ESO is currently in the final phase of the standardization process for PC-based Programmable Logical Controllers (PLCs) as the new platform for the development of control systems for future VLT/VLTI instruments. The standard solution used until now consists of a Local Control Unit (LCU), a VME-based system having a CPU and commercial and proprietary boards. This system includes several layers of software and many thousands of lines of code developed and maintained in house. LCUs have been used for several years as the interface to control instrument functions but now are being replaced by commercial off-the-shelf (COTS) systems based on BECKHOFF Embedded PCs and the EtherCAT fieldbus. ESO is working on the completion of the software framework that enables a seamless integration into the VLT control system in order to be ready to support upcoming instruments like ESPRESSO and ERIS, that will be the first fully VLT compliant instruments using the new standard. The technology evaluation and standardization process has been a long and combined effort of various engineering disciplines like electronics, control and software, working together to define a solution that meets the requirements and minimizes the impact on the observatory operations and maintenance. This paper presents the challenges of the standardization process and the steps involved in such a change. It provides a technical overview of how industrial standards like EtherCAT, OPC-UA, PLCOpen MC and TwinCAT can be used to replace LCU features in various areas like software engineering and programming languages, motion control, time synchronization and astronomical tracking.
Most of the real-time control systems at the existing ESO telescopes were developed with "traditional" methods, using
general purpose VMEbus electronics, and running applications that were coded by hand, mostly using the C
programming language under VxWorks.
As we are moving towards more modern design methods, we have explored a model-based design approach for real-time
applications in the telescope area, and used the control algorithm of a standard telescope main axis as a first example.
We wanted to have a clear work-flow that follows the "correct-by-construction" paradigm, where the implementation is
testable in simulation on the development host, and where the testing time spent by debugging on target is minimized. It
should respect the domains of control, electronics, and software engineers in the choice of tools. It should be a targetindependent
approach so that the result could be deployed on various platforms.
We have selected the Mathworks tools Simulink, Stateflow, and Embedded Coder for design and implementation, and
LabVIEW with NI hardware for hardware-in-the-loop testing, all of which are widely used in industry. We describe how
these tools have been used in order to model, simulate, and test the application. We also evaluate the benefits of this
approach compared to the traditional method with respect to testing effort and maintainability.
For a specific axis controller application we have successfully integrated the result into the legacy platform of the
existing VLT software, as well as demonstrated how to use the same design for a new development with a completely
Control of primary segmented mirror of an extremely large telescope with large number of actuators and sensors
and multiple control loops is a complex problem. The designer of the M1 unit is confronted to the dilemma of
trade-off between the relatively though performance requirements and the robust stability of the control loops.
Another difficulty arises from the contradictory requirements of the stiffness of the segment support system and
position actuators for wind rejection on one hand and vibration mitigation on other hand. The presence of low
frequency mechanical modes of the back structure and possible interaction of the large number of control loops
through such structure could be a limiting factor for achieving the required control bandwidths. To address these
issues a better understanding of dynamical behavior of segmented mirror is necessary. This paper addresses the
trade-offs on dynamical aspects of the M1 segmented mirror and the robust stability conditions of various control
Associated to tracking capabilities, the main axes control system of the E-ELT is the first correcting system in the
chain of control loops for reducing the image motion (tip/tilt) caused by perturbations on the telescope. The main
objective of the closed-loop performance analysis of the axes is to evaluate the trade offs for the choice of control
system hardware, i.e. specification and location of the motors and sensors (encoders/tachometers). In addition,
it defines the design constraints and requirements (actuator stroke and bandwidth) of other correcting systems
in the chain: the field stabilization (M5 unit) and adaptive deformable mirror (M4 unit). In this paper the main
axes control analysis of E-ELT is presented and the performance of telescope in face of external perturbations
such as wind and imperfections of the drive (cogging/ripple) and sensing (noise) systems is evaluated. The
performance metric is the wavefront error at the focal plane which is derived from the mechanical motion of the
telescope's optical elements together with their respective optical sensitivities.
The 2.6m VST telescope is in installation phase in the ESO observatory of Cerro Paranal. After preliminary tests in
Europe performed jointly by INAF and ESO the tracking system was considered at the readiness level to be shipped to
Chile. The motion control system has already been reintegrated in Chile and is operational again. The final tuning is
going to be performed after the integration of all the telescope subsystems, still in progress. Therefore here the focus is
especially on tests performed in Italy. This paper describes the solutions adopted for the telescope main axes control as
well as the preliminary tracking results. Available test data are related to encoder feedback. Tests have been performed
tracking coordinates of virtual objects. A comprehensive test case to evaluate the performance of different controllers
was needed to proceed in a systematic way. A tracking map derived from the VLT commissioning experience has been
used, spanning all the different conditions for axes positions and speeds, including meridian crossing and tracking near
the blind spot.
