Blanco Primary Mirror Radial Support Repair

Transcription

1 Blanco Primary Mirror Radial Support Repair ACTR Project #450 Project description. Project Manager: Tim Abbott Project Engineer: Andres Montane Date:5 Feb 2008 Note: this document was abstracted from section 2 of the Dark Energy Survey subproject CTIO Facilities Improvement Project (CFIP) Technical Design Report (TDR) which may be found in the ACTR DocDB as document #77. Page 1

2 Table of Contents Executive Summary... 3 Repair of the primary mirror support system... 4 Four Unexpected Discoveries The primary mirror moves on its cell Radial support attachment to the primary mirror is marginal Radial supports bend Misalignments have accumulated with each radial support repair The plan for a repair... 8 Page 2

3 Executive Summary The Blanco telescope primary mirror radial support system consists of 24 passive actuators distributed around the mirror circumference. As the telescope moves around the sky and the gravity vector changes with respect to the primary mirror, the support system applies forces to the mirror to support it and restrain movement. Ever since the telescope was built there has been a problem of supports breaking their glued bond to the mirror. Often, breaking off the bond damages the glass which consequently must be repaired. Loss of a support allows the primary mirror to move in a limited, but uncontrolled manner with respect to the prime focus corrector, with negative consequences for image quality. Re-gluing a broken-off support is an exacting procedure that requires removal of the primary mirror from the telescope. A detailed analysis of the support system has pointed to design problems as the root cause of the failures, exacerbated by the high alignment accuracy of the support positioning that is needed to make a marginal system perform semi-reliably. By-products of this analysis has been creation of an FEA model for the telescope tube assembly, and the measurements of flexure between the primary mirror and the prime focus. The repair involves re-building all 24 actuators, correcting the design flaws, and positioning the actuators accurately on virgin glass, which will be obtained by rotating the primary mirror about its axis by 5 degrees. It will guarantee that the primary mirror will be correctly supported and positioned and thus meet the technical specifications of DECam. Page 3

4 Repair of the primary mirror support system The Blanco primary mirror is supported at its edge by 24 gravity-driven levers (Figure 1) designed to spread their load appropriately to minimize distortions of the mirror s surface when the telescope moves away from the zenith. While they do not directly constrain motions of the primary on its cell, these supports are intended to be balanced to minimize the possibility of such motions (which are then limited by friction with the cell). FIGURE 1 LEFT: SCHEMATIC OF RADIAL SUPPORTS DISTRIBUTED AROUND THE PRIMARY MIRROR. RIGHT: TWO RADIAL SUPPORTS ON EITHER SIDE OF ONE SEISMIC CLIP ON THE SIDE OF THE BLANCO PRIMARY MIRROR From time to time, a support has become detached from the primary mirror. Under those circumstances, the primary mirror will then move around on its cell introducing coma into the delivered images. While it is possible to compensate for these movements by active control of, for example, the prime focus alignment, these motions are inherently uncontrolled and dependent on the current and recent positioning of the telescope. Moreover, uncontrolled movements of any significant part of the telescope structure represent a threat to its continued health. Therefore, through a series of shutdowns and investigations, CTIO has been developing its understanding of the mechanical properties of this system of radial supports with a view to repair in which broken supports become a rarity rather than the norm. This understanding has been hard won. It is now clear that the radial support system is more vulnerable to internal misalignment than was believed when the first supports began to fail. It is also clear that the original understanding of the radial support design was incomplete. Page 4

