RTC Magazine

They call them “sky islands,” these mountainous outcroppings that rise suddenly from the Arizona high desert. As if separated in an ocean, each has its own ecosystem and unique character. The highest of these, Mount Graham, rises to an altitude of 10,700 feet above sea level. It is here that the University of Arizona’s Steward Observatory in partnership with the Max Plank Institute and other participants, is building what will for a time be the world’s most powerful optical telescope, the Large Binocular Telescope (LBT).

When it is completed, the LBT will have two 8.4 meter mirrors, each of which is over half again the diameter of Palomar. When used together, the mirrors will provide a diffraction-limited image sharpness of a 22.8 meter aperture. Astronomers will be able to use the mirrors in tandem or individually. For example, different instruments on each mirror could capture optical and infrared images of the same object. The binocular design enables a number of unique instrumentation and research possibilities.

The LBT weighs approximately 580 metric tons and has a combination of focal stations for different instruments and observation tasks (Figure 1). Accurately pointing, focusing and controlling a precision instrument of this size is an enormous challenge, especially in the light of external factors such as temperature variations and the buffeting of high winds at such an altitude. It is a challenge that could only be met and overcome through the application of powerful embedded and real-time computing technology consisting of a combination of COTS and custom hardware and software. The open-loop, full-sky pointing accuracy is currently 0.3 arcseconds with a tracking performance of 0.01 arcseconds.

Like the end users of any other embedded system, astronomers do not want to have to worry about the details of pointing and controlling the telescope. They want it pointed and focused accurately and reliably so that they can effectively use their instruments and obtain the data needed for their science projects. So the engineering issues in question here are with those tasks rather than with the scientific instrumentation.

Basically, the LBT is what is called an “altazimuth” mount. That is, one axis turns in relationship to the surface of the Earth, the azimuth, while the other moves up and down in elevation. This is different from more traditional telescope equatorial mounts in which one axis is aligned with the axis of Earth’s rotation. Equatorial mounts become prohibitive in size for telescopes beyond a certain aperture. To point an altazimuth telescope, a pointing control system must translate the standard location of an object in the sky, which is cataloged in terms of right ascension with respect to 0 degrees longitude, and declination with respect to the equator into altitude and azimuth coordinates for the telescope. These are based on its geographical coordinates and the local time. But that’s the easy part. Once pointed, the instrument has to track the object, incrementally changing its altitude and azimuth settings, as the Earth moves on its axis.

In addition to the pointing problem, there is also the issue of optical accuracy. Corrections must be done through a system of adaptive optics to cancel the distortion of the Earth’s atmosphere and the stresses from tilting, and temperature on the 8.4 meter primary mirrors must be sensed and corrected to keep a perfect shape. In the case of the primary mirrors these are small increments, but we are dealing here with the wavelengths of light, and each mirror weighs 41,500 pounds.

Accurately moving such a massive piece of equipment is done by essentially “floating” it on a film of oil supplied by hydrostatic bearings and using four powerful motors on each axis with position information gathered from a set of inductive strip encoders mounted on the outer radius of each axis. It also involves a dynamic balance system, which operates by shifting water in and out of ballast tanks to keep the telescope in balance. Three major control systems include the Mount Control System (MCS), the Adaptive Optic System (AOS) and the Primary Mirror Controller (PMC). These operate under the main Telescope Control System (TCS), which includes, among others, the MCS GUI and the Pointing Control System (PCS).

The mount control system operates the altitude and azimuth controls, the swing arms, which are used to bring instruments and the secondary and tertiary mirrors into one of the focal points of the telescope, the instrument rotators and the enclosure rotation system. In addition, it interfaces with the stand-alone hydrostatic bearing system, mainly to control the pumps, which provide a film of oil at 1800 psi so that the enormous thing can be moved at all.

The mount control system computer is a CompactPCI system from BittWare (Figure 2), which includes an Intel processor-based CPU board running Linux and other boards supporting ten Analog Devices SHARC 21160 DSPs, two Xilinx Virtex II FPGAs as well as subsystems for real-time clock and analog inputs and outputs. Each axis has a cluster for four DSPs plus one FPGA card dedicated to it. Non-real-time state logic and communications tasks run under Linux. The card is a BittWare HH6U cPCI card with eight SHARC processors plus two daughter boards each with an FPGA and yet another processor for I/O. So this one cPCI card handles the servo logic for both the azimuth and the elevation axes. Much of the DSP code and all of the Linux-to-DSP interface was written by Dan Cox.

According to software engineer, Tom Sargent, with a 2 kHz control loop and four dedicated DSPs, there in no need for a real-time version of Linux or an RTOS. “With four DSPs and all kinds of high-speed interfaces between the link ports, which we use, we can have them all operate round robin deterministically—just by using the link ports and interrupts and telling them, ‘I need you to do this now,’ and the DSP will go ahead and do whatever it needs to do.”

