Building an Automated Telescope with High Photometric Accuracy

Transcription

1 PUBLICATIONS OF THE ASTRONOMICAL SOCIETY OF THE PACIFIC, 120: , 2008 September The Astronomical Society of the Pacific. All rights reserved. Printed in U.S.A. Building an Automated Telescope with High Photometric Accuracy ALBERT D. GRAUER 1 Lunar and Planetary Laboratory, University of Arizona, Tucson, AZ; algrauer@mac.com A. WILLIAM NEELY NF/ Observatory, Silver City, NM; neely@mail.nfo.edu AND CLAUD H. SANDBERG LACY Department of Physics, University of Arkansas, Fayetteville, AR; clacy@uark.edu Received 2008 May 29; accepted 2008 July 10; published 2008 August 20 ABSTRACT. The NFO WebScope is a Web-based observatory which has been in use since 2004 December. The telescope is an 0.6 m Group 128 Cassegrain reflector with a new drive and control system. By optical modifications and appropriate image treatment, we have been able to reach a differential photometric accuracy of about mag per observation in work on eclipsing binary stars. In addition, middle school, high school, college, graduate students, and public school teachers use this telescope to conduct their astronomical research projects under a NASA IDEAS grant. Online material: color figures 1. INTRODUCTION Wide-field-of-view reflecting telescopes are likely to suffer from varying responsivity across their field of view for a number of reasons. These variations will limit their photometric accuracy and may produce considerable systematic errors in photometric work. Light scattered and/or vignetted by the telescope, corrective optics, dewar window, and other optical elements from diffuse sources, nearby bright stars, planets, artificial sky-glow, and/ or the Moon may illuminate the detector in a way different from the point sources which are imaged by the telescope. These factors may compromise the effectiveness of conventional flatfielding procedures. We discuss a number of techniques for reducing photometric errors which were developed as a result of constructing the 0.6 m NFO WebScope and using it in work on eclipsing binary stars beginning in We find with modifications to the telescope structure and appropriate image calibration procedures, the standard errors for differential photometry may be reduced to approximately mag. The combination of the good weather in southern New Mexico (elevation of 6200 feet) and an efficient control system has allowed the observatory to be very effective. In the first 3 years of operation, the NFO WebScope has taken over 120,000 images for a total exposure time of more than 2700 hours. 1 Professor Emeritus, University of Arkansas at Little Rock; Affiliate, Arkansas Center for Space and Planetary Sciences; Mailing address: P.O. Box 2143, Silver City, NM NFO WEBSCOPE CONSORTIUM In 2002 March a consortium agreement for the construction of the NFO WebScope was signed by NF/ Observatory LTD (NFO) 2, the University of Arkansas at Fayetteville (UAF), the University of Arkansas at Little Rock (UALR), and the Arkansas School for Mathematics, Sciences, and the Arts (ASMSA). The Silver City School System of Silver City, New Mexico was added to the group. The purpose of this agreement was to enable the construction and operation of a robotic telescope which students could use to conduct research projects via the Internet. 3. TELESCOPE, DOME, AND CONTROL SYSTEM The telescope is a Group 128, 24 inch f=10 classical Cassegrain and dates from the late 1960s. The mirror was refigured in 1989 and was recoated in A thin silica overcoat was added for durability. The original drive system was designed for primitive stepper motors with two worm gears in series. The combination of this design and the lack of pre-loading made the telescope unusable for electronic imaging and computer control. The secondary worm gear was removed and replaced with a micro-stepping motor (Fig. 1). The resolution of this motor is approximately 2 NF/ is the cattle brand for the Neely-Frasca ranch, which is where the original observatory is located. 992

2 AUTOMATED TELESCOPE WITH HIGH PHOTOMETRIC ACCURACY FIG. 1. Micro-stepper motor and spur gears on the decl. worm provide precise computer control of the telescope. A similar arrangement is used to control the R.A. axis of the telescope. See the electronic edition of the PASP for a color version of this figure. 0.1 arcsec per step at the telescope axis. An anti-lash preloading system was added in both R.A. and decl. (Fig. 2). To correct for periodic error in the R.A. gear, an encoder was added to the R.A. worm shaft. Tracking corrections applied in the telescope control software eliminated the 12 amplitude periodic worm error (Fig. 3). The Group 128 German equatorial mount (GEM) is very stout (700 Kg) and employs 10 inch bearings in both R.A. and decl. The slewing errors are very reproducible. At the University of Arkansas at Fayetteville (URSA telescope, [Lacy et al. 2005]) a formula was derived to correct for mount error coupled with pole alignment error and refraction. This technique was applied to the NFO WebScope control software and FIG. 2. Anti-lash pre-loading weights on the R.A. drive eliminate erratic telescope motions. The decl. worm has a torque motor and a pulley to achieve the same result for this axis. See the electronic edition of the PASP for a color version of this figure. 993 FIG. 3. Inherent R.A. worm periodic error is shown in this figure. These data and an encoder were used to correct tracking to better than 1. has reduced its slew error to less than 1.5. The same formula is used to reduce the telescope tracking error over time. The telescope is operated unguided and, yet, produces round star images for exposures up to 10 minutes in length. The secondary mirror is moved with a 1=4-20 screw and stepper motor combination to achieve precise focusing. During the telescope break-in period a scheme for automatic focusing was developed and data were collected over a wide range of temperatures. A guide star between eighth and tenth magnitude was selected and 10 images of it were taken at different focal settings. The focal setting with the minimum FWHM (full width at half-maximum) and temperature were recorded. A consistent relationship between the ambient temperature and the focal position was derived. This relationship allows the telescope to be refocused before each image is taken based on the ambient temperature. At the beginning of each night a focus run is done on a guide star and these observations are added to the temperature versus focal position data base. A regression analysis is performed periodically to optimize the focusing algorithm. These procedures produce sharp images with a minimal use of on-the-sky time. The NFO WebScope is the second automated telescope produced by the NF/ Observatory LTD. The first instrument (Neely and Treasure 1989) used a rotating dome which was responsible for most of the operational problems. The NFO WebScope enclosure is a roll-off commercial corrugated steel structure (Fig. 4). The concept and initial roll-off components were designed by Dave Sperling, a mining engineer. The new enclosure has proved to be robust. Although, a traditional dome offers more protection, the NFO WebScope is able to operate in winds up to 5 mph with gusts of less than 12 mph. The telescope is unable to operate 10 to 15% of the time, during clear weather, due to excessive wind. Since the observatory runs 11 months a year, independently of an online operator, several fail-safe

