with the removal slot and removal door.
7. Engage the manual lock for the grating rotator.
8. Open the access door. This door is part of the sealing system that keeps the interior of DEIMOS clean and dry, and so care should be used not to contaminate this area.
9. Unlock the mechanism which holds the grating and sub-cell into the grating rotator mechanism.
10. Insert the grating cover.
11. Withdraw the old grating from the rotator, through the guide slot and into the service cart. Since the largest gratings are quite heavy, wheels may be needed to help do this. Dampers will probably be utilized in the cart to prevent large deceleration loads. The grating will be 56 inches from the floor, and so consideration will be given to what is needed to lower or move the grating into the storage position in the cart to eliminate any tendency of the cart to tip over.
12. Re-position the service cart so that the new grating now lines up with the access slot and the new empty position in the grating rotator.
13. Install the new grating and sub-cell, making sure the cover is in place. Operate the manual clamps. Remove the cover, and store it in the cart.
14. Follow the above steps in reverse order until the instrument is again ready for use.
15. All of the above steps apply to narrow-band-imaging mirrors as well as to gratings.

General Notes: · It is our plan to have all gratings for side A fit into side B as well. · The grating service cart will have a clean interior and have sealing doors just like DEMOS. · All necessary hand tools will be kept on the service cart. · When not in use, the service cart must be constrained and kept in a safe place on the Nasmyth platform so that it does not collide with DEIMOS or roll off the platform. Since some of the gratings may have silver coatings, the grating service cart will need to be nitrogen flushed after being opened.

5.6.1 The Slide Philosophy:

The grating slide is shown schematically in Figure 5.13. The slide motor moves all four units in a linear slide. One option for the slide mechanism is wheels on tracks, as illustrated. The structural support for these tracks is the large drive disk. When the required grating reaches the operating position, the motor stops and locking actuators move into an independent structure which now defines the grating more precisely (Figure 5.13.A). This is an important operation because our deflection tolerance is very small, and the weight of all four gratings and their cells, rotation and encoding systems is over 350 pounds. The structural design will attempt to cancel the bending of the drive disk due to the grating mechanism by the weight of the camera lens cell and CCD dewar, which are located on the opposite side of the drive disk (see Figure 5.1). Each of the three rotating units has its own motor. It is not clear yet if a brake or clamp or lock is needed, but if so, then each unit will be provided with one.

5.7 TV Guide and Acquire

TV guide and acquisition is unchanged from the that outlined for PDR, except that one of the TV cameras has been omitted to save funds. The option to add a second TV when funds become available has been retained. The slit-viewing mode of operation will be implemented. A field lens will be incorporated into the TV-guiding reticle assembly, but the details of the TV relay lens design have been deferred pending completion of higher-priority optical activities. The key elements are:

1. The center line of the TV field of view is about 9' off axis. This is to have better rotational control than on-axis viewing. The trade-off is that the image quality is poorer than on-axis.
2. Incoming light hits a 45 pick-off mirror about 6 inches ahead of the focal plane. This mirror is approximately 4 inches by 6 inches. This is followed by a 45 folding mirror that is about 6 inches x 8 inches.
3. The relay lens is next. This could be the same as that used by HIRES, namely a Canon 200mm f/1.8, or it could be a Nikon 135mm lens.
4. The folded field is focused onto a transparent reticle. This will have an illuminated grid, where the illumination is remotely variable in intensity. The field lens is mounted close to the reticle, and reduces the required diameter of the relay lens. This has the great advantage that it will probably make it practical to utilize a commercial lens (as in #3 above, even with the large field specified).
5. The CCD camera follows, proceeded by an 8-position filterwheel between the lens and the CCD camera.
6. The focus will be remote-controlled with a motor driving the built-in focus stage of the lens.
7. The aperture control will also be remotely controlled with a motor driving this particular function in the lens as well. The Canon lens, for example, has a range of f/22 up to f/1.8.

The pixel scale (with the Canon lens) would be 0.135" per pixel, and so with a 1024 by 1024 CCD operating in full frame mode, the sky coverage will be 2.3' by 2.3' (24 m pixels and 4.08 de-magnification). This is 5.3 square" of sky, and there will thus be no need to move the FOV for more sky coverage. That gives, no X -Y or R-Theta TV stages are required.

The PDR proposal stated that Photometrics was our first choice vendor for the CCD camera, as for the Keck I instruments. Since then, a new system designed and built at Lick may be an option. We will compare the system performance for both the Lick and Photometrics cameras before deciding. The software impact, as well as sparing will also be a factor in our decision.

5.8 TV Layout

Figure 5.14 shows the TV layout. This differs from that shown at the PDR in that the current baseline is for only one TV. However we are allowing for a second camera in the design should the funds become available. The far right panel of the figure shows the operation of a two- mirror mechanism which lets the stationary TV camera look at a slitmask with a long single slit. This mask will have reflective areas around the slit for guiding. With the two-mirror mechanism in the slit-viewing mode, the slitmask on the other side will be partly obscured. The reticle will not be used in this mode, but the slit edges will serve the same purpose. The single slit will be located in the unvignetted area as close to the optical axis of the telescope as practical; provisionally the center of the long slit is about 4' from the optical axis.

