Implications of Wind Testing Results on the GSMT Control
Systems
David R. Smith
MERLAB, P.C.
Hierarchical Approach
• If errors can be arranged hierarchically, then the control system can be as well.
• Large, high payload, long stroke systems can be slow and less precise.
• Higher bandwidth systems can be smaller stroke and capacity.
Hierarchical Approach (cont.)
• Keeping high-bandwidth control on smaller systems eliminates control-structure interactions.
• Intent is to keep cost/risk low by combining simpler and more standard control systems and components.
Errors• Large, slow errors (m-mm, <0.01-0.1 Hz)
– Gravity– Thermal– Mechanical misalignments– Wind
• Medium-sized, rate (<~10 m, <~10 Hz)– Wind– Vibrations
• Small, fast errors (<1 m, >~10 Hz)– Wind– Vibrations– Atmosphere
Controllers (example)
• Main Axis
• M1 Gross/Fine Position
• M1 Segment warping
• M2 Positioner
• M2 Fast tip/tilt/position
• M2 Deformation
• Downstream AO
Assumptions
• Most systems don’t interact• Separated physically and in bandwidth• Final image corrected by AO• Each previous system used to offload mean
positions.– E.g., M2 offloads AO to ~5 Hz– M1 fine offloads M2 to ~1 Hz– M1 gross offloads M1 fine to ~0.001 Hz
Assumptions (cont.)
• Separability of systems has limits– Motion of slow systems may induce vibrations– Some systems are partially redundant, so must
‘agree’ on how to remove certain errors (e.g., pointing)
• Some systems can’t avoid interaction– M2 fast positioner
Assumptions (cont.)
• Input must allow hierarchical approach
• Roll-off of errors must allow separation of high-bandwidth control from large structures.
• Wind is a key unknown– Magnitude of errors– Frequency content
Wind Data
• Gemini South 8m (Optical)– Structural (modal and operating)– Pressure on primary– Wind speed (on structure and dome)
• Nobeyama 45m (mm-Wave) – On-sky pointing– Structural (operating)– Controller
Gemini Data
• First round data (CD produced)– Modal Test– Operating Data– Wind pressures– DOE results
• Second round data (analysis beginning)– Wind speed and pressure only– Better coverage of parameter space
Nobeyama Data
• Goal was to investigate pointing– Pointing data analyzed– Structural data quick-look only
• Deformations relevant to GSMT– Similar size– Similar natural frequencies
Wind Effects
• Generally assumed to be low frequency– For 10m/s wind at 10m height
• Davenport Spectrum peaks at ~0.01 Hz• Antoniou spectrum peaks at ~0.1 Hz
• Roll-off is slow– Slope of -2/3 in typical approach to plotting
• Vortex generation from structure• All frequencies are affected
Wind Effects (cont.)
• All structural frequencies excited
• Amplitude drops as 1/²
• If a specific mode isn’t driven by a vortex, then deformations are unimportant above some frequency.
Nobeyama Results
• Deformation of the primary– Motion normal to surface– Rigid body tilt removed
• Motion of the secondary– X,Y,Z of typical point
Conditions of Tests
• Parked, calm (<2 m/s wind)– Benchmark case
• Tracking, calm– Effects of controller and motion
• Parked, windy (6-8 m/s)– Effects of wind
• No data tracking in wind
Deformations of the Primary
• Raw acceleration signal
• Removal of rigid body tilt
• Comparison of RMS deformation at/above a given frequency
Parked Telescope, Calm Wind
Tracking Telescope, Calm Wind
Parked Telescope, Wind 6-8m/s
RMS Comparison
Implications: Primary
• Total RMS error can be 10’s of microns
• Tracking is as important as wind– Hydrostatic bearings– Motion planning essential
• After ~3-4 Hz, residual is <1 m– Control of M1 would interact with structure– Low spatial frequency errors: M2 correction
Motion of the Secondary
• Accelerations in X, Y, Z
• RMS comparisons at/above a given frequency (X, Y, Z)
Parked Telescope, Calm Wind
Tracking Telescope, Calm Wind
Parked Telescope, Wind 6-8m/s
RMS Comparison, X
RMS Comparison, Y
RMS Comparison, Z
Implications: Secondary
• Twist motions much smaller• Tracking and wind cause same scale errors• Lateral and focus/tilt motions: 10’s of m• Most effects (>1m) below 3 Hz• M2 probably must correct ~3Hz effects
– Deformation– Position/tilt– Implies interaction with structure
Conclusions
• Data indicate likely size of errors
• Frequency range includes structural modes
• Seems to support hierarchical approach
• Interaction problem at M2
