Документ взят из кэша поисковой машины. Адрес
оригинального документа
: http://www.apo.nmsu.edu/Telescopes/eng.papers/milt/structure.html
Дата изменения: Mon Dec 13 18:49:06 1993 Дата индексирования: Sun Apr 10 00:30:52 2016 Кодировка: |
Item Allowance (arc sec) Wind loading Structure 0.05 rms Control system 0.02 Misc control system error Uncorrelated with wind 0.04 Total tracking error 0.08
In this budget the first and second items are assumed to be correlated and to add algebraically. The third item is summed in quadrature.
The above budget applies for a mean wind speed of 5 m/s. While it is a simple problem to calculate structural deflections due to a constant velocity wind, it is much more difficult to calculate the image degradation due to realistic (dynamic) wind loading. One could imagine doing a full dynamic finite element structural analysis using a sample of experimentally measured wind loading vs. time as the driving function, calculate the image position at each time step and thereby derive a rms spot size. While this would be laborious and would require a lot of computer time, the results would only apply to the particular driving function which was chosen and would not be generally applicable unless an impractically long time history was calculated.
For now we take the simple approach of assuming the rms image motion due to amean wind speed of 5 m/s is equal to the image displacement due to a constant wind velocity of 5 m/s. This is a crude approximation and will no doubt be replaced later with a more satisfactory way of relating mean wind speed to image motion.
A uniform temperature change causes defocusing of the image due to the difference in the coefficients of expansion of glass and steel. Refocusing compensates for this effect. Temperature gradients cause the OSS to distort slightly causing pointing uncertainties. This effect can be largely prevented by designing the OSS to have a high degree of symmetry, by making the wall thicknesses of members equal where appropriate so that cooling rates will be uniform, and by insuring that the members have low infrared emissivity to prevent asymmetries in the IR radiation field from coupling into the OSS.
Wind loading is by far the most troublesome effect, both quasi-static loading which elastically deforms the structure and dynamic effects which tend to excite the natural vibration modes of the structure. Nelson (1983) has summarized the effects of wind loading on image motion.
The effects are the rotation of the optical axis (defined by the primary mirror), the tilt of the secondary with respect to the optical axis, and the translation of the secondary with respect to the optical axis. To this list could be added the displacement of the detector, which is usually ignored since it is very small in reasonable designs.
These motions of the optical elements are very small so their effect on telescope collimation is negligible. Instead the motions produce time varying motion of the stellar image as the wind changes direction and speed. This results in image enlargement for integrating imaging detectors and the loss of light hitting the slit in the case of a slit spectrograph. Nelson gives formulas relating image motion to the motions of the optical elements.
Measurements by Forbes (1983) at the MMT indicate that the top end of the telescope is only slightly protected from wind loading by the MMT building when the telescope is pointed into the wind. Our enclosure is likely to be "efficient" in the same sense as the MMT (it will fit the telescope closely for reasons of economy) and is not likely to provide much shielding either. The wind problem is aggravated because the telescope is likely to be pointed toward the west and southwest preferentially as astronomers chase objects toward the horizon. This is the direction of the prevailing winds at SPO.
Traditionally wind screens have been used to protect telescopes from wind loading. Beckers and Williams (1982) have reported that the MMT wind screens degrade the image quality by preventing the flushing of heat out of the telescope chamber. For this reason the wind screens at the MMT are not normally used.
If use of wind screens must be avoided to preserve image quality then wind loading effects will be very difficult to overcome. (The MMT side steps the problem by essentially building a large telescope out of a number of small and therefore stiff telescopes.)
The problem is compounded by relatively new realization that ground based image quality can be very good and that ground based telescopes should be designed to take advantage of exquisite seeing (Woolf 1982). As a result the overall error budget for image quality is tight and the portion allotted to wind loading is meager.
Our response is to use wind screens to shield the telescope when the mean wind speed is higher than 5 m/s (11 mph). The enclosure design minimizes the amount of heat generated and stored in and around the telescope. We expect to have a greatly reduced need for air flushing as compared with the MMT and the higher wind speeds will provide effective forced ventilation even with wind screens in use.
Meeting the structure performance requirement still means dealing with wind loading effects beyond what is needed in the design of traditional well-protected large telescopes. The top end (including its supporting truss) will have minimal effective area for wind loading constrained by the requirement that the lowest natural frequency of the structure be greater than 10 hertz.