5.9 Collimator Mirror Cell

5.9.1 Mirror Support

It was shown through finite element analysis that deflections in the collimator mirror will be less if it is supported from the hole in the middle rather than from three points on the edge. We will use a counterweight to provide 90% of the support to the center, regardless of the rotated position. There are also radial supports, but these supply only 10% of the load bearing. The radial supports provide all of the radial positioning. There are two radial hard points and one spring-loaded radial soft point. The center support is a counterweight which will move slightly as DEIMOS rotates, and thus cannot be used for positioning. There are also three axial supports, but these are used primarily for positioning and carry the weight of the collimator mirror only when the cell is removed. Retainer springs are used to keep the mirror against the hard points without overconstraining it. The cell is supported by a bulkhead which is welded to the cylindrical outer shell of DEIMOS. The mirror and its support system are shown in Figure 5.18.

5.9.2 Mirror Alignment

The mirror is positioned radially by the adjustment screws on the radial supports. These are 1 inch screws with a pitch of 40 threads per inch. The axial supports have similar screws. The glass has a 1/4 inch wide reference flat ground into the front surface just outside the clear aperture. This flat was made perpendicular to the optical axis at the time of generation by Kodak. We can use this flat to physically measure the tilt/tip with respect to a datum on the instrument centerline. A reference surface was also ground in the center by Kodak, which can be used to measure and adjust the concentricity with respect to the centerline datum.

5.9.3 Access/Removal

The mirror will be removed from DEIMOS while still in its cell. There will be a removable hatch in the side of the shell. DEIMOS will be rotated so that the hatch is on top. After the hatch cover is removed, a crane will be attached to lifting points on the cell. The taper pins which ensure the positional accuracy of the cell are removed, followed by the bolts, and the cell is lifted from the instrument. The mirror is removed from the cell by laying the cell flat, removing the front stiffening flanges, unbolting the radial soft point and pushing the mirror up through holes in the back of the cell.

5.10 Tent Mirror & Support

5.10.1 Mirror Design

There is insufficient room for a mirror cell, so the tent mirror will be supported by special mounts bonded to pockets machined into the sides of the glass. The rear corners have been cut off since no light will fall on that area and the weight savings were desirable. The mirror has been made thick enough (2 inches) to prevent out-of-tolerance deflections. The front edge comes to a wedge to accommodate an opposing mirror and a 1-1/2 inch diameter hole will be put in the glass around the DEIMOS axis to allow a sight line through the instrument for alignment purposes. Zerodur was chosen for its small coefficient of thermal expansion.

5.10.2 Support

The mirror is held by two flex pivots which define an axis through the center of gravity. By holding the mirror at the center of gravity, only small transient gravity loads will be placed on the third defining point which is the piezo actuator. The flex pivots are designed to be stiff in all directions, except for rotation about the "Y" axis (see Figure 5.19). The flex pivots will in turn be held by kinematic mounts so that the mirror can be removed for re-silvering without realigning the instrument. The mirror can be aligned in the "X" and "Z" directions by adjustment screws which will move the kinematic mounts. These screws can be locked when satisfactory alignment has been achieved. Adjustment in the "Y" direction is not necessary since it is in the plane of the mirror. Rotation about "X" and "Z" will be accomplished by differentially adjusting the two "X" and "Z" screws. Gross adjustment about "Y" can be done with a screw, thread on the piezo mount. Fine adjustment will be done with the piezo actuator. Most of the mirrors weight is supported by mounts fastened directly to the drive disk. The piezos will be attached to a 4 inch x 4 inch x 30 inch steel stiffening bridge bolted to the center of the drive disk. In addition to serving as a solid mounting surface, this bridge reinforces the drive disk which has numerous cutouts near the center.

5.10.3 Flexure Compensation Drive.

Steering the beam about the "Y" axis with the tent mirror is part of the flexure compens tion system. This system is described in greater detail in Section 5.13 and in Section 4.

5.11 Camera Cell Design

5.11.1 Cell Design

Since all of the cameral lenses are round, the camera cell is naturally cylindrical. The largest lens, element number three, will be held in place by the camera cell, but the other elements will be held in subcells. Element number nine, which follows the shutter and filter wheel, is considered part of the dewar and will be part of that design. The camera cell is made from a series of cylinders which will be bolted together. This is partially due to available material sizes, but results from the procedure being developed for assembly of the camera. The lenses and subcells will be held axially by machined surfaces on one side and springs on the opposite side to avoid over-constraining the lenses. Each lens will be held radially by two adjustable hard points and one soft point. There will be three thin (0.006 inch) resilient pads to prevent glass-to-metal contact on the axial defining surfaces.

