Skip to main content

Sub-Millimeter Positioning in Segmented-Mirror Handling Systems

Installing a segment in a segmented primary mirror is a positioning problem with three stages: a crane that moves the segment to within millimeters, a handling fixture and positioning system that close the last millimeter, and the telescope's own sensors and actuators that take it from there to nanometers. UTEC Industrial designs, engineers, machines, fabricates, and installs custom material handling systems for aerospace and heavy industry from its Spokane Valley, WA facility, integrating Allen-Bradley PLC and motion control with in-house CNC machining, heat treating, and stress relief. This article explains how published observatory requirements and lessons divide that work, what they ask of crane creep speed, hoist drives, load transfer, fixture interfaces, sensing, and verification, and why sub-millimeter placement belongs to the handling system rather than the crane hook. The equipment follows the same chain as any heavy handling system, design → engineering → parts machining → fabrication → assembly → weld fatigue → stress relief → drives → controls → tuning → monitoring, and the positioning budget is set at the design link and spent at the tuning link.

Why can't the crane alone place a mirror segment to sub-millimeter accuracy?​

Because published crane requirements, such as TMT's, are measured in millimeters and a segment's capture window is about one millimeter. The Thirty Meter Telescope (TMT) Observatory Architecture Document, a 2018 design-requirements document for a telescope not yet built, gives a published example of how an observatory specifies precision at the hook. For its 20 t enclosure crane, which the document says will principally service instruments on the Nasmyth platforms, it specifies:

TMT enclosure-crane requirementSpecified value
Minimum hook speed0.3 m/min or less
Vertical acceleration in a limited mode0.05 g or less
Vertical hook position increment2 mm or less
Horizontal hook position resolution25 mm diameter or less

Compare that with what a segment's own sensing can capture. In a 2014 paper on the European Southern Observatory's (ESO's) M1 Test Facility, the edge sensors that measure a segment's piston against its neighbors had a coarse catching range of ±1 mm, with nanometer-level precision over a ±200 µm range, while gap and shear were monitored over a range of several millimeters. The two documents describe different telescopes and different lifting equipment, so the comparison is one of scale rather than a single budget: a 2 mm vertical increment equals the full width of a ±1 mm piston window, and a 25 mm horizontal resolution circle is wider than the several-millimeter range over which gap and shear are monitored.

The engineering conclusion is the framing for the rest of this article. The crane does the coarse move at creep speed; the handling fixture, the positioning system, and any lifting jack do the sub-millimeter work; and the telescope's active control does the nanometer work described in Observatory and Scientific-Instrument Handling Equipment. A specification that asks the crane hook for sub-millimeter placement asks for the wrong thing from the wrong machine (TMT International Observatory 2018, §4.6.1 [REQ-1-OAD-6210]; Dimmler et al. 2014, §3).

How is positioning split between the crane, the lifting fixture, and the positioning system?​

TMT's architecture document defines the M1 segment handling system (SHS) as an integrated system of three parts: a segment lifting fixture, a positioning system, and a crane or other means to raise and lower segments. Its requirements show how the work is divided:

  • The crane raises the segment between the observing floor and the mirror cell. A separate crane of at least 0.5 t handles segment assemblies, against the document's mass estimates of 210 kg for the segment assembly and 150 kg for the lifting talon.
  • The SHS has six motorized degrees of freedom (three translations and three rotations); as engineering reasoning, that lets it align the segment independent of the hook.
  • The telescope is locked in a zenith-pointing orientation for installation and removal.
  • The transfer to the handling cart happens only after the segment is leveled, and the document's discussion says the number of transfers should be minimized.
  • The mirror must not be contacted under any condition, including loss of power.
  • Contamination, such as dust and oil, is strictly minimized.

Read as a positioning budget, the crane's job ends when the segment hangs close to its slot at creep speed, and the six-axis positioning system then brings it within the capture window of the segment's own sensing. As engineering reasoning, the no-contact-on-power-loss requirement means every motorized axis in that stage has to hold position when power is removed, which makes brake and drive behavior part of the positioning design rather than a safety afterthought (TMT International Observatory 2018, §4.2.3.8 [REQ-1-OAD-1610 to 1642] and §4.6.1 [REQ-1-OAD-6230]).