During the past year the control of the 42m segmented primary mirror of the E-ELT has been studied.
This paper presents the progress in the areas of M1 figure control and control hardware implementation. The critical
issue of coupling through the supporting structure has been considered in the controller design. Different control
strategies have been investigated and from a tradeoff analysis modal control is proposed as a solution addressing the
topics of wind rejection as well as sensor noise in the presence of cross-coupling through the supporting structure.
Various implementations of the M1 Control System have been studied and a centralized architecture has been selected as
baseline. This approach offers maximum flexibility for further iterations. The controller design and main parts of the
control system are described.
Direct drive motors are finding their way into telescope drive designs and have many advantages over more traditionally used friction and rack/pinion drives. They are implemented as curved linear motors custom made to the actual diameter required by the telescope design. Custom design is by definition a costly process and this paper discusses the possibility to use standard straight linear segments to approximate the curvature. Detailed analysis is given concerning differences in the electro-magnetic properties compared to a custom design motor.
Wind disturbance is an important parameter in the assessment of telescope performance, especially for extremely large telescopes like TMT. In the process of estimating the response of the TMT telescope structure to wind buffeting, we were carrying out measurements on the Keck and Gemini observatories in order to better understand the possible variations in the wind response of these different telescopes. The measurements were used to validate the TMT wind model as well as explaining the mechanism through which the wind disturbances affect the pointing of the elevation axis. We were estimating the wind torque acting on the elevation axis from the measured Disturbance Transfer Function and Position Error of the mount control. Furthermore, the quality of the correlation between wind speed and torque disturbance were investigated to evaluate the feasibility of mitigating the wind disturbance by introducing feed-forward or adaptive control schemes in the mount control.
Traditionally telescope main axes controllers use a cascaded PI structure. We investigate the benefits and limitations of this and question if better performance can be achieved with modern control techniques. Our interest is mainly to improve disturbance rejection since the tracking performance normally is easy to achieve. Comparison is made to more advanced controller structures using H-∠infinity design. This type of controller is more complex and needs a mathematical model of the telescope dynamics. We discuss how to obtain this model and also how to reduce it to a more manageable size using state of the art model reduction techniques. As a design example the VLT altitude axis is chosen.
On March 17, 2001, the VLT interferometer saw for the first time interferometric fringes on sky with its two test siderostats on a 16m baseline. Seven months later, on October 29, 2001, fringes were found with two of the four 8.2m Unit Telescopes (UTs), named Antu and Melipal, spanning a baseline of 102m. First shared risk science operations with VLTI will start in October 2002. The time between these milestones is used for further integration as well as for commissioning of the interferometer with the goal to understand all its characteristics and to optimize performance and observing procedures. In this article we will describe the various commissioning tasks carried out and present some results of our work.
As a further step to improve the excellent tracking performance of the VLT telescopes, the intrinsic errors in the telescope drive systems are analysed. These errors fall into two categories, torque disturbances and sensor errors and they have different impact on the performance. Models for the errors are developed and algorithms for on line adaptive parameter identification are presented. The models can be used to significantly reduce the influence of the errors and also to monitor parameters like friction and unbalance. The VLT servo model is used to test and verify the models and algorithms. It follows a description of the real-time software aspects of the algorithms, which have been implemented for VxWorks-based systems. The software design allows various options for the adaptation of the process coefficients, either running permanently in background, only on demand through maintenance procedures, or fixed off-line modeling based on recorded process data. Finally, real test data are presented.
The large direct drive motors and encoders form together with the control system a high performance telescope exhibiting very high tracking accuracy. This paper describes the integration and fine-tuning of the VLT Drive Systems. It discusses the different problems encountered during the integration. The servo model that was used to simulate the problems and to find new solutions is described as well as test results and advanced analysis methods.
This paper outlines the design of an auto guiding system and points out the potential pitfalls that often are not observed. Rather than describing the mathematics in detail, the fundamental principles are discussed at a more basic level, leaving the interested reader to explore the details in the mathematical textbooks himself. The diagrams in this paper all made through simulations rather than mathematical calculations.
Tracking errors of any kind will degrade the image quality of a telescope. By using a CCD camera to record image motion during blind tracking it is possible to analyze the resulting data and identify the origin of these errors. The part of the image motion which is caused by seeing effects has been analyzed in numerous papers and is not discussed in this report, which will instead concentrate on telescope and servo errors. A strategy for analyzing and identifying these errors is outlined and suggestions for further analysis are given.