5.11.2 Subcells

The optical design has two doublets and a triplet which will require couplant. All three of the lens groups will have their own subcell each of which will have its own seals, etc. Each individual lens will be held radially and separated axially from its neighbor(s) by shims of the required thickness. The subcells will be able to be removed from the camera without disturbing the couplant or lens alignment. Access holes will be needed in the side of the main cell so that radial adjustments can be made to each lens while the subcell is mounted in the camera.

5.11.3 Couplant Seals.

The cell design is heavily influenced by the type of couplant used in the doublets and triplet. After considerable research on different types of couplants, the current approach is to use a liquid. A realistic test using a potting-gel material (Silgard 528) to couple two elements of appropriately different coefficients of thermal expansion, a large calcium fluoride element and a heavy flint, failed rather dramatically. It was subsequently realized that the inability of such materials to accommodate volume change under differential expansion of the lens surfaces largely precluded their use. This will be discussed in more detail at the CDR. Fluid couplants present a number of problems, but none that appear insurmountable. In particular, fluid couplants require careful sealing to prevent the couplant from leaking out. The most promising design approach so far appears to be to seal the glass-to-metal rings using an RTV type of sealant, and then sealing the rings to an outer cylinder with rubber O-rings. The design will require us to be able to thermally match the glass with the metal rings to minimize the stresses on the glass due to thermal expansion.

In addition, to allow for thermal expansion of the couplant itself, each subcell will require a volume compensator. This is a flexible reservoir made from a metal bellows. As the fluid expands, it flows into the volume compensator which in turn expands to accommodate the larger fluid volume. Each subcell will also have two bleeder valves that will allow us to fill the cavity with couplant and bleed off the air. There will be little or no air left in the cavity. As noted, this critical area will be discussed in more detail at the CDR.

5.11.4 Support

The camera and dewar assembly, along with the shutter will be mounted as a single unit. The camera will be mounted to two flex pivots that will define an axis through the center of gravity. The pivots will be supported by mounts cantilevered from the drive disk. The cantilever will balance other loads on the opposite side of the drive disk. As for the tent mirror, the flex pivots are part of the flexure compensation system. A piezo-electric actuator will be the third mounting point for the camera/dewar unit.

5.12 Filter Wheel

5.12.1 Filters

The filters will be held in a wheel, arranged as shown in Figure 5.15. There are numerous advantages of this system over a slide mechanism. There are seven filter positions, each one six inches in diameter. One of the locations is a "clear" position with a coated window. This is to ensure that there is minimal focus shift, but more importantly, that no image degradation occurs. The filters are optical elements and are part of the camera design.

5.12.2 Support

The filter wheel rotates about a virtual center and is supported by a thin, ring bearing. This design provides a large area in the center for the dewar mount. Part of the mount for the dewar to the camera reaches through the opening in the center of the wheel, thus stiffening quite substantially the very important mechanical link between the dewar and camera. The bearing is held in place by a mount fastened to the adjacent bulkhead (see Figure 5.15).

5.12.3 Drive

The wheel is driven through an internal ring gear mounted on the inner diameter of the wheel. The motor is mounted to the adjacent bulkhead. The motor is a servo motor of the same type as the other servo motors used throughout DEIMOS. There will be a series of optical switches to give positional feedback to the control system.

5.12.4 Decoupling

The filter wheel is not physically attached to the camera/dewar assembly. This removes the mass of the filter wheel unit from the main dewar/camera assembly. Since there may be potential light leaks around the filters, an opaque shroud may be required to seal off light from this part of the camera/dewar system.

5.12.5 Access & Changing

There is a cover over the part of the wheel that penetrates through the cylindrical shell. Each of the filters is accessible through a slot in the cover. Five of the positions will have "permanent" filters that are the basic imaging set and a nominal order-blocker, while two will be typically selected by the user. However, all the filter holders will be fabricated to be identical, and so a much larger number could be considered "user" filter slots.

5.13 Focusing and Active Flexure Compensation

5.13.1 Focusing the Dewar

Focussing will be done by moving the dewar with respect to the camera. The shutter and the filter wheel are not moved, only the dewar itself. We don't yet have a firm design for either the camera or dewar so specific details have not yet been worked out. There are two areas that reach through and around the filter wheel. These areas are for mounting the dewar to the camera. All of the necessary parts for a focussing stage, including a slide, drive motor, and positional sensors will (and can) be located within these two areas.

5.13.2 Flexure Compensation Overview

Since DEIMOS is a rotating instrument, the gravity loads will be constantly changing. Although every effort will be made to make the instrument as stiff as possible there will always be some transient residual flexure of the components. In anticipation of this problem, we have included an active flexure compensation system. It is possible, even likely, that testing will prove that an active system is unnecessary, but we will have a relatively low cost solution if needed.

