|
Документ взят из кэша поисковой машины. Адрес
оригинального документа
: http://www.apo.nmsu.edu/Telescopes/eng.papers/milt/telescopecontrol.html
Дата изменения: Mon Dec 13 18:50:17 1993 Дата индексирования: Sun Apr 10 02:22:57 2016 Кодировка: |
Several peripheral microcomputers implement control functions locally near the telescope. The microcomputers effectively relieve the TCC from a large amount of real time control and create convenient modularity of system hardware functions.
Figure 4.1 shows the interconnection of the components of the telescope control system. The system is hierarchical with the telescope control computer (TCC) at the top.
The TCC is connected to the azimuth controller (AzC), the altitude controller (AltC), the secondary controller (SC), tertiary controller (TC), the bright star camera controller (BSCC), the two instrument changer controllers (ICC1 and ICC2), the weather monitor controller (WMC), and the wind screen controller (WSC).
The AzC and AltC are identical hardware and run essentially identical software except for the values of a few constants. Each implements the digital state variable control system for that axis. The SC controls the tilt and piston of the secondary. The TC controls the position angle of the tertiary, field rotation for the two Nasmyth field cameras, and instrument selection in the two instrument changers.
ICC1 and ICC2 each control field de-rotation for one instrument changer for those instruments which require this service. The BSCC locates and centroids the brightest star in its field and returns this value to the TCC upon request. The WMC supplies the TCC with real time weather data including temperature, pressure, and humidity. The WSC controls the position of the wind screens.
The TCC and BSCC are located in the telescope control room in the main building. The WMC is mounted along with its transducers on top of a tower near the telescope enclosure. The other controllers are mounted in the telescope enclosure in close proximity to the device(s) they control.
The datalink between the controllers and the TCC is a hierarchical bus which uses the RS-232 protocol for asynchronous serial data communication. The long link between the control room and the telescope will be a fiber optic which will fan out into a fiber optic star configuration at each end thereby providing the necessary ports.
The TCC will be the bus controller and the other devices will send data only when requested by the TCC. This will prevent collisions and allow the TCC to schedule time critical tasks first when necessary.
Electrical code transition angles for the absolute encoder repeat to about 1arc sec. The coupling ratio for the incremental encoder is determined so that each increment corresponds to 0.04 arc sec of telescope axis rotation.
Each encoder is interfaced to a real time microprocessor controller. Using the incremental encoder as an interpolator between absolute encoder transitions, the system software synthesizes in effect a virtual absolute encoder with 25 bit resolution, 1 arc sec accuracy over large angles and high relative accuracy over smaller angles.
The Smithsonian Astrophysical Observatory (SAO) catalog is readily available in machine readable form. In a 3.5 meter telescope all 260,000 SAO stars are bright enough for real time acquisition using an inexpensive unenhanced CCD surveillance camera operating at the focus to encode a small patch of sky and serve as a 'crosshair'. Since they are so inexpensive, it may be desirable to provide one at each Nasmyth focus devoted entirely to the offsetting function and fed by retractable pickoff mirrors.
We have conducted controlled tests of our own in the UW Physics machine shop where several machine tools utilize retrofitted friction wheel micrometers to encode tool motions over distances of 0.5 to 1 meter.
These devices reproduce position readings to no worse than 1 micron (the dial reading limit) after many arm-tiring manual cycles of motion. At the edge of a 4 meter disk, 1 micron over 0.5 meter corresponds to about 0.1 arc sec over 14 degrees.
This suggests tracking with less than about 1 arc sec accumulated error may be possible over the entire sky (180 degrees). Existing telescopes with friction encoders do not currently perform this well all night. Thermal gradient drive disk distortion is surely a factor here as well as contact microslip. Our conservative encoder system design does not depend on the incremental encoder to do offsets over angles greater than about 40 arc min, the angle between two absolute encoder transitions.
The incremental encoder will be used for movements over all angles for which accumulated errors are smaller than the absolute encoder accuracy (1 arc sec rms).
The very familiar alternative of star-tracked offset guiding of course will not be excluded.
There are of course offset stars available during the day, and observable using the CCD encoding camera. The smaller number of stars available will correspondingly degrade offsetting precision.
