Primary mirror edge sensor project for the Southern African Large Telescope
•
0 likes•41 views
A presentation given to SWLUG about the software development of the primary mirror alignment system for a large telescope.
Report
Share
More Related Content
Similar to Primary mirror edge sensor project for the Southern African Large Telescope
Similar to Primary mirror edge sensor project for the Southern African Large Telescope (20)
Recently uploaded
Recently uploaded (20)
Primary mirror edge sensor project for the Southern African Large Telescope
- 1. Primary Mirror Edge Sensor Project for the Southern African Large Telescope by Deon Bester
- 2. SALT Infographic Key features 1. 91 segment primary mirror 1. tracker with payload at the optical focus of the primary mirror 1. CCAS tower housing the Shack-Hartmann wavefront sensor for aligning the mirror 1. primary mirror surface is tilted at 37° from the zenith
- 3. Telescope at twilight during mirror alignment Dome shutter open and pointing at the CCAS tower, the Centre of Curvature Alignment Sensor. Telescope chamber has been conditioned to predicted outside temperature at twilight and louvres opened to facilitate natural air flow.
- 4. Primary Mirror Segmented Array ❖ 91 segments ❖ hexagonal shape ❖ 1m edge-to-edge ❖ 273 actuators ❖ 480 edge sensors ❖ size 10m x 11m ❖ 37° from the zenith
- 5. Primary Mirror Alignment System
- 6. Primary Mirror Alignment System High Level Flowchart Open Loop Alignment 1. Measure with Shack-Hartmann sensor 2. Process CCAS 3. MACS calculates GRoC 4. SPS positions mirror segments 5. repeat steps 1-4 until ... Global Radius of Curvature Achieved Set Sensor References: edge sensor heights & gaps zeroed Closed Loop Control 1. Acquire height & gap 2. Process data with MARS producing tip, tilt & piston values 3. MACS verifies dGRoC and sends displacements 4. SPS postions mirror segments 5. process repeated automatically to maintain mirror alignment until next referencing is required, (usually about every 5 days).
- 7. ★ Hexagonal mirror segments 1 to 91 ★ Central Segment = Segment No. 1 ★ Segments numbered in clockwise rings from the centre to the outer ring Segment Numbering System
- 8. ★ 6 Sensors per Segment (one per side) ★ Sensors numbered as 1 to 6 on segment, clockwise from top ★ Globally sensors numbered from 1 to 480 ★ Arithmetically >>> 91 x 6 = 546 sensors ★ No sensors mounted on the outer edge of outer ring where there are no adjoining segments. Sensor Numbering System
- 9. ★ 3 Sensor Segment Corner config with complimentary Transmitters and Receivers on adjacent segment edges ★ 3 hardware channels per rack module optimises cabling ★ Requires remapping of hardware channels to sensor numbers ★ The sensor number always refers to the Receiver device Sensor Configuration
- 11. Software Flowchart Description 1. Acquisition & identification of UDP datagrams. 2. Conversion of binary data to real values: Z, Y, Ro, hardware status. 3. Mapping of channel to sensor data. 4. Boundary limit checks: dead or disconnected sensors. 5. Sensor linearisation and application of temperature coefficients. 6. Adjust data averaging integration period according to temperature slope. 7. Average height/gap data to remove noise. Set sensor references for mirror alignment maintenance. 8. Sensor outlier detection using a matrix derived rogue sensor detector. 9. Sensor drift detection using a threshold hysteresis comparator. 10. Matrix transform using singular value decomposition and least squares method. The matrix transform calculates the required segment (actuator) displacements to restore the primary mirror to its original spherical surface/shape. 11. Calculate tip, tilt and piston for the mirror segments and transmit to the mirror alignment control system.. 12. Visualisation tool for analysing primary mirror GRoC errors (on the HMI). 13. Data logger for system events and nonlinear and linear sensor data. 14. Logging of ICD data to MySQL database (by the network Event Logging System).
- 12. Software Coding Schematic Illustrated here to show consistency with the Software Architecture diagram that follows.
- 14. Some features of the software architecture 1. Modular reusable software components 2. Subsystem communication using a publish & subscribe paradigm (DataSocket & http server) 3. Data exchange using type-defined clusters 4. Multithreaded application separated into: two data logging threads hardware communications thread sensor data acquisition & processing thread control system command thread system status broadcast thread state machine & execution engine thread 1. Queues for inter-process data transfer 2. Common data such as config info updated via ODMs (functional variables) 3. Global variables only used for system reset & software exit
- 15. Data Structure: Arrays Each parameter type for all sensors is contained in an array 480 elements long. The advantages are: 1. simple processing of parameters in a For Loop 2. known paradigm that is already in use in the ICD of the telescope control system. The drawbacks of the 480x1 array method are: 1. the data processing loops would have to run 480 times in memory 2. the 480 element arrays are not human reader friendly, 3. it is not easy to identify and correlate related data errors 4. they are difficult to navigate in terms of viewing complete mirror segment data, and 5. a lookup table is required for processing the sensor parameters by mirror segment number.