What duty cycle does a segment handling system have to be designed for?​

Segment handling recurs throughout a telescope's life, since mirror coatings are renewed on a schedule (the Keck example below). TMT's architecture document sets the SHS duty as:

  • any 10 segments exchanged in a 10-hour day;
  • 2,000 operations during construction;
  • 10 exchanges per 8-hour day every two weeks for 50 years, which the document totals at 13,000 exchanges; and
  • a proof test at 2 times rated load every six months for 50 years.

The proof-test requirement alone amounts to about 100 proof tests over a 50-year life, at two per year; as engineering reasoning, that is far more load testing than equipment tested only at acceptance.

W. M. Keck Observatory's own account of its program gives an operating example at a smaller scale. When everything goes smoothly, its crew removes and replaces three segments in a day; the aluminum coating has to be stripped and reapplied about every two years; and with 36 segments in each primary, twelve days of segment exchange give one mirror a complete recoat. The ESO figure for its Extremely Large Telescope and the resulting yearly operation count are worked through in the observatory companion article.

As engineering reasoning, a failure mode is equipment specified for installation duty only: a fixture and positioning system sized for one careful commissioning campaign, then used for thousands of exchanges, wears its screws, brakes, and locating features until it can no longer deliver the segment inside the capture window (TMT International Observatory 2018, §4.2.3.8; Cooper 2011, W. M. Keck Observatory).

Why does the crane's slowest speed matter more than its fastest?​

A precision lift's final approach is made at creep speed, and the handling hazard is overshoot, not delay. TMT's requirements for its 20 t enclosure crane, repeated for its 10 t jib crane, show how creep performance is specified in separate terms:

  • Minimum hook speed of 0.3 m/min or less. That is 5 mm/s. At 5 mm/s, a 2 mm move takes 0.4 s, so reaction time of the operator or the control loop directly sets how far the hook travels past its target.
  • A minimum vertical increment of 2 mm or less. This is a separate requirement from minimum speed; as engineering reasoning, it reads as a controlled step, an inching function, rather than continuous creep.
  • A mode that limits vertical acceleration to 0.05 g or less. That is about 0.49 m/s², a ceiling on the jerk and bounce a suspended segment sees when motion starts and stops.
  • A segment-maintenance hoist continuously variable from 0 to 4 m/min, with its trolley from below 0.2 m/min. The same document applies creep requirements to the monorail crane in the M1 segment maintenance area.

Hook creep speed is set by the hoist drive, its gearing, and its reeving together, so the requirement belongs in the crane specification as a hook-level value, not a motor-speed value. As engineering reasoning, a failure mode is a crane bought on capacity and top speed, whose drive can hold 0.3 m/min only unsteadily with a light segment on the hook, so the operator inches by jogging and the load bounces at every stop (TMT International Observatory 2018, §4.6.1 [REQ-1-OAD-6210, 6213 and 6271]).

How do hoist drives hold a load at zero speed and inch it into position?​

Holding a suspended optic motionless, then moving it a few millimeters, is a drive function as much as a brake function. Rockwell Automation's reference manual for PowerFlex 750-series drives with TotalFORCE control describes the lifting and torque-proving (TorqProve) functions of the PowerFlex 755T drive:

  • Torque proving before brake release. Before releasing the mechanical brake, the drive checks motor output phase continuity and verifies proper motor control.
  • Brake proving before the drive lets go. The drive proves the brake holds before it drops its own control of the load, and it monitors for motion after the brake sets.
  • With an encoder: brake-slip detection, float, and micro positioning. Brake-slip detection lowers the load slowly if the brake slips; float holds full torque at zero speed, so the load hangs on the motor rather than the brake; and micro positioning is listed alongside them; as engineering reasoning, it suits the slow, small moves of the final approach.
  • Encoderless mode. Micro positioning is available, but float and brake-slip detection are not, and the manual limits encoderless torque proving to applications where personal safety is not a concern.