The closed-loop approach is straightforward and quite cost effective. There will be a hole in the slitmask which we will illuminate as needed to check for flexure. The image of this spot will wander on the detector as the instrument flexes. The system will steer this spot back to its desired position. This requires active control over two degrees of freedom. One axis can be controlled by tipping the tent mirror (see Figure 5.16). Tilting the camera controls the orthogonal motion. These motions will be small (our estimates are that 42" will be adequate). The actuators we have chosen for this task are piezo-electric actuators which have a range of about 90 m with positional accuracies at the sub-micron level. These low voltage piezos produce much less heat than most. The piezos are capable of adjusting the image position many times per second if necessary (we are designing for 10 hertz).

5.14 Thermal Analysis

A preliminary analysis of the thermal sensitivity has shown us that if we insulate and thermally de-couple the structure from its surrounding, we are likely to have nearly isothermal conditions inside the instrument. We want to keep the interior as isothermal as possible to avoid differential expansion. We estimate that gradients should not exceed 1C to ensure repeatable, stable calibrations.

The maximum temperature differential that we could envisage occurring across the instrument from top to bottom is 11C, driven largely by the cold dark sky as a thermal sink. In order to do a basic analysis we assumed the temperature distribution would be a thermal gradient of 5C from the outside top of the instrument into the body, 1C through the instrument body itself, and 5C from the inside bottom to the outside bottom of the instrument. The instrument is assumed to be surrounded by 2 inches of insulation. This model is shown schematically in Figure 5.17.

Our analysis shows that the heat flow from top to bottom through the 2 inches of insulation over the 6 feet between the drive disk and the collimator would be of the order of 11 watts. Heat can also flow through the interior of instrument by conduction and radiation. Convective heat flow is difficult to evaluate and was assumed to be small. The heat conducted around the inside of the instrument through the steel tube was calculated to be 1W. The radiative heat transfer between the top and bottom, assuming a 1C difference, is of the order of 15 watts. The heat flow out the bottom would be the same as is incoming through the top due to symmetry.

The heat flow through the instrument must balance the heat flow in and out. The analysis indicates that the equilibrium condition inside would come to support a gradient of only about 0.5 C, even with the extreme condition of an 11 degree gradient across the outside. Any convection would help in reducing the internal thermal time constant and also lessen this already small differential.

The drive disk on the end potentially provides a path for a significant amount of thermal energy input into the instrument. Given the thermal gradient in this case alone, 13W could be flowing through it, raising temperatures across its internal surface area well above or below the 1C goal. To the extent possible we will insulate the sides of the disk and reduce the area that can couple to the surrounding environment.

Basic calculations show that, given our planned level of insulation and the mass, that the thermal constant is in the order of 50 hours.

We plan to actively sink any heat that we introduce into the instrument from any electrical/ electronic sources. The electronics, TV cameras, and possibly the calibration lamps will all be housed in thermally controlled areas of the instrument, with the heat being transferred to the telescope glycol system. None of the motors or actuators will run long enough to introduce more than a fraction of a watt.

6. ELECTRONIC / INSTRUMENT CONTROL

6.1 Electronics

The electronics for DEIMOS are designed around a modular architecture. This architecture minimizes the number of different types of spare parts required for maintenance and allows quick swapping of parts to aid technicians in debugging and servicing the electronics. The architecture also allows for easier access to the subassemblies for servicing, minimizing the "down time" of the telescope. The DEIMOS electronics can be broken into six categories: computer communications; CCD electronics; mechanical stage control; AC wiring; and environmental monitoring.

The instrumentation computer communicates to the different mechanical stages of DEIMOS via two Lantronix terminal servers (one for each beam). Each terminal server allows communication from the ethernet link to four independent RS-232 ports, for a total of eight serial communication ports for the DEIMOS stages. The CCD electronics, which have not been finalized, are connected through a fiber-optic link to a VME chassis that is connected to the instrumentation computer through an ethernet link. The guide camera is (tentatively) connected directly to the computer through a separate fiber-optic cable. Figure 6.1 gives an overall view of the instrumentation computer's interfaces.

The baseline electronics for driving the CCDs and for signal acquisition is expected to be handled by a new, updated version of the Leach CCD Controller, developed by Bob Leach at San Diego State University. Leach systems are in use for LRIS and HIRES and are expected to be used for ESI, so the continuing use of these systems will minimize support requirements for CARA personnel. The updated version is expected to meet the demanding requirements of our multiple CCD system. The guide camera for DEIMOS is tentatively a Lick-developed camera (but the performance cost and interface trade-off have not yet been done between the Lick and the Photometric TV systems and so this is still an open item). All of the electronics associated with these systems are either in design, under development, and/or evaluation and so a number of issues remain to be resolved before final designs are available.