- 16. Data Processing for an Array Software coding schematic for processing the data in a 480 x 1 array.
- 17. Data Structure: Super Cluster
- 18. This Super Cluster data structure has the following defined properties: 1. There are 8 parameters defined for each sensor 2. These parameters, plus a validity flag, are encapsulated in a data structure known as a cluster 3. There are 6 sensor clusters (s1 - s6) for each mirror segment 4. The 6 sensor clusters are encapsulated in a larger mirror segment cluster 5. The mirror segment clusters are then contained in an array with 91 elements (S1 - S91) The advantages of the super cluster are: 1. extremely human reader friendly 2. all sensor data for one segment can be viewed and compared in a table-like form 3. navigating between mirror segments is simple 4. correlation of sensor number to segment side is obvious 5. additional parameters can be added or removed at anytime, requiring minimum changes to the data processing structures in the software.
- 19. Data Processing for a Super Cluster
- 20. With the structure processed as in the diagram above the loop only needs to run 91 times whilst processing the sensor parameters simultaneously in identical parallel reentrant functions. The disadvantages of the super cluster are: 1. the case statement containing the 6 sensor data processors must be duplicated 90 times when coding, however this is a once off event 2. initial coding is slightly more complex, but with proper planning pays off in the long run Although this data structure has a processing speed ~3 times longer than the array and uses more memory, the benefits of ease of use in the data processing software and it’s human friendly readability outweigh it’s more costly use of computer resources.
- 21. Data Display in a Table This is the super cluster translated to a table on the user interface, which in fact is a direct mimic of the data structure. Left is a table of the mirror actuator displacement information ordered by segment number.
- 22. Typical Software Coding Schematic Typical coding diagram of a threshold hysteresis comparator that can be found on the web.
- 23. Threshold Hysteresis Comparator Note: this implementation is specifically for the mirror segment edge sensor which, when the primary mirror is perfectly aligned, should produce an output very close to zero. The detector has two upper and two lower threshold limits which allows an input (for example, a sensor parameter) to drift out of an inner limit without generating an invalid flag, but if it drifts outside an outer limit it then becomes invalid. For the reverse, when the input crosses back into the outer limit it remains invalid until the value falls inside the inner limit, the hysteresis effect. For simplicity, only one input parameter is provided for both the inner & outer limits, and is negated for the lower limit, as generally tolerance limits are symmetrical to the nominal.
- 24. Demonstration of a threshold hysteresis comparator
- 25. Software Design & Development by Deon Bester with collaboration from Dr John Menzies, Hitesh Gajjar, Janus Brink, Deneys Maartens, Anthony Koeslag, Stephen Hulme, Paul Rabe Paper: D Bester “Software architecture for processing a large data array” CCAS Centre of Curvature Alignment System DAQ Data Acquisition GRoC Global Radius of Curvature ICD Interface Control Definition MACS Mirror Alignment Control System ODM Object Data Manager PMAS Primary Mirror Alignment System MARS Mirror Alignment Reference System SDP Sensor Data Processing SHS Shack-Hartman (wavefront) Sensor SPS Segment Positioning System UDP User Datagram Protocol ACRONYMS
Editor's Notes
- Today I’d like to present my edge sensor project which I completed on the Southern African Large Telescope this year. This is SALT, located in the Northern Cape province of South Africa, in the arid Karroo desert at an altitude of just over 1700m (~5700 feet).
- Some key components of the telescope are: 91 segment primary mirror tracker with payload at the optical focus of the primary mirror CCAS tower housing the Shack-Hartmann wavefront sensor for aligning the mirror primary mirror structure is tilted at 37 degrees from the zenith
- Sensor Numbering System 6 sensors per segment (one per side) Sensors numbered as 1 to 6 on segment, clockwise from top Globally sensors numbered from 1 to 480 Arithmetically >>> 91 x 6 = 546 sensors Sensors mounted on outer edge of outer ring where there are no adjoining segments are disabled in the system
- Sensor Configuration 3 Sensor Segment Corner configuration with complimentary Transmitters and Receivers on adjacent segment edges 3 hardware channels per rack module optimises sensor cabling Requires remapping of hardware channels to sensor numbers in the software The allocated sensor number always refers to the Receiver device, as this is effectively the sensor, whilst the Transmitter acts as the reference signal.
- Demonstration of the threshold hysteresis comparator.