The design of a sufficiently stiff OSS is eased greatly by the choice of a very fast f/ratio for the primary mirror. A still faster f/ratio would be desirable from the standpoint of the structural design but is much more difficult from the optical fabrication point of view. It is likely that f/1.75 is a good compromise between the optical and structural constraints.
New requirements may demand new solutions. Novel approaches under consideration include shielding the members of the upper structure with with close fitting shields attached to the OSS center section thereby greatly reducing the wind loads applied to the truss and the secondary support structure. Alternatively the state variable tracking control system we propose to implement (see section 4-2) can be modified fairly easily compared to a classical control system to accept more inputs. This suggests that wind loads could be sensed with appropriate transducers and pointing errors could be corrected by the control system. Clearly a novel approach is justified only if there is no other cost effective solution, the risk is reasonable, and the requirements cannot be relaxed.
A number of elements of this design require additional explanation. The secondary support vanes are tensioned to a strain of about 0.0002. Though their cross-section is only 150 mm^2 the natural frequency of vibration is high. The vane tension insures that the forces in the vanes always remains tensional. The vanes act to stiffen the peripheral square frame. The vanes extend nonradially from the secondary backup structure; the attachment points are not 90 degrees apart. This adds torsional stiffness to the secondary cell. The MMT secondaries are supported in this manner. Note that the diffraction pattern produced by the four sets of vanes will be degenerate producing four spikes as in traditional telescopes. The spikes will not be at right angles to one another however.
Each axis of the telescope will be driven by a small roller friction coupled to a large (2 m radius) drive disk attached to the driven structure (section 3-4). Friction roller drives are supplanting the traditional large diameter precision spur gears for drive purposes because of the high cost of the latter (~$300k each for the RGO 4.2 m).
The square frame at the top end is also a departure from tradition though the 1.8 meter Steward transit telescope also will have a square top end (McGraw et al. 1982). A square frame is cheaper to fabricate and is a more efficient structure than the traditional circular front ends since the members act as truss elements rather than beams.
The attachment of the truss to the OSS center sections also requires comment. The truss attachment points are at the midpoints of the sides of the center section instead of at the traditional four corners. This change results in very short and direct load paths from the truss to the altitude bearings and drive disk (if the drive disk is mounted at the midline of the back of the mirror cell).
The more traditional geometry with the drive disk mounted near one axle and with the truss mounted at the four corners of the center section was also investigated. In this configuration the center section required much more mass to prevent wind loading on the truss from causing objectionable torsion near the axle opposite the drive disk. In the current design the deformation of the truss is the dominant cause of image motion due to wind induced OSS deformations.
Clearly when the telescope points at the zenith the center section will sag under the static load. Early in our analysis we were concerned that this deformation would propagate to the top end and cause degradation of the performance of the secondary support assembly; the deformation could be sufficient to cause some of the vanes to go into compression where upon they would buckle. Fortunately finite element modeling indicated that in a typical model the tension in the vanes changed by less than 30% in going from the horizon to the zenith, and this is perfectly acceptable.
The present attachment points for the truss results in an efficient envelope for the OSS motion; if the truss were rotated 45 degrees the corners of the square top end would extend well beyond their current positions.
The sag of the primary cell is small; the drive disk can be mounted to partially compensate for the sag in order to minimize the compliance required in the drive and encoding mechanisms.
Traditionally (since the design of the 200 inch) the secondary and primary mirrors of large telescopes have been mounted at opposite ends of a Serrurier truss designed to sag equally at both ends thereby keeping the two mirrors collimated.
For modern fast primaries this strategy is neither practical nor necessary since the center of gravity of the OSS lies very close to the primary mirror. In the case of short telescopes, tube deflection can be kept so small that simply retilting the secondary mirror restores high image quality (Barr et al. 1979).
The following table lists the element types used in the construction of the model. The center section is built of quadrilateral membrane shell elements. It is a honeycomb-like structure with the large inside and outside surfaces separated by orthogonal sets of elements forming six-sided cells. The actual structure would be fabricated in a somewhat different way but a welded plate structure with appropriate stiffeners and webbing should have approximately the behavior predicted by the model. The mirror cell is constructed in a manner similar to the center section. Membrane shell elements were also used to construct the drive disk. Membrane shell elements have no bending stiffness. In the analysis described below the performance of the center section, mirror cell and drive disk should be dominated by the in plane properties of the elements so membrane elements should be appropriate.