Among the things the DSPs are juggling are the inputs from the encoders, which have a specified resolution of 0.005 arcseconds but thanks to interpolation performed by the DSPs, this has reached a repeatable resolution of 0.002 arcseconds. The DSPs resolve the commands and polynomials sent from the PCS into control algorithms and commands to the positioning motors as well as taking in aberrations caused by wind and making minor corrections, plus tracking the object as the Earth moves while also controlling the instrument derotators (Figure 3).

The latter are used to correct a problem inherent in altazimuth mounts. In an equatorial mount with one axis aligned to the axis of the Earth, the telescope can track an object in the sky without the field rotating around it. With an altazimuth mount, the camera or instrument must be rotated in the opposite direction to compensate for field rotation. Tracking is done, according to Sargent, by the PCS emitting polynomials for both axes and the appropriate derotators, which fortunately have drive motors and shaft encoders similar to the motors controlling the axes.

In the case of the axis motors, one pair supplies torque to move the azimuth axis on its hydrostatic bearings, while the other pair supplies some reverse force to eliminate gear play. The circular azimuth axis as well as a pair of large C-rings used to drive the altitude axis, are mounted with large rim gears driven by the motor pinion gears. The motors are also equipped with optical shaft encoders.

The observatory building moves independently of, but in synchronization with, the telescope. It has large, vertical openings that are open during observing sessions in order to keep the optics as close to ambient temperature as possible, but it still presents a huge profile to the wind. The building moves on four large bogies (Figure 4), each of which has two large AC motors. The building, under control of the enclosure rotation controller (ERC), must follow the telescope within =/- 1.5 degrees under all conditions including wind speeds up to 80 km per hour.

The interface between the MCS and many of these subsystems, except for the main axes and instrument derotator hardware, takes place via an Allen-Bradley ControlLogix programmable logic controller (PLC). Referring to the MCS, Tom Sargent remarks, “It’s amazing how this thing can make a 600-ton telescope twitch.”

There are certain other factors that affect the pointing accuracy of the telescope but are not under the control of the MCS. The MCS can detect and correct for wind disturbances on the telescope (as opposed to the enclosure) in real time by means of the inductive encoders and the MCS software. However, the effects of temperature and flexure of the structure cannot be detected by the MCS and must be corrected optically. These involve the primary mirror controller (PMC) and the adaptive optics system (AOS). Flexure in the structure is about 1 mm top to bottom and can mostly be taken care of by adjusting the primary mirror. The primary mirror is big and heavy and does not lend itself to real-time control

The AOS corrects for atmospheric turbulence by changing the secondary mirror shape to cancel the refractive effects of the atmosphere. The application of adaptive optics corrections to the secondary mirror as a facility instrument is a first in astronomy. Normally, it is done by means of a much smaller forth or fifth mirror in the optical path. The AOS consists of a wavefront sensor, which detects the optical distortions in the atmosphere and sends data to a custom-built DSP box called the “slope computer” that computes the needed shape of the secondary mirror in terms of an array of slopes. The slope computer then sends commands to an array of 672 actuators that bend the mirror at an update rate of 2 kHz. That is, the system can sense, compute and reconfigure the mirror 2,000 times per second, including the latency of the actuators.

The computational task is immense. The wavefront sensor consists of a pyramidal lens over a two-dimensional sensor. The sensor is looking at a reference star during observation. The patterns formed on the sensor must be computed as two-dimensional parameters then into slope data and finally into positioning commands to the actuator array. Each DSP controls four actuators for a total of 168 DSPs in the system. A slope is formed by the position of sensors in relationship to each other over the back of the mirror. In addition, the software includes safety checking so that no combinations arise that could break the glass mirror. The secondary mirror in the LBT is 911 mm in diameter and the glass shell is only 1.6 mm thick.

The third major control system in the LBT is the primary mirror controller—one for each mirror. Each primary mirror is made of a single piece of borosilicate glass weighing 41,500 pounds and is 0.894 meter thick at the edge, narrowing due to the convex shape to 0.437 meter at the center. The mirrors are thinner at the center (where there is a 1-meter hole for a focal station) because the concave surface was formed by melting the glass in a 10-meter diameter rotating furnace and then cooling it. During this process a honeycomb pattern was formed in the back by the melted glass shaping around forms set up in the bottom of the mold. A mirror of this size will experience some change of shape due to sagging when the telescope is repositioned and temperature can also affect the mirror shape (Figure 5).