5 Four Unexpected Discoveries Four unexpected discoveries were made in the course of this work: 1. The primary mirror moves on its cell. The original telescope design incorporated four radially constraining load cells between the central chimney and the primary mirror. Early tests showed that the chimney flexed with respect to the primary mirror sufficiently to break these load cells and they were removed. This left the primary mirror unrestrained against radial motions on its cell except for friction against the axial load cells and its balance with the radial support system. It was apparently believed that this would be sufficient. However, once a radial support fails, the balance in the system is lost and the mirror is free to move on the cell. Ultimately, it is limited in these motions by the seismic clips at four points around the mirror which act as hard limits. In practice, the motions are limited by the lateral slack available in the radial support mechanisms. This amounts to about 1mm. If that 1mm of free space is used up on any given radial support, it is in danger of bearing a significant portion of the weight of the mirror in a direction in which it is not designed to do so, and it may bind up. Both of these conditions may in turn induce strain beyond the designed limits on its epoxy joint with the primary mirror and cause it to fail. 0.8 N(mm) laser pole 07 Mitut.pole 07 laser pole 06 Mitut.pole 06 4hr West hr East 6hr East 6hr West hr East E (mm) 2hr West FIGURE 2 MOVEMENT OF THE PRIMARY MIRROR ON ITS CELL AND WITH RESPECT TO THE PRIMARY MIRROR. THE TELESCOPE WAS POINTED AT THE SOUTH POLE, THEN ROTATED 6 HOURS EAST FOLLOWED BY SIX HOURS WEST AND THEN BACK TO THE MERIDIAN. RED AND BLUE LINES INDICATE MOVEMENTS OF THE PRIMARY MIRROR ON ITS CELL. YELLOW AND GREEN LINES INDICATE MOVEMENTS OF THE PRIMARY MIRROR WITH RESPECT TO THE PRIME FOCUS CORRECTOR. MEASUREMENTS MADE OCT Page 5

6 2. Radial support attachment to the primary mirror is marginal. Models have shown that the design of the assembly by which the radial supports are affixed to the primary is intrinsically marginal. Each radial support is attached to the primary mirror via four invar pads which are epoxied to the glass. An H-shaped invar bar (the Hbar, Figure 3) is bolted onto these four pads and the radial support applies traction or pressure to this bar at its center. Finite element analysis (Figure 4) shows that the H-bar must bend; this redistributes the forces under each pad such that when it is at maximum tension, the peeling forces experienced by the epoxy are very close to its limit. FIGURE 3 THE ORIGINAL DESIGN OF THE RADIAL SUPPORT "H-BAR" WITH THE FOUR INVAR PADS SHOWN FIGURE 4 FINITE ELEMENT ANALYSIS SHOWING H-BAR DEFORMATION UNDER LOAD AND PEELING STRESSES INDUCED IN THE INVAR PADS TO WHICH IT IS BOLTED 3. Radial supports bend. When the telescope moves away from the zenith, each radial support bends elastically with gravity causing the counterweight on the end of the lever arm to displace with respect to the telescope superstructure (Figure 5). If, as a result of that bending, the counterweights come into contact with the cell the radial support system will lose balance and the primary mirror will tend to move on the cell. In around 1995 (?) a plenum was installed around the radial support system to help direct cooling air. At high zenith angles, the radial supports will have come into contact with this plenum. The plenum was removed during the 2005 shutdown. During the same shutdown, at least one radial support was observed to come into contact with the cell at high zenith angle. Also, the radial supports were augmented in an effort to achieve a finer balance between the primary mirror and its radial supports, this augmentation will have (slightly) exacerbated the bending of the supports. Once a radial support has failed, the unbalanced mirror will move on the cell in the direction of the gravity vector, lifting the remaining radial supports as it does so, masking the effects of the bending in those supports. Page 6

7 FIGURE 5 DISPLACEMENT OF RADIAL SUPPORT COUNTERWEIGHTS WITH RESPECT TO THE CELL DUE TO FLEXURE IN THE RADIAL SUPPORT, COMPENSATED FOR MIRROR MOVEMENTS ON ITS CELL. Page 7

8 FIGURE 6 MISALIGNMENTS OF RADIAL SUPPORTS, BEFORE AND AFTER SHUTDOWN SUPPORTS MARKED WITH RED ARROWS ARE NOW BROKEN. 4. Misalignments have accumulated with each radial support repair. Failed radial supports have been repaired over the years during re-aluminization shutdowns. Unfortunately, due to misunderstanding and miscommunication, inappropriate techniques have been used to align the repaired supports. As a result, misalignments have accumulated and the radial support system no longer has the intended lateral freedom of motion that was originally built in (Figure 6). Thus any motion of the mirror on its cell is more likely to drive the system into an over-stressed or bound condition. Indeed, in some directions there may be little to no freedom of motion and the radial supports already in a critical condition. Given the design of the telescope and these four conditions, failed supports are inevitable. Without hard restraints on the location of the primary mirror on its cell, some movement is inevitable whether through imbalance, abuse or earthquake. Eventually, one such movement will drive a support into a laterally stressed or a bound condition, causing it to break. Moreover, if the telescope is shocked while at high zenith angle, the supports under maximum tension at the top edge of the mirror are close to the epoxy design limit and may fail. It is instructive to note that most, but not all, the broken supports appear on the southeast side of the mirror and these are the supports which most commonly experience the highest stresses under traction because the prime focus service station is found on the northwest side of the telescope. When a support has failed, it has been re-glued onto the primary with poor alignment. As these misalignments have accumulated, the mirror has come ever closer to a binding or laterally stressed condition, so the frequency of broken radial supports has increased. The Mayall telescope, with its identical design of radial supports, has suffered fewer breaks because the repair techniques used when they break has resulted in smaller misalignments specifically, the Kitt Peak team has used an alignment pin during repairs whose significance was not known to the Cerro Tololo team. The plan for a repair It is not possible to constrain the primary mirror on the cell so that it cannot move. To do so would over-constrain the mirror and probably result in reduced image quality if not damage the telescope. We can only rely on the small friction with the axial load cells and the balance of the radial supports. However, as noted above, some motion is inevitable. How, then, to ameliorate the situation? The following lines of attack present themselves: 1. Modify the system to prevent (or at least reduce) interference between the radial supports and the telescope structure. Page 8