OSS Subsection Element Type Mirror cell membrane shell Drive disk membrane shell Center section membrane shell Truss 3-D spar Top ring 3-D beam Secondary vanes 3-D spar Secondary supports membrane shell
The mass of the primary and its support structure is modeled by mass elements at the nodes of the front surface of the mirror cell. Similarly the secondary mass is modeled by a mass element at the central node on the appropriate surface of the secondary mirror support structure. The model includes only the weight of the mirrors. In the real structure the mirror weight also applies a torque to the structure since the mirror CG is well in front of the surface of the mirror cell. A better model would explicitly include these torques but should not change the results described here very much.
Constraints due to bearings are infinitely stiff in the model. These will be replaced by springs of the appropriate stiffness as parameter values become available.
The tertiary supporting structure has not yet been included in the model; consequently all analysis of image motion due to wind loading assumes no motion of the tertiary.
All plates are 6.35 mm (0.25 inch) thick steel except for the drive disk which is 12.7 mm (0.5 inch) thick. The truss elements are 76 mm (3 inch) steel tubes with 12.7 mm (0.5 inch) walls. The top ring elements are 76 mm (3 inch) steel tube with 6.35 mm (0.25 inch) walls. The spider vanes have across-sectional area of 150 mm^2 (0.23 square inches) each. The mass of the primary mirror and its supports is astatic supports is 3150 kg and the secondary mirror mass is 44 kg.
The mass of the half model including mirrors is 4980 kg and its moment of inertia about the altitude axis is 17400 kg m^2.
Three loading cases are described below.
Symmetric boundary conditions at the symmetry plane were used for all three cases. The results are summarized below.
Raw values Secondary displacement 1.711 microns Secondary rotation 0.0 Primary rotation -0.0397 microrads Primary displacement -0.0113 microns Relative to primary optical axis Image motion Secondary displacement 1.481 microns 0.041 arc sec Secondary rotation -0.0397 microrads -0.004 arc sec Primary rotation 0.0397 microrads 0.008 arc sec total 0.045 arc sec
The static displacement of the secondary mirror with respect to the optical axis is within the range of motion which can be optically compensated by retilting the secondary.
We have not yet done a modal analysis (calculation of the lowest natural frequencies of a structure), though the static analysis indicates that all modes should be well above 10 Hz. Hand calculations have shown that the truss elements and vanes have natural frequencies above 10 Hz.
No temperature loading has been applied to the structure as of this writing. These calculations will indicate the sensitivity of focus and collimation to temperature gradients in the OSS. The effect of temperature gradients on pointing will be explored also.
Specifications of radial runout, smoothness and friction, repeatability of errors, axial loading, etc are only available for off-the-shelf versions. Substantial improvements in all specifications can be achieved at moderate cost by selection and remanufacture.
For an altitude-azimuth telescope the support geometry need only handle pitch in a vertical plane. A very elegant concept which satisfies this requirement has been suggested by Roger Angel (see Figure 2.9) and is currently under development by Larry Barr and colleagues at Kitt Peak for use with a 1.8 meter honeycomb mirror to be tested in the MMT late next year.
The mechanism consists of links forming a pantograph type geometry. The position of one parallelogram pivot is fixed with respect to the mirror cell and serves as a fulcrum for an effective lever whose loads are the counterweight at one end and the mirror at the other. The pantograph geometry preserves a constant lever ratio despite changes in the shape of the parallelogram. In this way the mechanisms may accommodate themselves individually to each mirror load point for all angles in an altitude quadrant between 0 and 90 degrees while applying at all times a constant reaction to gravity.
The airbag alternative support is mechanically less complex and somewhat less massive than mechanical lever supports but requires an actively regulated air supply. Since it inherently provides only axial support some supplementary in-plane support would be required.
The number and positions of such 'astatic' supports necessary to preserve the optical mirror figure at all operating orientations is determined by finite element structural analysis and depends on mirror geometry, thickness, and optical performance constraints. Such an effort will be needed for the MILT 3.5 meter honeycomb mirror and will be commissioned in early 1984.
The mirror is a Pyrex honeycomb structure 80 cm in diameter attached at the three corners of the triangle labeled A, B, and C. The mirror and the structure represented by the triangle then move as a rigid unit.
Two of the three links lying in the plane normal to the optical axis just behind the mirror provide roughly orthogonal mirror offsets by virtue of their length control actuators. In Figure 3.7 the triangle is shown offset as a result of changing a single link length. Notice that even over this exaggerated range of motion, the link end and the triangle (ie. mirror) center follow roughly parallel paths of roughly equal lengths.