These elements and their coupling can be represented by linear, ordinary differential equations, thus providing a state space representation of the system (Kuo 1982 and Electrocraft Corp. 1980). For the selected motor and drive characteristics this model will be 4th order for each axis. From this representation, transfer functions can be derived to permit frequency response analysis and the application of classical linear control theory. A major concern in the derivation of these models will be to specify the range of variation of parameters in the models (e.g. inertia) so that controllers of sufficient robustness (i.e. with sufficient margins) can be designed to assure stability and good performance under all operating conditions.
Characteristics to be expected in the design will include: 1) some form of notch filter to improve damping of LRRF while providing a relatively wide bandwidth (e.g. 5-10 Hz) to counter turbulent wind loading, and 2) sufficient integral control to permit following a varying rate command without significant pointing error. The use of a digital microprocessor to implement the compensation, and the inherent discrete character of the encoder output, suggests a direct digital design for the compensators.
Computer tools exist at UW to perform both multivariable and classical single loop analysis and design for digital control systems. Thus, compensator design is expected to proceed in a straightforward manner.
Among the types of effects that will need to be assessed are the following:
To assess fully the impact on performance of these effects, a detailed computer simulation of the system will be needed. Such a simulation can be used in a variety of ways, including assessing performance of controller designs in the presence of the above nonlinearities, verifying control algorithm software before it is implemented in the telescope computers, and aiding in the assessment of performance anomalies once the telescope is in operation. A trade study on the approach to development of this simulation capability will need to be carried out.
The output of the encoder buffer goes into a 16 bit parallel processor port. The output of the absolute position encoder goes into another processor parallel port. The absolute encoder is used to implement software limits to motion by the axis controller and is sent on demand to the telescope control computer which uses it as part of its determination of the zero point for the coordinate system defined by the incremental encoders.
The velocity of the motor shaft is sensed by a tachometer integrated with the motor. This signal is digitized by an A/D converter. Information from both the motor shaft and the telescope drive disk allows the control system to calculate the amount of wind up in the drive train and to correct for this effect.
The processor implements the digital state variable control system. This sampled data system operates at approximately one hundred cycles per second; well above the bandwidth of the servo loop. Each cycle the processor estimates the current state of the system based on the transducer outputs. It calculates the desired state of the system based on commands from the telescope control computer. Then it generates an output to achieve that state.
The digital output is converted to an analog output by a D/A converter. The low level analog output command is amplified by the motor controller and used to drive the motor. The motor is coupled to the telescope axis by azero backlash speed reducer which turns the drive roller. This roller is pressed against the axis drive disk with a contact force sufficient to insure that slippage will not occur.
The TCC will be largely insulated from the real time requirements of the task by the controller components of the system. The TCC provides the user interface between the astronomer and the telescope or between the astronomer's computer and the telescope. As such it will accept high level pointing and tracking commands. The SAO catalog will be read accessible to the users and utilities will be provided which will allow the users to create and edit their private catalogs of objects which they wish to observe. We expect that most observers will wish to grant read access to their program object files but file protection will be available. The telescope will point to objects specified by their coordinates in arbitrary epoch or will search in the designated files for the coordinates of objects specified by name.
The SAO catalog is readily available in machine readable form and is a compilation of objects from other catalogs. Because of this its overall precision is nonuniform, but on the average is better than about 0.5 arc sec; estimated position errors, proper motion and the estimated error in the proper motion is included with each entry. The SAO catalog contains more than 250000 stars and nearly every point on the sky is within 0.5 degree of a star in the SAO catalog.
The TCC will provide a status display indicating where the telescope is pointed in right ascension and declination and galactic coordinates, the number of air masses in the optical path, the current time, etc.
During aquisition of a new object the TCC will periodically reset the zero points of the alt-az coordinate system by pointing the telescope at a SAO catalog position reference star and requesting the field camera controller for the position of the star centroid in the field. Alternatively it can request the axis controllers for data from the absolute encoders which in turn can be used to reset the zero points.
During tracking the TCC sends the axis controllers the velocity commands necessary to track a celestial object. It takes into account the alt-az RA-dec coordinate transformation, telescope flexure, atmospheric refraction, mechanical telescope misalignments, etc. The TCC will perform closed loop tracking control by periodically requesting the telescope position from the axis controllers and sending correction commands as required.