9 2. Re-align the system to maximize the permissible motion without driving the system into an untenable state. 3. Strengthen the join between the radial supports and the primary mirror. 4. Improve the balance between mirror and radial supports. Point (1) has already been partially addressed by removing the plenum so there is now much more space for a radial support to move away from the primary mirror as it bends under gravity (when the telescope moves towards the horizon on that support s side of the mirror). It should also be possible to re-shape the radial support counterweights so that the support can bend further without the counterweight coming into contact with the cell (when the telescope moves away from the horizon on that support s side of the mirror). A redesign of the radial supports to reduce bending is probably not necessary and may even be untenable. Point (2), to re-align the system, is an obvious step, but may not be trivial. The correct reference frame for alignment of the radial supports on the primary mirror is not to space them equally, every 15 degrees around the circumference of the primary. Rather, it is the location of the 24 sockets in the inner circumference of the ring girder of the telescope superstructure which receive the plugs at the top of the radial supports when the mirror is installed in the telescope. We have assumed that they are precisely spaced at every 15 degrees, but we have no definitive evidence that this is the case. Any alignment procedure must start by verifying this assumption if not use the actual reference frame provided by the sockets in order to re-align the supports. Re-aligning the radial supports requires that they be removed and re-glued. At least half of the supports must undergo this treatment and many of these have damaged glass from prior breaks under the invar pads. As such, a more extreme approach is attractive in which the primary mirror is rotated on the cell to present fresh, undamaged glass on which the invar pads may be glued with appropriately precise alignment. There is sufficient space between the supports to permit a 5 turn of the mirror on its cell to reach fresh glass, thus the procedure may be repeated in the future should it become necessary. Page 9

10 FIGURE 7 REDESIGNED RADIAL SUPPORT H-BAR WITH ROTATIONAL FREEDOM OF MOVEMENT ON EACH LEG The join between the radial supports and the primary mirror can be strengthened (Point (3) above) through attention to the design of the H-bars. Indeed, in the shutdown of 2005, four H-bars were replaced with new H-bars which were modified to introduce a rotational freedom in their legs with a hinge between the H-bar and each Invar pad (Figure 7). This redesign significantly reduces the peeling forces experienced by the epoxy join with the mirror and none of those four radial supports have failed in the past year, despite being sited over some of the most damaged glass 1. It has recently been verified that the hinges do not introduce significant backlash into the mechanism. A complete repair plan would therefore involve replacing the remaining 20 H-bars of the original design with the new. Note that the new design of H-bars requires replacing the Invar pads which simplifies the repair by not requiring that we remove the pads which are currently epoxied in place. Finally, achieving a finer balance between the radial supports and the primary mirror (Point (4) above) is a matter of careful measurement and adjustment. Some of this was done in the 2005 shutdown and the counterweights increased slightly. Initial results seen before the radial support broke immediately after the shutdown (#15) suggest that the balance is better. Plans for the repair will include a procedure for verifying this. The telescope shutdown required to make these repairs is anticipated to last 3 weeks. It will be scheduled for semester 2008B or 2009A, giving more than a year for any problems to be resolved before DECam is delivered to CTIO. The work required prior to the shutdown is constituted by the manufacture of the 20 new H-bars and definition of the detailed repair plan. 1 Note: voids in the glass were filled with an epoxy/glass spherule mix to more closely match the properties of the missing glass than epoxy alone. Page 10