Length control of one additional link provides roughly orthogonal freedom of positioning. The third link has a fixed length.
The three remaining links lying parallel to the optical axis provide piston and tilt. Piston is converted by the linkage geometry into harmless secondary revolution about its optical axis, allowing the offset links to freely move out of a plane without any tension or compression buildup.
Obviously all motions interact and one would hesitate to construct such a mechanism were it to be adjusted by hand. Modern microprocessor control technology happily allows full exploitation of the mechanical simplicity of this arrangement. Its control firmware will synthesize the combined link motions necessary to allow effective orthogonality of control commands from ahost computer.
Moreover since the secondary and its mechanism will at all times be shielded from wind, the dynamic performance requirements are comfortably minimal.
Linear offset is given by
where f1 is the primary focal length, f is the telescope focal length, and is the image angular offset.
For the 3.5 meter f/1.75 - f/10 telescope secondary mirror tilt is given by
where D2 is the secondary diameter and D1 is the primary diameter. For the same telescope this value is
The 6 mm range will help simplify initial focus at telescope assembly time, allow sufficient range to actively compensate differential sag of primary and secondary supports, and compensate thermal extension and contraction of the telescope OSS (partially compensated by changes in the optics with temperature). The piston range required for tilting is negligible by comparison.
A radial offset range of +- 3 mm min will also be provided, again to facilitate initial telescope collimation and relax some manufacturing and assembly tolerances.
The offset range necessary to compensate gravity deflection is only about 0.25 mm.
This mechanism is highly repeatable, has few and simple moving parts, uses all standard industrial components (actuator, solenoid, etc.), and dissipates no heat. More glamorous servo technology was considered but was rejected because it costs more, and servo deadband jitter adds to the image size.
Before removing the primary mirror it will be simple to replace the baffle top segment with a stiff plate having a lifting eye at its center. The entire tertiary structure including structural baffle, rotation mechanism and mirror can lifted away vertically with complete safety.
A rolled or segmented shell with a conical outline connects this bearing with the large diameter friction drive disk about 4 meters above. The rim of the drive disk is ground cylindrical and constrained by guide rollers around its periphery, at least one of which transfers azimuth drive motion from a DC servo motor via a friction reduction drive train.
The cone and drive disks were modeled using quadrilateral shell elements which have both bending and membrane capabilities while the platform and forks were modeled by quadrilateral membrane shell elements which have no bending stiffness. The performance of the forks and platform should be dominated completely by the in plane properties of the elements so membrane elements should be appropriate.
The displacements of the nodes at the base of the cone were constrained vertically and radially. The two nodes on the outer diameter of the drive disk which were on the x-z and y-z symmetry planes were constrained radially. In addition one of these nodes was constrained tangentially. These constraints will be replaced by springs as the stiffness of bearings and guide rollers are determined.
All plates were 12.7 mm (0.5 inches) thick steel except the azimuth cone where the plates were 25.4 mm (1.0 inch) thick. Only a minor attempt was made to optimize thickness of members. No attempt was made to optimize the geometry of the structure. The mass of the 1/4 fork was 4762 kg and its moment of inertia about the vertical axis was 18100 kg m^2.
2) With the telescope pointing at the zenith, wind blowing the x direction causes x loading at the tops of the fork tines as well as anti-symmetric z loading across the yz plane on the fork tines. The effect of both the x loading and the z loading is to to twist the fork about the y axis. Distortion due to this load case should not be as serious as for case 1 since the enclosure should be effective in shielding the telescope from much of the x component of the wind. Figure 3.10 shows the deformation of the fork due to x loading.
The results of the finite element calculation are given below. The applied forces refer to full model forces rather than those applied to the 1/4 model.
Case 1: Force on fork tines Applied y force 200 N Resultant fork y displacement 0.77 microns Case 2: Force on fork tines Applied x force 200 N Resultant fork z displacement 0.65 microns Applied y torque 514 N m Resultant fork z displacement 0.27 microns
From these results it is straightforward to calculate pointing errors due to wind loading.
Air density (2800 meters altitude) 0.90 kg/cubic meter Wind loading (5 m/s with drag coef = 1) 11 N/square meter Fork tine separation 3.8 meters Drive disk radius of curvature 1.8 meters
We consider wind loading on the upper end of the OSS including the truss and secondary baffling.