Galil motor controllers and drive amplifiers are used to drive the various mechanical stages of DEIMOS. Earlier versions of Galil systems have been used on HIRES and MOS, and have proven to be very reliable, with only one drive-axis failure to date. The latest version of the Galil motor controller, the DMC-1500, has been tested in a cold chamber and is being used for prototyping control, and has shown no problems or failures. Each DEIMOS beam utilizes two Galil motor controllers that are connected to two of the terminal server serial ports. The different mechanical stages are driven by Galil motor amplifiers that receive their inputs from the Galil motor controller unit. A preliminary configuration for one beam is shown in Figure 6.2. Note that up to eight stages can be driven by one Galil motor controller. For the layout in Figure 6.2, one side of DEIMOS has 13 stages: TV filter; TV focus; TV aperture; slit view mirror; "X" slitmask; "Y" slitmask; filter slide; grating tilt #1; grating tilt #2; grating tilt #3; grating tilt #4; grating slide; and CCD focus. The number of stages would approximately double in a final, two beam, two TV configuration for DEIMOS. There is also a provision added to the Galil motor amplifiers that allows direct movement of any stage via a hand-held "paddlebox" facilitating the testing and servicing of the stage electronics.

The piezo actuator controller maintains alignment of the tent mirror and the camera through two piezo devices, controlling both the "X" and "Y" axes. The controller communicates with the instrumentation computer through an Optomux analog rack that is connected to the third serial port of the terminal servers.

Each one of the mechanical stages consists of a servo motor, a fiducial for sensing wheel position, and a motor encoder. Depending upon the stage, both primary and secondary hardware limit switches are also incorporated. The four grating stages have an additional encoder for high resolution sensing of the grating position.

The AC wiring, as shown in Figure 6.3, is that for the electronics vault and one of two beams of the DEIMOS instrument. The vault will house two Uninterruptable Power Supplies (UPS) per beam. One will power the Leach CCD Controller, while the other will power the Galil/ instrument controller. The AC wiring and the equipment in the vault are identified in Figure 6.3. Again, we are using components that allow us to use the same spares as provided for the instrument controller.

The environment of the vault and the instrument electronic enclosures (mounted to the periphery of the spectrograph) are monitored and controlled using the same approach. A Galil motor controller in concert with an analog input card will monitor enclosure temperature, coolant temperature, and ambient/dome temperature. The instrument control software will moderate the cooling each enclosure by monitoring the temperature input and controlling a set of cooling fans. The fans will move air around the electronics and across a heat exchanger that is connected to the Keck II coolant supply system. In the case of a failure of the cooling system, there is a thermostat in each instrument controller enclosure, and in the vault, that will de-activate a control relay upon reaching a set temperature, disconnecting the power to all the electronics in that particular enclosure. Removing the power can also be done through software, but requires a manual push-button switch to reset the power. The extra channels on the vault controller are used to monitor the coolant level in each of the two CCD dewars and the 50 liter dewar. In the instrument controller, extra channels are used to monitor the two servo motor power supplies and the external logic power supply.

Some miscellaneous functions are also performed by the vault controller. It monitors the status of each one of the UPSs. It controls an air solenoid, that is used for pressurizing the 50 liter liquid nitrogen dewar when filling the CCD dewars. It also controls a lamp, that indicates that an observation is in progress, allowing maintenance personnel to determine locally if the spectrograph is in use. The only motion that is controlled by the vault Galil controller is the spectrograph rotation. Schematically, this is the same as any of the previously mentioned stages except that the drive motor will be larger.

7. Slitmask

7.1 Summary/Introduction

As noted in Section 5, we have changed how we plan to cut slitmasks since the PDR. At the time of the PDR we thought the best way to do this would be to use a laser cutter. However, we were disappointed by both the edge roughness and the speed that the material could be cut with the laser systems. On the other hand, experiments at Lick with our NC mill with a high speed chuck and small diameter tools proved that we could cut masks relatively easily with a very well established technology. In addition, the cost of an NC machine is less than a third the typical cost of a laser cutter, and the risk of it not meeting the required performance is much less.

Since the PDR we also changed the shape of the slitmask from spherical to cylindrical. This allows us to use much simplified manufacturing and mounting processes. This section covers the relative differences of the laser cutter and NC processes, describes the mask and frame, and looks at the expected costs and timescales.

7.2 Laser Cutter vs. NC Machine

7.2.1 Cutting Speed

Generally the cutting speed of the laser cutters is slower than the mill. All laser suppliers thought they could meet our minimum requirements of one mask in 20 minutes, but that our minimum time would likely be at their maximum speed. None of the suppliers contacted had the correct equipment to cut slits, but thought they could make a custom machine for our application. The major issues that needed to be researched were the power of the laser and its cost, edge smoothness, cutting speed and the cost and frequency of maintenance.

The NC mill with the high speed spindle cut a DEIMOS-sized mask in 15 minutes without optimization of the tool paths or cutting speeds. With some additional testing and development it is thought that the time could be reduced to 10 minutes.

7.2.2 Knowledge of Technology

Although not new to some in industry and to CFHT, using a laser to cut slitmasks would be a new technology for us and for CARA. The equipment would generally have to be custom built. Combined with the limited experience base in our organization, as well as in the astronomical community, there would be a potentially long learning curve for such equipment.

We currently have a NC machine that we have used for several years. Most of its problems are well known to us, and we continue to use the machine as part of our normal work load. If an NC machine were to be used to cut masks in Hawaii, we would be able to give technical support to those operating the machine.