As engineering reasoning, for segment handling the encoder-based mode is the relevant one: float and micro positioning together let a hoist hold a segment still above its slot and then inch it down. NASA-STD-8719.9C, NASA's lifting standard, sets a related rule for cranes used for NASA Critical lifts: either two holding brakes, each capable of bringing a rated load to zero speed and holding it, or one holding brake with a motor drive that automatically monitors brake functionality and motor torque. As engineering reasoning, a failure mode is a hoist that holds only on its brake between jogs, so each restart is a small drop as the drive takes the load back (Rockwell Automation 750-RM100D-EN-P, Ch. 5, pp. 151–152 and p. 160; NASA-STD-8719.9C, 2024, §5.2.2.1–5.2.2.2).

What happens to the load when a jack or support takes it from the crane?​

The hand-off is a point of risk for a segment's supports, and ESO's M1 Test Facility paper records one reason. The facility, a section of the primary mirror with four prototype segment subunits, served to study and get familiar with calibration, alignment, and handling procedures long before the mirror's components are commissioned, and its lessons are published:

  • Test load transfer at every inclination. Load transfer when the extraction jack is lowered has to be tested at different inclinations, because tight tolerances, partly driven by earthquake safety, could create extraction forces large enough to damage the segment support.
  • The crane hand-off was not prototyped in detail. The facility had no segment crane; a fork lift was used to transport segments, so load transfer to the segment crane was not prototyped in detail there.

As engineering reasoning, the first lesson means the load path during transfer is not simply the segment's weight moving from one device to another. Friction and binding in tight locating features can put forces into the segment support that neither device is designed to carry, and those forces change with the angle at which the transfer happens.

The second lesson is a caution for any program. A handling rehearsal that substitutes a different lifting device for the real crane leaves the crane-to-support hand-off, with its creep speed, drive behavior, and hook compliance, untested until the real segment is in the air. As engineering reasoning, a failure mode is an extraction or insertion that binds partway, where continued hoisting builds force in the support instead of moving the segment (Dimmler et al. 2014, Abstract and §2).

Why does re-seating a segment on a different support frame cause clocking errors?​

Segments are removed, coated, and reinstalled, and they do not always go back onto the same hardware. ESO's M1 Test Facility found that the repeatability of segment-assembly integration needs to be tested with different segment assemblies on one fixed frame: its tests showed problems with clocking, the segment's rotation about its own axis, which became apparent only once it was considered that a replaced segment will not be mounted on the same fixed frame again. The same paper reports a measured membrane stiffness more than twice the nominal value, a reminder that the as-built behavior of the support can differ from its design model.

For the handling system, clocking is a positioning requirement in its own right:

  • Clocking is set by locating features. Pins, pads, and reference faces on the fixture and on every frame the segment visits have to define the same rotation, so a segment moved between frames lands at the same clocking each time.
  • Interchangeable frames need interchangeable tolerances. If several fixed frames or storage stands receive the same segment, their locating features are machined and inspected to one drawing and one datum scheme, not fitted individually.
  • As-built checks matter. Stiffness and position are measured on the built hardware, because the test facility measured a membrane stiffness more than twice its nominal design value.

UTEC Industrial machines fixture interfaces to tolerances as tight as ±0.001 in and inspects them by CMM. As engineering reasoning, a failure mode is a segment that seats correctly on the frame it was aligned on and out of clocking on every other, so each exchange starts with a realignment the handling system was supposed to prevent (Dimmler et al. 2014, §2).

How does a lifting fixture spread load into a fragile optic?​

A mirror cannot be lifted like a steel part, and the fixture's design load can differ from the mirror's weight. The University of Arizona team that casts 8.4 m segments for the Giant Magellan Telescope (GMT) describes the lift after casting:

  • The mirror is lifted by a steel frame bonded to its front surface with compliant adhesive on 36 disks.
  • The frame then turns the mirror into a vertical plane.
  • Most of the silicon carbide floor tiles and all of the ceramic fiber boxes are still attached at that step, nearly doubling the weight supported.

Two design lessons follow as engineering reasoning. First, the load is spread through many compliant pads rather than a few hard points, so no single pad overstresses the glass and small errors in pad height are absorbed by the adhesive. Second, the design load is everything attached at the moment of the lift, which may be close to twice the finished optic.