Each PMC is built on a VME rack that includes a Cougar-10 board based on the PowerPC 7410 running at 550 MHz from Curtiss-Wright Controls Embedded Computing. The Cougar-10 boards run VxWorks from Wind River Systems. According to Christopher Biddick, the software engineer responsible for the PMC, the I/O subsystem includes RS-422 serial interface cards, a 48-bit parallel card and a 12-bit A/D card all from Acromag. These go out to custom cards based on the 18F252 PIC microcontroller from Microchip Technology that control the actuators in the primary mirror cell.

One VME card controls one mirror via 160 pneumatic actuators glued to the back of the mirror. 108 of these are dual-axis actuators—one vertical and one attached at a 45° angle—and another 52 are single-axis actuators. This arrangement gives the mirror six degrees of freedom: in the X, Y and Z directions and in “moments” around each X, Y and Z axis. Six so-called “hard points” measure the position and force exerted by the mirror on them and this data is used to control the actuators.

When the telescope tilts, there are slight but measurable effects from gravity that change the mirror’s position and its shape—again, we are dealing with the wavelengths of light. The data from the hard points is computed to issue commands to the actuators in order to equalize the forces on the hard points and assure a perfect shape. As with the secondary mirror, there are combinations of forces, which, if applied, could break the mirror. So the software must guard against any and all such cases.

The effects of temperature are also monitored and controlled by the PMC, mainly by maintaining the mirror as close to ambient temperature as possible. The specifications call for a temperature gradient of less than 0.1°C across the entire surface of the mirror. The honeycomb pattern in the back of the mirror is ventilated and the system monitors the temperature to keep it within parameters.

According to Biddick, there are other corrections the PMC can make. “If, for example, there is astigmatism or other aberrations, then we have the math that will take Zernike coefficients from a camera or imaging system and make corrections.” These coefficients are converted into what are called “surface displacement errors.” The bending modes of the mirrors were calculated using finite element analysis, and, given that knowledge, when there is a need to change the surface displacement the system is able to apply the proper forces based on the bending mode distribution.

The Large Binocular Telescope has so many degrees of freedom and so many opportunities for adjustment that a sort of macro-system may be needed to negotiate among the possibilities for the optimum solution. So much precision and control has been applied to such a large device because its assignments will be truly astronomical. The realistic hope is that with this Earth-based instrument it will be possible to get far better images than currently possible with the Hubble telescope based in space. The instrument packages that will be used with it demand the utmost in pointing precision and image stability. These are solutions that have been and are being achieved with the concerted application of embedded and real-time computing technology through which we may actually find new worlds supporting life.

Why a Binocular Telescope?

There are a number of instruments and experiments that can take advantage of the Large Binocular Telescope. For example, there is a high-resolution spectrograph with dual independent polarimeters that will be able to simultaneously linearly and circularly polarize light with high spectral and temporal resolution. Another advantage is interferometry–the ability to combine two input waves of light to produce a different output wave. By combining the waves at the same or different phases, they can be added to each other or made to cancel each other.

One interferometer, called LINC-NIRVANA, will provide a beam combiner to allow imagery over a wide field of view. When used with the LBT in so-called Fizeau mode, it will provide the sensitivity of a 12-meter telescope and the spatial resolution of a 23-meter telescope over a field of approximately 10 to 20 square arcseconds.

Another interferometer is the NASA-funded LBT Interferometer (LBTI). It is partly being developed to study the technique that will be used later in NASA’s orbiting terrestrial planet finder (TPF). On the one hand, the LBTI will be able to combine the light from the two beams for an effective aperture of 23 meters resulting in an ability to produce images ten times the sharpness of the Hubble Space Telescope. Its other use for planet detection will be as nulling interferometer.

The difficulty with discerning terrestrial-size planets around stars is that such planets are so small with respect to the star and so relatively close that they cannot be seen due to the glare of the starlight. The LBTI incorporates a universal beam combiner, which brings in the two beams to a common focal plane and the nulling interferometer, which adjusts the phases of the wavelengths of starlight to suppress the light from a star. A nulling optimized mid-infrared camera (NOMIC) will then be used to form an image of the field around the star and detect infrared emissions from any surrounding disks of dust and/or planets thanks to the high resolving power of the telescope.

The plan is to survey a large sample of nearby stars to try to determine what percentage of stars may have make-ups similar to our solar system and which stars may potentially have terrestrial-sized and potentially life-bearing planets. The holy grail of such a search would, of course, be to detect the spectral lines for water.

The LBT is an international collaboration among institutions in the United States, Italy and Germany. The LBT Corporation partners are:
• The University of Arizona on behalf of the Arizona universities
• Istituto Nazionale di Astrofisica, Italy
• LBT Beteiligungsgesellschaft, Germany, representing the Max Planck Society, the Astrophysical Institute Potsdam, and Heidelberg University
• The Ohio State University
• The Research Corporation, on behalf of The University of Notre Dame, University of Minnesota and University of Virginia

Large Binocular Telescope Observatory
University of Arizona.
[www.lbto.org].