7.2.3 Edge Roughness

One of the key issues for slitmasks is edge roughness. This appears to be a problem with laser cutting. Our goal is to have the edge roughness (i.e., jaggedness of the edges of the cut slit) be less than 1% of the slit width. The same specification also applies to larger-scale width variations along the slit. For a slit that is 0.5" wide (the smallest that is likely to get much use), the slit width would be 0.015 inches (0.36mm), the desired edge roughness and width variation is 0.00015 inches (~ 4 m).

We examined samples cut by several systems, namely Advance Recording Technologies (ART), Directed Light, Canada France Hawaii Telescope (CFHT), the LRIS punch, and the Lick NC mill. The results are shown in Table 1.

Table 1: Edge Roughness inches Edge Roughness microns DEIMOS Desired 0.00015 4 A. R. T 0.00125 32 Directed Light 0.0027 68 CFHT 0.0008 20 LRIS Punch 0.002 50 Lick NC Slow Spindle Speed 0.00054 14 Lick NC High Spindle Speed 0.0002 5 It is thought that the principle cause of variations in the edge of the slits cut with the high speed spindle at Lick is the way it is held down. We are planning to order a vacuum vise that will hold the piece more securely.

7.2.4 Cost

We have estimated the cost of the laser cutter to be about $160,000. As this would be a custom machine, the actual cost could vary considerably from this estimate, depending on the details of the actual machine specified. It is entirely possible that the machine we would need for our application would cost as much as $200,000. From our discussions with vendors there appears to be little chance it would cost less than our estimated amount.

A good NC machine for our application with a high speed spindle and vacuum vise would cost about $50,000. The market is quite competitive for these machines and we know of several that would meet our specifications. Since such NC machines are "production" items, warranties, service, sharing and maintenance are much less of a concern and will have lower operational cost than a laser cutter. The end mills come in a wide range of sizes with 0.015 inch tools being about $15; larger sizes even cheaper. We were initially concerned about the wear rate, but our first 0.030 inch end mill (0.8") lasted for over 30 LRIS masks.

7.2.5 Maintenance

For a laser cutter the laser head has a known life of about 2,000 hours. The cost to replace this item will depend on the size of the laser chosen for our application, but would likely be on the order of $4,000 to $5,000. As we have no experience with laser cutters, the other maintenance issues are not well known to us.

The maintenance history for NC machines has been very good. They need to be maintained in typical machine shop condition, as we have maintained our current machine. It is about 15 years old, and other than the recent replacement of some of its control cards, it has required very little in the way of maintenance expenses. It cut the demonstration masks and continues to be used on a daily basis. The high speed spindle does have a life expectancy of about 2,000 hours, and would cost $1,500 to rebuild or $3,000 to replace. (This represents more than 5,000 DEIMOS masks and so the cost per mask is comparable to the tool cost per mask, together they come to ~ $1 - $2 per mask.)

7.2.6 Safety

We would need to have the supplier of the laser cutter train both our initial operators and our maintenance personnel. It will require that special safety shields and eye protection be worn by all personnel in the room when in operation. When the cutter is being used, the area will likely need to be locked due to the class of the laser and the required safety precautions.

We will need clear safety shields for the NC machine that is used in Hawaii. People near the machine will not need to take any particular precautions. Servicing the machine will not require any particular area precautions and, except for the general mechanical/electrical safety procedures that are well understood, nothing special is required.

7.2.7 Floor Space Requirements

The laser would need to be in a room at least 10 feet by 15 feet that can be locked to control access.

The NC machine can be kept in a typical machine shop environment and will need a floor space of about 10' square.

7.3 Slitmask

7.3.1 Dimensions

The slitmask is planned to be 29 inches long by 8.5 inches wide with some extra length and width to allow mounting on the frame. It will be cut flat and then bent into a cylindrical shape with about an 83.5 inch cylindrical radius. See Figure 7.1 for an example. It is expected that the mask material will be inexpensive 0.010" aluminum that is easy to procure, handle and store.

7.3.2 Frame

The frame will bend the mask into a cylindrical shape to simplify manufacture and mounting of the slitmasks. We will fabricate about 40 frames initially (double the number needed for a full loading of both beams in the two-beam instrument). The project will be able to request more if required.

7.3.3 Location and Relocation On Frame

The plan is to cut the masks flat and then install them using locating pins on the frames. The goal is to be able to dismount and re-mount the mask at a later date. Life testing needs to be done to verify that this is practical.

7.3.4 Materials

We plan to build the mask frames out of aluminum. If we were to use a laser cutter, we would likely need to use stainless steel of about 0.005 inches thick for the mask, since the laser energy is absorbed by stainless better than by aluminum. However, aluminum would be our choice for the mask material when using an NC mill to cut the masks. For aluminum the NC machine could cut stainless, but the cutter life would be shorter by some unknown amount. The thermal expansion expected in the course of one night would be in the order of 0.0005 inches (5C temperature change) across a mask perpendicular to the slits. This corresponds to a negligibly small change in slit-to-slit positions (only a 0.02" shift).