Keck's segment exchange uses a different pickup, a fixture placed around the segment's edges, as described in the observatory companion article; the principle of engaging defined, load-rated features rather than the optical face is the same. Whichever pickup is used, the fixture is a below-the-hook lifting device: ASME BTH-1-2023 provides minimum design criteria for such devices, and ASME B30.20-2025 covers their marking, inspection, and testing among other provisions, both cited here at the level of the standard and discussed further in What Is Mechanical Ground Support Equipment (MGSE) for Aerospace Handling? (Martin et al. 2020, §4 and Fig. 8; W. M. Keck Observatory 2010; ASME BTH-1-2023; ASME B30.20-2025).

How do load cells confirm that the load has transferred, and what limits their readings?​

A load cell under the hook or in the fixture is how the controls know which device is carrying the segment. As the positioning system lowers a segment onto its support, the hook load should fall toward the weight of the fixture alone; a residual reading above that means the fixture is still carrying part of the segment, the kind of binding the ESO test facility's extraction-force lesson points to.

The vocabulary for specifying such a load cell can be taken from OIML R 60-1:2021, the international recommendation for load cells. It is written for load cells in measuring instruments subject to metrological controls, so it is cited here for its terms, not as a requirement on crane load monitoring:

  • Accuracy class. Load cells are ranked by overall performance capability into one of four accuracy classes, A to D.
  • Creep. Creep is the change in a load cell's output with time under constant load, with other conditions held constant, which matters when a segment hangs on the hook for minutes during alignment.
  • Influence quantities. Temperature and other influence quantities affect the relation between the indication and the result, and a dome at night is not a calibration laboratory.

NASA-STD-8719.9C defines a load measuring device as a below-the-hook device used to indicate the weight of the item being lifted, such as a load cell; as engineering reasoning, that makes the cell part of the lift record as well as the control logic. A related failure mode is a transfer confirmed by a load reading taken from a cell whose creep and temperature drift were never stated, so a residual force in the support goes unnoticed (OIML R 60-1:2021, §2, §3, and §5; Dimmler et al. 2014, §2; NASA-STD-8719.9C, 2024, §3.2).

What brake, limit, and fail-safe rules apply when the optic is a critical lift?​

A one-of-a-kind optic is a high-value lift, and NASA's lifting standard gives a published model for how such lifts are controlled. NASA-STD-8719.9C classifies an operation as a NASA Critical lift when failure or loss of control presents an elevated risk of serious injury, loss of life, or loss of one-of-a-kind articles or high-dollar items whose loss would have serious programmatic or institutional impact. Its classification rules and its equipment rules for those lifts, covering the equipment hazard analysis, holding brakes, emergency stops, upper and lower limits, and proof loads, are set out in Handling Airframe, Engine, and Spacecraft Assemblies. They are NASA practice rather than a universal legal requirement, and two of them bear directly on segment positioning:

  • Fail-safe control. The standard recommends control design in which a single failure cannot drive the crane faster than commanded or in a direction other than commanded. As engineering reasoning, over a mirror that behavior can decide whether a fault ends in a stop or a strike.
  • Proof load. Cranes are proof-tested with a mock load of 1.20 to 1.25 times rated capacity, and load positioning devices by holding a mock load in the same range or as recommended by the designer with concurrence from the Center's Lifting Devices and Equipment Manager (LDEM).

TMT's architecture document adds a proof test at 2 times rated load every six months and the no-contact-on-power-loss requirement, which as engineering reasoning makes fail-safe behavior a positioning requirement. A failure mode, again engineering reasoning, is a single-fault overspeed: a failed speed reference or contactor that drives the hook at full speed toward the mirror while the segment is centimeters above it (NASA-STD-8719.9C, 2024, §4.2.2, §5.2.3.9, §5.3.1.1, and §13.3.2.6; TMT International Observatory 2018, §4.2.3.8).

What sensing, controls, and interlocks run the fine-positioning stage?​

The fine-positioning stage is where the intelligence layer does the work the crane cannot. One architecture combines:

  • Shared permissives. The telescope is locked at the handling attitude, zenith in TMT's case, and the handling controller reads that state before any motion near the mirror.
  • Six-axis servo positioning. Each of the positioning system's degrees of freedom runs on a servo axis with encoder feedback. Allen-Bradley Kinetix 5700 inverters support DSL and Hiperface encoder feedback, with position, velocity, and torque loop axis configurations and a current regulator loop, and feature a safe torque-off function.
  • Level and attitude sensing. An attitude check confirms the segment is level before the transfer to the cart, turning TMT's level-before-transfer requirement into a permissive.
  • Load sensing. Load cells confirm pickup weight and the completion of each transfer, as described above.
  • Capture-window logic. The final approach ends when the segment is inside the capture range of its own sensing, about ±1 mm in piston for the edge sensors in ESO's test facility, and the handling controls hand off to the telescope's control system from there.
  • Deterministic logic. Logix 5000 controller tasks can be configured as continuous, periodic, or event, and a periodic task executes at a preconfigured interval, so permissives placed in one are evaluated on a known schedule.
  • Safety basis. ISO 12100:2010 covers the risk assessment, ISO 13849-1:2023 covers the design of safety-related control parts, and IEC 60204-1:2016 covers the electrical equipment, each cited at the level of the standard.

UTEC Industrial, a Rockwell Automation Recognized System Integrator, builds Allen-Bradley ControlLogix and CompactLogix control, servo drives, and UL 508A panels into the handling equipment it fabricates (TMT International Observatory 2018, §4.2.3.8; Rockwell Automation 2198-UM002E-EN-P (2018), Kinetix 5700; Rockwell Automation 1756-RM094N-EN-P-2025; Dimmler et al. 2014, §3; ISO 12100:2010; ISO 13849-1:2023; IEC 60204-1:2016).

How are fine-positioning axes rehearsed, tuned, and verified before the first real segment?​

The ESO M1 Test Facility is a published example: four prototype segment subunits on a representative section of the primary mirror, used to study and get familiar with calibration, alignment, and handling procedures long before the mirror's components are commissioned. A handling program can apply the same idea at the equipment level:

  • Dummy-mass rehearsal. Each transfer is run with a mass dummy of the real segment assembly, including the fixture and any hardware attached, so drive tuning, load-cell thresholds, and the hand-off sequence are proven before glass is lifted.
  • Tuning at load. The Kinetix 5700 commissioning procedure includes an axis tuning step, and the manual notes that tuned bandwidths can need adjustment once the motor and load are connected. As engineering reasoning, axes tuned on an empty fixture can overshoot with a segment on them.
  • Rehearse the real hand-off. The crane-to-support transfer is rehearsed with the real crane, because that is the step the ESO facility did not prototype in detail.
  • Measure the result independently. The positions the handling system delivers are checked with an instrument whose own performance is known. A laser tracker used for that check is evaluated under ASME B89.4.19-2006 (R2015), whose ranging, length and two-face tests are explained in the MGSE companion article.

As engineering reasoning, a failure mode is commissioning by the first real exchange: the first time the full sequence runs at full mass is with a segment on the hook (Dimmler et al. 2014, Abstract and §2; Rockwell Automation 2198-UM002E-EN-P (2018), Kinetix 5700; ASME B89.4.19-2006 (R2015); Muralikrishnan et al. 2009).

Where does a segment handling system sit in the design-to-monitoring chain?​

Each link of the chain has a positioning consequence:

  • Design and engineering split the positioning budget among crane, positioning system, and telescope sensing, and define the design load with everything attached.
  • Parts machining produces the locating features that set clocking and repeatability across every frame and stand.
  • Fabrication and weld fatigue decide whether fixture and cart frames survive a duty measured in thousands of exchanges and a proof test every six months.
  • Stress relief keeps machined interfaces where they were machined, as explained in Stress Relief for Machine Bases and Frames Before Final Machining.
  • Drives, controls, and tuning deliver creep speed, float at zero speed, micro positioning, and interlocked hand-offs.
  • Monitoring trends brake-slip events, load-cell readings at pickup and transfer, and proof-test results across the equipment's life.

UTEC Industrial has built precision handling fixtures for segmented-mirror telescope components for W. M. Keck Observatory. The monitoring link closes the loop on the duty cycle: a drive that detects and reports brake slip, as the PowerFlex reference manual describes for encoder-based operation, gives the maintenance program an early signal before a brake problem becomes a lowered segment (TMT International Observatory 2018, §4.2.3.8; Rockwell Automation 750-RM100D-EN-P, Ch. 5).