Effective area of upper end of OSS 3 square meters Moment arm 5 meters Wind force on upper end of OSS 33 N Wind moment 170 N m
For case 1 the force on the fork ends is 38 N which results in a y displacement of 0.15 microns. Due to the encoder location on the drive disk and its reference to the apex of the azimuth code the control system prevents telescope motion other than rotation about the apex of the azimuth cone (see Figure 4.5 for the geometry). This results in an angular motion of 0.004 arc sec.
For case 2 the x force on the fork arms is 33 N giving a z displacement of 0.11 microns. The moment applied to the tops of the fork arms is 170 N m giving a z displacement of 0.09 microns. Superposition of the two deformations results in a pointing error of 0.008 arc sec.
The fork errors will be correlated to a large extent with the errors due to the deformation of the OSS and the error in the control system due to wind loading and will add algebraically rather than in quadrature.
In both cases the largest stresses were in the short beam section of the platform between the top of the azimuth cone and the base of the fork tine suggesting that the performance of the structure could be substantially improved by adding some material in this region. Also the webbing thickness in the platform and the fork is probably much thicker than it needs to be and it is likely that the webbing weight could be reduced.
Random errors in radial definition arise typically from nonuniformities in roller dimensions and poor dynamic roller alignment. These errors can be minimized by factory selection of stock bearing with these performance requirements in mind and by selecting a brand with the best roller retainer design.
The extent to which radial runout under the telescope load is repeatable in amplitude and phase will be studied with help from factory engineers before acommitment to this azimuth cone design is made.
(Note that two types of encoders are referred to in this paper; an absolute encoder generates a signal indicating the position angle of its shaft. An incremental encoder indicates only changes in position angle of its shaft but not the absolute position.)
Very stiff and close fitting seats for the bearing race rings are necessary to prevent bearing distortion under load. No serious design difficulty is apparent here since there are essentially no space limitations that would prevent using components of the required dimensions.
Since Nasmyth instrument loads will be balanced at all times there will be no azimuth overturning loads except for some small residuals and those caused by wind loading on the exposed parts of the telescope.
Wind loads resolve into worst case radial loads of less than 1000 N. A balanced radial preload of 10000 N will be applied to the guide rollers. This will increase stiffness since one also serves as drive roller and will provide sufficient tangential friction to resist slipping due to wind load induced torques. The method of preloading the guide and drive rollers is TBD.
Figure 3.12 shows a proposed drive roller mechanism. The driven roller is cylindrical and contacts the drive tire on a line parallel to the rotation axis of both. To optimize drive stiffness its design enforces the condition that the axis of the tangent column pass through the drive contact point and through the flex-hinge at the base point. The mechanical design of this mechanism will undergo a thorough stiffness analysis and optimization.
Guide rollers will contact the ground edges on the drive tire. This arrangement simplifies the alignment of the drive roller with respect to the drive tire, and also localizes stick-slip effects associated with residual misalignment of the drive disk and drive roller axes.
Drive contact preload is applied precisely toward the drive disk axis to guarantee zero moments about the contact line. The preload also establishes uniform pressure along the contact line automatically establishing parallelism of the roller and disk axes. A sufficient degree of torsional compliance in the flex hinge allows this freedom.
Given that the entire mechanism can be manufactured to precision tolerances it can easily be adjusted in place before final locking down of the flex hinge fixed point. This will leave the hinge essentially free of strain other than pure compression and tension along a tangent to the contact point.
The azimuth disk is a continuous 360 degree cylinder and along with a number of defining rollers serves as the upper defining bearing for azimuth rotation.
In each case there is a 'track' which is never contacted by the friction drive roller. This clean track is utilized exclusively for friction roller incremental encoding.
The disk must be radially stiff to minimize 'flower' distortion due to guide roller preload. It must also be torsionally stiff to preserve high torsion mode frequency. In the direction through the disk sufficient stiffness is required to preserve high fundamental mode frequencies and control static sag under gravity. The disk contributes significantly to the azimuth moment of inertia so it deserves to be structurally optimized.
At its center is a reference bore and face concentric with the drive tire to which the absolute encoder driveshaft hub attaches. Loads due to the telescope OSS, fork and instruments are transferred through the disk and distributed around the top rim of the conical shell structure below.
The effects of stresses stored as a result of welding is TBD. Some advice from shop experts here is needed.
The effects of a 0.005 inch range in different diameters across the disk amounts to a maximum absolute pointing error of about +- 20 arc sec, most of which repeats in phase with absolute encoder angle and sums to zero after one complete rotation. It is expected therefore that any 'machined in' effects will be largely computer-correctable.