7.4 Life History of a Slitmask

7.4.1 Cost For Basic Material

The aluminum is expected to cost on the order of $2.50 for each mask and $10 each if stainless steel is used. If an NC machine is used, the end mills (cutter) cost about $10 - $15 each (depending on size) and we would expect to get in the order of 10 masks per mill cutting aluminum. The expectation is that we would only get ~ 3 masks per cutter if we used stainless.

7.4.2 Time To Cut and Handle Mask

The NC machine demonstration mask was cut in about 15 minutes, plus about 5 minutes to mount and dismount the material on the machine for a total of about, 20 minutes per mask. For the worst case night where we might integrate for 1 hour per pointing plus set up we would expect to use 8 masks per side, or 16 total. To cut 16 masks would take a total of 5 hours 20 minutes. Typically the number would be 1/2 to 1/3 of this.

We expect that mounting each mask on a frame will take 5 minutes, and that loading all the masks in the cart will take another 10 minutes for a total of 1 hour and 40 minutes.

Loading and de-loading DEIMOS masks in the instrument in expected to take 20 minutes per side. It will require the instrument to be rotated to the appropriate loading position, locked in place, the previous night's masks removed and the masks for the current night installed in the handler. The instrument will need to be rotated to load the second side.

7.4.3 ID Numbers and Bar Coding

We plan to bar code the masks at the time of cutting, and log the mask and its field at that time. When installed in the instrument, DEIMOS will automatically read the bar codes on each mask in the handlers and record them for selection by the user.

7.4.4 Storage and Re-use Of Masks

The masks are being designed for re-installation into a frame. We are currently doing tests to determine how well the masks relocate, and the normal amount of use the mask will take before it no longer adequately relocates into a frame.

8. SCHEDULE

The basic project schedule for two beams is shown as Figure 8.1. We are still planning for first light through the spectrograph and telescope for the double beam instrument in August 1998, as we were at the PDR one year ago. The major change in the schedule presented last year is that for the fabrication of the optics. We elected not to proceed immediately after the PDR with the purchase of optical materials. Based on the then current camera design, a further iteration on the design was needed. In addition we undertook to improve the blue response of the camera to better match the spectrograph wavelength coverage to that of a silvered telescope (as is planned for Keck II). Harland undertook (very successfully) the very difficult task of improving the image quality and throughput in the blue, while ensuring that it could be built with available glasses. By March the iterations on the camera design to match available glasses from Ohara was completed, and the materials were ordered in April. We are expecting delivery of the glass from Ohara in early December and the start of fabrication in January, after the camera has been re-optimized based on the actual melt sheet data.

To partially compensate for this change in plan we acquired the collimator blank earlier than originally planned. The blank was generated to the basic shape by Kodak and delivered to Lick in late September. Fine grinding has now started at Lick. We are planning to complete this mirror by February 1996, just after the work on the camera optics begins.

The critical path for the two beam project (Figure 8.2) is through the camera optical fabrication. David Hilyard has reviewed the estimates, and at this time we believe the schedule is realistic and that we will meet the planned commissioning schedule. Fabricating optics is unpredictable at some level however, especially for aspheric surfaces, and we will need to monitor the progress closely. Figure 8.3 shows the general plan for fabrication of optics and Figure 8.4 is a historgram of the effort planned for the Optical group.

The current engineering effort (see Figures 8.5 & 8.6) on the project is currently concentrated on the structure, slitmask handler, and grating cell holder and changer. The design of the collimator cell has been completed and the cell fabricated. Due to the problems that arose with optical coupling the camera elements, some additional work is currently being carried out on the mechanical design of the camera. As noted previously this effort is to determine the feasibility of using alternate couplants, such as fluids and greases, and to clarify the design and operational impact of their use. It is not expected to delay significantly the start of major mechanical fabrication, however.

The start of the design and fabrication cycle of the dewar has been delayed from what was originally planned, largely due to other activities in the Engineering Lab. We currently plan to start design of the structure for the CCD mosaic in February, and of the dewar assembly in March, resulting in fabrication of the dewar by the end of the first quarter of 1997. This is a slip of a little more than one quarter from the original schedule, but it is not expected to have an effect on the overall project schedule.

The Instrument Lab has fabricated the prototype shutter and the collimator mirror cell and are currently working on the prototypes for the slitmask handler, and for the grating cells and slide.

The major fabrication effort on the project is expected to start during the 2nd quarter of 1996 with the frame and skin for the support structure of DEIMOS. It is planned that the fabrication of the slitmask handler would also start about that time (see Figures 8.7 & 8.8).

As discussed above in Section 7, we have identified a very economical way to produce slitmasks of the required quantity. This approach, using an NC mill also imposes minimal operational loads on CARA compared to the previous approach. Soon after the CDR we plan to prepare an order for an NC mill that will be used for a major part of the fabrication effort on DEIMOS, for development of the software and procedures for mask making and ultimately be used to provide for mask production capability.