There will be three levels of encoder calibration: At installation time, when initial setup parameters for the telescope are determined; after occasional power failures; and every time a slew occurs to a new object.
Next a bright star is boresighted onto the CCD camera frame. Initially this will require operator participation. Noting the time and the pixels on which the star image falls, the computer, using the known star position, now calibrates telescope coordinates. In this context the CCD pixels serve the function of a crosshair. At this time the fiducial positions in alt and az space (referred to the TV frame) are determined once and for all. These values are also stored on non-volatile storage and the telescope system is now ready to perform a power up cold start at any time.
The scale of incremental encoder tick per telescope axis rotation and the relative positions of absolute encoder transitions in practice will fail to repeat at some level. Frequent re-determination of these values is expected to be a routine engineering function of the TCC software.
At a cold start, the telescope computer first rotates the telescope until an absolute encoder CHANGE is detected. The incremental encoder buffer register is initialized at this position. The 25-bit virtual absolute encoder is now initialized 1 arc sec rms accuracy.
We propose to largely avoid this problem by incorporating software in the axis controllers which will check commands from the TCC for possible errors before execution. The axis controllers will also check for consistency between the telescope position and rate transducers since the position signal can be differentiated to give a rate. The axis controllers will assume that an error in position which cannot be corrected in a reasonable time is due to a fault such as D/A or amplifier failure or an external obstruction or load and will cut power to the motors and set brakes.
A watch dog circuit will be implemented for each controller. In normal operation the controller software will reset the watch dog every 100 milliseconds. If the controller fails to reset the watch dog for any reason (e.g. if the program bombs) this circuit will cut power to the drive motors. The TCC will provide high level monitoring of the axis controllers for faults.
The philosophy in the control system design is to detect all single point failures and stop the telescope before any damage or injuries occur and to identify the failed module to facilitate rapid repair or replacement.
Other alternatives include using a pair of spur gear reducers driven by two separate motors with a DC offset so that one is preloaded against the other. The extra complication of this design is to be avoided if possible. Another possibility is that the drive roller could be driven directly with a150 N m torque motor. A trade study will be done to choose between these alternatives.
One constraint which has not been investigated in detail at this time is the operating temperature range specification. Since we require operation down to -20 C we require better temperature specs than standard consumer specs. The effect of this requirement on cost and availability has not yet been determined; it can be avoided by providing temperature controlled enclosures for the electronics. Since the enclosures would be small they could be insulated very well and would require very small amounts of heat to meet temperature specs of consumer parts. Still the simplest solution is to buy electronics which will run at the ambient temperature.
The software for the axis controllers will be written in FORTH, a language which is unexcelled for real time control applications. FORTH executes rapidly; time critical functions can easily be recoded in assembly language when necessary.
The TCC will be a 68000 class microcomputer interfaced to a hard disk drive with approximately 20 megabytes capacity. The TCC will be connected to the data aquisition and analysis computer and its disk and tape drives most likely via Ethernet. The TCC software will be written in a high level language such as Pascal or C under a standard operating system such as UNIX.
Figure 4.3, provided by the encoder's designer Paul Johnson of BEI shows the results of high precision optical autocollimation measures of 32 code transition angles for the 5V682A both before and after a rugged environmental test. The mean of before-after residuals is 0.44 arc sec with arange of 2.5 arc sec peak to peak.
The cost is $3800 for the 16-bit version. A newer less expensive 16-bit encoder, model M25 ($2200) is also available. It is equally repeatable and more rugged. These prices represent upper limits to our cost as we do not anticipate the need for the 16-bit version.
After azimuth axis installation, a fixture is bolted concentrically to the stub providing complete load isolation between the driveshaft and the encoder. The hollow tubular driveshaft is suspended at its ends by lash free flex-disk universals. The flex disks will be stiff enough to support the drive shaft weight without themselves distorting excessively.
The isolator shaft couples telescope rotation to the encoder via a precision bellows flex-coupling allowing some shaft misalignment tolerance without lash and isolating the encoder from all significant shaft loads. Once installed, the isolation fixture remains in place and allows independent installation and removal of the encoder and driveshaft.