What should a segmented-mirror handling specification define?​

A specification that asks for "sub-millimeter crane positioning" puts the requirement in the wrong place. A complete specification separates the stages:

  • Coarse stage, crane: minimum hook speed, vertical increment, horizontal resolution, and an acceleration-limited mode, written at the hook, plus the brake and drive functions for holding at zero speed
  • Fine stage, positioning system: degrees of freedom, travel, resolution, and the capture window the segment must reach before hand-off
  • Hand-off: load-transfer sequence, inclinations at which it is tested, allowable residual force, and load-cell class and creep
  • Interfaces: locating features that fix clocking, and the datum scheme shared by every frame, cart, and stand
  • Duty: exchanges per day, per year, and over the equipment's life, and the proof-test interval
  • Safety and controls: permissives shared with the telescope, emergency stop, limits, fail-safe behavior, and the risk assessment they trace to
  • Verification: dummy-mass rehearsal, tuning at load, and independent position measurement

UTEC Industrial performs factory acceptance testing and on-site commissioning, so the hand-off sequence, interlocks, and dummy-mass tests can be demonstrated before a segment is lifted (TMT International Observatory 2018, §4.2.3.8 and §4.6.1; Dimmler et al. 2014, §2 and §3; ISO 12100:2010).

Related Articles

References​

  • TMT International Observatory. Observatory Architecture Document, TMT.SEN.DRD.05.002.CCR34. TMT Systems Engineering, 2018.
  • Dimmler M, Marrero J, Leveque S, Barriga P, Sedghi B, Kornweibel N (2014). "Improved E-ELT subsystem and component specifications, thanks to M1 test facility." Proceedings of SPIE, 9145, 91451K.
  • Cooper, A. Keeping Keck Telescopes Shiny. W. M. Keck Observatory, 2011.
  • W. M. Keck Observatory. A Mirror's Perfect Reflection. W. M. Keck Observatory, 2010.
  • Martin HM, Ceragioli R, Jannuzi BT, Kim DW, Kingsley JS, et al. (2020). "Manufacture of 8.4 m segments for the GMT primary mirror." Proceedings of SPIE, 11451, 114514F.
  • Rockwell Automation. PowerFlex 750-Series Products with TotalFORCE Control Reference Manual, 750-RM100D-EN-P. Rockwell Automation, 2025.
  • NASA-STD-8719.9C: Lifting Standard. National Aeronautics and Space Administration, 2024.
  • OIML R 60-1:2021: Metrological regulation for load cells — Part 1: Metrological and technical requirements. International Organization of Legal Metrology, 2021.
  • ASME BTH-1-2023: Design of Below-the-Hook Lifting Devices. ASME, 2023.
  • ASME B30.20-2025: Below-the-Hook Lifting Devices. ASME, 2025.
  • ASME B89.4.19-2006 (R2015): Performance Evaluation of Laser-Based Spherical Coordinate Measurement Systems. ASME, 2006.
  • Muralikrishnan, B. et al. (2009). "ASME B89.4.19 Performance Evaluation Tests and Geometric Misalignments in Laser Trackers." Journal of Research of the National Institute of Standards and Technology, 114(1), 21-35. DOI 10.6028/jres.114.003.
  • ISO 12100:2010: Safety of machinery — General principles for design — Risk assessment and risk reduction. ISO, 2010.
  • ISO 13849-1:2023: Safety of machinery — Safety-related parts of control systems — Part 1: General principles for design. International Organization for Standardization, 2023.
  • IEC 60204-1:2016 (Ed. 6.0): Safety of Machinery -- Electrical Equipment of Machines -- Part 1: General Requirements. International Electrotechnical Commission, 2016.
  • Rockwell Automation 1756-RM094N-EN-P-2025: Logix 5000 Controllers Design Considerations. Rockwell Automation, 2025.
  • Rockwell Automation 2198-UM002E-EN-P (2018): Kinetix 5700 Servo Drives User Manual. Rockwell Automation, 2018.

Ready to Discuss a Material Handling System?​

UTEC Industrial designs, engineers, machines, fabricates, and installs custom material handling systems for heavy industry, from the stress-relieved structure and drives to the Allen-Bradley PLC controls, tuning, and monitoring that run them, at its Spokane Valley, WA facility. Send UTEC the application, loads, and duty cycle to start a system review.

Request a Quote →

Questions? Call (509) 922-1832 or email sales@utec.co

Last updated on