The software efforts are still not well-defined, and so their schedule is not presented here. A software PDR is planned for early 1996. The detailed software schedule will be presented at this time.

Based on the original schedule we had hoped to be able to make the decision to go ahead with the construction of both beams by June 1995. This was primarily based on the optics fabrication schedule. Since the optics fabrication schedule has slipped we are currently showing this date as January 1996. The actual date is still likely to be determined by the optical fabrication schedule of the camera, but we expect that a final decision will need to be made in the first half of 1996, and preferably in the first quarter. Specifically, we would like to know if we are fabricating two sets of optics for mid-1998, or if one set would be installed at a later date. This has significance both in the sequence that optics would be fabricated, and in the scheduling of other projects into the optics lab. The fabrication of the slitmask handler is scheduled to start in the first quarter of 1996, and the grating slide and cells in the second quarter. We will need to know if we are building one or two of each of these items, meaning that there is at best only one quarter of float in the time that the decision to fabricate one or two beams can be made. If funds are not established by early 1996 we will default to building only one beam for 1998, with the other being installed in Hawaii at a later date.

The schedule to deliver one beam is shown as Figure 8.9. Fabricating, assembling and testing the parts for only one beam will likely result in a reduction in the delivery schedule of the instrument of about six months, to nominally the first quarter of 1998. The critical path is no longer through the camera optics, but rather the detectors and dewars (see Figure 8.10). There is only a couple of months of float in this schedule, and if the detectors are removed from it, the engineering/instrument fabrication activities become the critical path.

We have not yet considered in detail what might be entailed in installing the second beam on the instrument after it has been delivered to Hawaii, or the increase in overall fabrication time that results from building the two beams separately. Clearly such a change will entail a net increase in cost as well. Assuming that the assemblies for the second beam are tested before shipping, an initial guess is that DEIMOS would be out of service for up to three months to allow integration and testing of the second beam.

9. BUDGET

The total predicted cost of DEIMOS is estimated to be $5,028,000. This figure is approximately $210,000 less than estimated at the PDR. We subsequently found that a commitment for some of the funding could not be met, and that a small allocation had already been spend on development activities. The contingency fund was decreased to reflect this change. Table 9.1 is a summary of the current budget as of September 30, 1995, while Table 9.2 shows the detail. In the Appendix there is a re-sorted version of this table which breaks many of the costs down by instrument component.

The original contingency fund was $680,000 and was revised to $470,000 in January 1995, as a result of the decreased funding. Although the estimated project total has remained constant, there have been several revisions in the detail of the budget that have resulted in the contingency fund further decreasing to a low of $407,000, and then gradually increasing to its present level of approximately $490,000.

The major unexpected (and unbudgeted) costs occurred because we decided to purchase optical glass for both beams to ensure they are the same. In addition, there was concerns about the long-term availability of some of the glasses, and that glass prices could increase substantially because of new environmental constraints on production of certain materials. In total we spent $130,000 more on optical glass for the camera(s) than was originally budgeted. To offset this increase we elected to eliminate one of the two TV cameras proposed for the instrument, to eliminate the window, and to reduce the budgets for optical machinery and optical materials. Implicit in doing this was a decision to fabricate the optical elements in a serial fashion rather than in parallel as allowed for in the original estimate. The net effect of the increase in the cost of optical glass, and the decrease in other areas was a deficit of $30,000.

Other major unanticipated costs included hiring a structural consultant, and the cost of the piezo electronics for the active flexure control which together totalled approximately $30,000. Several other areas were also found to be underestimated. The resulting total comes to $80,000, including production of the slitmasks, the cost of the model, and the cost of reviews.

Due largely to a re-estimate of the electrical engineering and fabrication, and a greater use of fabricated components, the manpower estimate for the project has been reduced by about 1,500 hours. This results in an estimated decrease in the cost of labor of approximately $65,000. We have also decided to discontinue our evaluation efforts on laser cutters for slitmask production. The use of a standard NC machine with high speed spindle and vacuum hold-down will result in an estimated savings of $95,000.

The net result is that the current contingency fund for the project stands at a little under 10% of the total estimated project cost, or about 12% of the current unexpended budget.

We currently have expended approximately $900,000 to the end of September 1995. Our original estimate had us spending about $200,000 more than this as is shown in Figure 9.1. The difference is in manpower costs.

Although the materials and supplies acquired to date have been somewhat different in detail than originally planned, we have spent approximately the predicted amount of $500,000 (see Figure 9.2) as of September 30, 1995. The major material expenditures on the project to date are the optical materials at $260,000, optical machinery and supplies at $100,000, the collimator at $50,000, the piezos for the active flexture compensation at $16,000, the commitment to consulting costs for the structure at $12,000, and the cost of reviews at $11,000.

The difference in manpower expenditure is largely due to the fact that we have not yet started the intensive software effort or started the optics fabrication (see Figure 9.3). The net result is that we have expended about a third less manpower than we would have expected at this point. The overall manpower estimate remains the same, however.

APPENDIX