3 994 GRAUER, NEELY, & LACY software pipeline. The fourth computer communicates with the Internet, controls the target selection, communicates with the DOS tracking computer, occasionally communicates with an on site astronomer, and runs the status monitoring webcams. All of the computers, except the tracking computer, run Linux. Processing software includes C++, C, PHP, and MySQL. The telescope uses a custom version of ATIS (automatic telescope instruction set) for target selection and instrument control. It is a modification of ATIS-2000 maintained by Don Epand. 3 ATIS was written in the 1980s by Boyd and Epand for photoelectric photometric telescopes and, later, upgraded for CCD use (Boyd et al. 1993). ATIS is intended to allow students from the fifth grade to the University level to use the NFO WebScope for imaging. The Web student interface with ATIS will be discussed later in this paper. FIG. 4. NFO WebScope telescope with the roll-off enclosure partially open. See the electronic edition of the PASP for a color version of this figure. sensor and software systems are employed to prevent damage to the robotic instrument. A commercial Davis weather station monitors the conditions while site-built sensors are used to detect lightning, rain, snow, and/or clouds. All the weather and telescope performance data are archived and retrievable in various formats. A 3 s software status loop checks for dangerous weather conditions or other problems and closes the dome if necessary. A dead-man switch will also close the dome in the event of a computer failure. The observatory is controlled by four primary computers. A single high performance machine could have been used. However, the use of multiple mass produced units is cost effective, enables accurate control of timing functions, and provides for relatively straight forward software and hardware maintenance. The telescope pointing and motions are produced by the micro-stepping motors which are controlled by a real-time tracking computer running DOS with software written in C. The second computer controls the dome and various weather sensors. The third computer runs the camera, shutter, filter wheel, and focus motor. The third computer, also, does the dark and flat-field image calibrations and uses WCSTools to calculate the astrometric position of the center of each frame (Greisen and Calabretta 2002; Calabretta and Greisen 2002). This processing takes about 3 s per image. The WCS calibration is successful about 90% of the time but can fail in a very crowded or sparse star field (e.g., in the Milky Way or for a short exposure time). Non WCS-corrected images still proceed through the 4. INSTRUMENT PACKAGE The instrument package was designed by Alan Uomoto and constructed in the UALR GIT (Graduate Institute of Technology) machine shop. It is located at the Cassegrain focus and utilizes a pixel Kodak KAF-4301E CCD chip in a Finger Lakes Instrumentation 4301E camera. It has 0.8 pixels and a 28 field of view. B, V, and R Bessel prescription filters from Custom Scientific and neutral density filters for bright objects are located in a filter wheel. The camera shutter employes a micro-stepper motor, encoder, and a rotating blade to achieve precisely timed exposures. Kodak chips are known to produce ghost images of bright, non-saturated objects, which appear on subsequent images. Our chip is no exception. These spurious images fade on each successive dark image which is read out. Since these artifacts can effect the photometry, our procedure is to read out a dark image of the chip every 2 s when the camera is not making an exposure. Thus, the chip is erased at least 4 times between science exposures. This technique is usually sufficient to remove the effect of bright sources on subsequent frames. The telescope is a classical Cassegrain with a parabolic primary and a hyperbolic secondary. A system of two concave-skyward anti-reflection coated meniscus lenses made of fused silica are employed to produce adequate (2 ) images over the entire CCD. Seeing conditions at the site are probably on the order of 1 with the limit being set by the quality of the primary and secondary mirrors. Figure 5 shows the instrument package attached to the telescope. The camera cooler is programmed to keep the detector at 20 C, except during June when it is set to 15 C. The variations in the temperature of the CCD are regulated to be less than 0.5 C during the night. A 23 s readout time for the CCD is used to provide high photometric accuracy with a minimum of dead time. 3 For information on ATIS, see