Basic code disk 2500 line pairs/rev Electronic multiplication 40x Effective counts/turn 100,000 Average resolution 13 arc sec/step Step size jitter 7 arc sec Disk runout error 20 arc sec Maximum output rate 300 kHz at output
Alternatively the system could be designed to use only the low resolution absolute encoder during slewing thereby relaxing the high speed requirement for the incremental encoder.
It would be feasible to operate a second redundant incremental encoder on each axis due to their low cost.
However since the encoder tangent forces will be nearly nonexistent, we violate this condition for the altitude axis in order instead to implement alonger column whose hinged end is fixed near the azimuth load bearing. This geometry is advantageous in dealing with wind loading and can be explained in the following way.
Wind enters the enclosure slit from the left in Figure 4.5. The telescope tube responds by attempting to rotate clockwise about the altitude axis and by some bending of the fork tine (not shown). This causes encoder rotation which the servo immediately corrects by forcing the telescope to rotate counterclockwise. Another way to view the effect is to note that the dashed triangle PQR is constrained to rigid body rotation about the point P because the servo maintains the distance PR (R is the encoder contact point) constant.
The azimuth incremental encoder is similarly coupled via a much shorter effective tangent arm whose hinged end is at an isolated point on the telescope pier. The geometry is essentially identical to the drive mechanism in Figure 4.5 except that it may be scaled down in mass because of the negligible forces involved.
At the actual contact points the pickoff roller is preloaded against a virgin portion of the drive surface, not contacted by a drive roller.
Instead the telescope will sample at about 10 times this rate (10 pixels per arc sec, one dimension). By adjusting the bandpass to around 3000 A the daytime sky at f/10 about half fills a pixel (50,000 count) in 1/60 second, a standard video frame time.
For a signal to noise ratio of 10 the following table gives daytime limiting magnitude as a function of seeing diameter in arc sec and presumes data from all pixels inside the seeing disk can be effectively binned into asingle pixel.
Seeing Magnitude 1.0 7.0 2.0 6.2 3.0 5.8
Nearly all the signal is being used with a 3000 A bandpass centered at say 5000 A. If the seeing disk is always larger than one pixel, further gains can only be made by an increased detector dynamic range. This can be accomplished in effect by summing frames in memory but to be helpful may require more time than could be tolerated for a "transparent" operation. Since the the catalog coverage already provided by these magnitude limits is clearly satisfactory for offset pointing in daylight there is little need to pursue significant camera improvements.
At night the camera becomes essentially dark current limited and easily covers the entire SAO catalog with more than ample S/N.
We plan to implement at first light a background seeing data acquisition program that records image information for each offset star and all user program objects. No such program has ever been attempted with a large telescope with continuous nightly quantitative coverage over many seasons. We feel the contribution of these data to future seeing research should be substantial.
It is also intended that improvements to SAO positions be made to the on-line catalog also as a background task. It is not out of the question that some amount of background research can be done as well. Examples are photometric monitoring of bright stars and proper motion measurements with order of 0.1 arc sec precision.
The low heat dissipation of the CMOS controller will allow it to be mounted in the secondary assembly. All cabling between the controller, sensors, and actuators is thus confined to the on-axis secondary structure. Long distance cabling is minimized and consists of power and serial control commands.
The minimization of electronic noise produced by the telescope control hardware is essential. Clean (filtered and conditioned) power will be reserved for the instruments and the telescope control system processors. All medium level and power components will operate from raw power (protected from lightning) preventing noise coupling to the instruments through the power supplies.
Electromagnetic interference (EMI) generated by the telescope control system will be minimized at the source. We will use shielded enclosures and all medium and high power signals will run through shielded cable with the return in the same cable. When practical the driven device will be electrically isolated from the telescope preventing currents from returning through the telescope. Cable routing will be planned to minimize interference.
Small self contained weather monitoring stations are commercially available with standard computer interfaces built in. Weather data will be logged by TCC periodically during each night as a matter of general policy.
The bus protocol will be implemented on a fiber optic bus. This device, currently offered by Codenoll Inc. has sixteen fiber connections arranged in pairs, one a transmitter and one a receiver.
Signals received on any receiver fiber are retransmitted on all eight transmitter fibers. Communications thus proceed bi-directionally as on any bus but without any electrical contention. Moreover bus star connections may be made or broken while data are flowing without disturbing communications among other bus devices.
TBD unused fiber optic bus connections will be provided initially.