In the presence of thermal gradients across the disk, scale errors will exist whose amplitude and azimuth dependence will be difficult to impossible to predict and correct. For a steel disk (CTE=12 ppm/deg C) a 1 arc sec pointing error over a 180 degree interval would be contributed by a uniform thermal gradient across the diameter of about 0.15 deg C max.
Since the absolute encoder is small, and thermally compensated, azimuth pointing need not be limited by thermal gradients in the drive disk any more than by thermal gradients elsewhere in the mount. Therefore unusual measures to control disk gradients seem unnecessary. Moreover, when offset pointing is appropriate, the allowable thermal gradient budget is increased far beyond anything likely to exist under the worst possible conditions.
Because of its large size the large disk will be flame hardened. We will exert considerable effort to ensure that the fabricator achieves a minimum specified uniformity. An experienced consultant will be employed for this job.
In case of an accidental skid at the drive contact point it may be possible to confine potential surface damage to the drive roller by making it intentionally softer than the main disk. Damaged drive rollers are easier to replace.
Cleanliness of the disk/roller assembly is essential; a suitable housing and dust seal system will be provided.
The drive disk/roller design task is essentially identical to the design of any rolling antifriction bearing. This is an exceedingly well developed technology and we will thus not need to take any risks. Relatively small loads and large dimensions will mean small stresses, much less than in a typical ball bearing at its rated load. An essentially infinite lifetime design should not be a problem.
Surface failure can also occur slowly due to corrosion. This calls for either the use of corrosion resistant steel or surface protection by active means such as by continuously wiping an oil film onto the surface.
Surface denting or 'Brinelling' can occur if the bearing surfaces are overstressed. Shock loading is a common cause for this type of failure. This type of damage is most likely to occur during assembly, disassembly and shipping.
Accidental surface wounds may be inflicted by encounters with sharp tools, etc. Since the drive roller comes into contact along a line many centimeters in length, some insensitivity of performance to this type is damage is available. Appropriate precaution and oversight will occur during manufacture, packing, shipping, and installation, however.
The situation with the azimuth axis is much simpler. Since the enclosure and the telescope are co-rotating clearances between the telescope and mechanical stops affixed to the enclosure can be made small. Relative angular velocities will be low; a stop will be encountered almost immediately in the event of a failure. The azimuth drive train will be backdrivable by the building without damage. Limit switches will cut power to the drive motors when a fault occurs. Mechanical stops will be shock absorbers. This system is implemented on the MMT.
Coolant lines for detectors are also contemplated. It is expected that detector cooling will be self contained wherever possible. A very compact Joule-Thompson type cooler using nitrogen as the expanding gas has recently been introduced by the MMR Technologies corporation and is being used to cool CCD's at Lick Observatory. The nitrogen gas circuit mentioned above anticipates routine use of these devices.
Where self contained cooling is not practical, cooling equipment will be supported by the enclosure floor near the instrument and rejected heat ducted away from the telescope (Ulich et al. 1983).
Instrument rotation resolution will be in ~1 micron increments measured at 20 cm from the center of rotation. This will accommodate the most extreme possible case requiring 0.1 pixel guiding precision for a CCD being used at the extreme field margin. This is equivalent to rotator angular resolution of 1 arc sec. This can be conveniently accomplished using a stepper driven worm gear. The exact angle per step is thus determined and since only rate information is required for de-rotation no additional absolute encoders are needed save for a single fiducial position marker arranged for detection by the microprocessor controller.
The azimuth cable windup will be somewhat more complicated than that of the MMT. The bearing diameter at the apex of the azimuth cone will be too small for the cabling. It must be a self aligning thrust bearing and a reference for the absolute azimuth encoder will take much of the space available. Still it is evident that there is room both vertically and horizontally for cable windup exterior to the azimuth cone and reasonably close to the rotation axis. Figure 3.15 shows a conceptual design for the cable wrapup. This appears to be a workable concept; we present it as an example of a solution to the problem, not as the final design.
The altitude windup is a simple problem in comparison. Not only is the range of motion much smaller (90 degrees rather than 540) but the amount of cabling is much less. A simple drape between the fork and the OSS will suffice.
This means there will be a standing design goal to weld and anneal all structural connections wherever possible. This may include some joints which are welded only after final installation at the site. On site welds will need to be kept small and confined to cases where annealing is not crucial. Joints where bolts must be used will be designed and fabricated to have low hysteresis.
(Hysteresis in the MMT structure is now well controlled and known to contribute no more than 0.3 arc sec to overall pointing error.)