4 AUTOMATED TELESCOPE WITH HIGH PHOTOMETRIC ACCURACY 995 preceding the evening run. These images are used to produce nightly median filtered bias, dark, and flat frames. Each day the median filtered bias, dark, and flat frame from the previous night is compared with the respective rolling average of the 10 previously accepted nightly bias, dark, and flat frames. If the average pixel value in the most recent nightly median filtered bias, dark, or flat frame varies by less than 5% from the average of the 10 most recent accepted nightly bias, dark, and flat frames, the new nightly frame is accepted and replaces the oldest of the 10 frames in the rolling average. Our Kodak CCD chip is cosmetically excellent. Dead and/or variable low sensitivity pixels are not a problem since they represent an insignificant fraction of the chip s light gathering surface. FIG. 5. NFO WebScope instrument package. See the electronic edition of the PASP for a color version of this figure. 5. PRELIMINARY PROCESSING OF IMAGES All CCD frames must undergo a process of calibration to obtain an image from which scientific measurements can be made. CCDs have a level of voltage in the readout amplifier, or bias, which must be subtracted. In addition, during the exposure each pixel in the image field gradually builds up electrons, even in the absence of light. This is called dark current and must be subtracted. Finally, each pixel has a slightly different sensitivity and is affected by inhomogeneities in the optical system (for instance dust over a specific area of one of the filters). Flat-field images are taken of a uniformly illuminated field, and a sensitivity correction factor is calculated for each pixel. Dome flats obtained by imaging a uniformly illuminated screen, twilight flats obtained by imaging the sky at twilight, and all sky flats obtained by median combining images of the night sky are the commonly used methods for obtaining the flat-field correction. Bias and dark frames are taken each morning after the night s observing. Twilight flats are used at the NFO WebScope and are constructed from images taken at twilight immediately 5.1. Bias Subtraction Bias images are constructed from 10 images each taken with a 0 s integration time and the same 23 s read-out time used to obtain the science frames. The value of each pixel is ranked relative to its value on the other nine frames. The top three values of this pixel are thrown out. The middle value of the remaining seven is kept as the value for the bias of this pixel in the new nightly median filtered bias frame. This image is then compared with the average of the 10 last median filtered nightly bias frames which have been accepted. If the median pixel value of nightly bias frame is within 5% of the average of the last 10 good nightly bias frames, it is accepted, and averaged in the rolling average which is then used to calibrate the images. This procedure compensates for cosmic rays, variable hot pixels, and the slow drift in the bias due to seasonal temperature variations Dark Correction and Flat-Fielding Ten raw dark frames are taken with a 120 s integration time for each frame. The bias rolling average frame is subtracted from each new raw dark frame. The resulting set of 10 new bias subtracted dark frames are processed like the bias frames, therefore, the fourth lowest value is adopted for each pixel. The resulting new nightly median filtered dark frame is compared with previous ones. If the median pixel value of the nightly dark frame is within 5% of the average of the last 10 accepted nightly dark frames, it is incorporated into the rolling average of the last 10 accepted nightly 120 s median filtered dark frames. The 120 s rolling average dark frame is scaled by integration time and subtracted from each image. This procedure compensates for cosmic rays, variable hot pixels, and the slow drift in the dark values due to seasonal temperature variations. Flat-field images are taken in the evening twilight, in each filter, with the telescope pointed at the zenith. They are obtained with the telescope in proper focus, as determined by the ambient temperature, with the optical tube on the east side of the mount. The exposure time for each frame is selected so the CCD is at the midpoint of its linear range to achieve a high signal-to-noise

5 996 GRAUER, NEELY, & LACY ratio. Twilight flat images can be problematic, because timing is very important. If the images are taken too early, the sky brightness can move the detector out of its linear range. If they are taken too late, background stars illuminate the CCD in a nonuniform manner. There is a very small window in time for these exposures. The capture of NFO WebScope twilight flat frames is carefully timed by the Sun s angle below the horizon. An integration time of 2 s is used to obtain 10,000 ADU with the B, V, R, and N flats being started when the Sun is 2 and 4 below the horizon, respectively. Filters are sequenced in proportion to the ambient light and taken in the order: B, V, R, and N (no filter). For each filter a series of four twilight flats are taken. The four images are then individually normalized by multiplying by the appropriate factor so the average value of a flat-field pixel over the whole frame equals 10,000 ADU. This procedure allows accurate comparisons of pixel values between the frames and speeds up later image processing by allowing integer arithmetic to be employed. For each pixel location, the top value of the four flat frames is discarded and the median value of the remaining three pixels is used as the flat pixel value. This procedure minimizes the effects of cosmic rays and variable hot pixels. These median pixel values are used to make the nightly flat-field image. The new nightly flat for each filter is then compared to the 10 previously accepted flats for that filter. This step is taken to make sure that poor sky conditions or clouds did not influence the calculation. If the median pixel value of the new nightly flat frame is within 5% of the average of the last 10 good nightly flat frames, it is averaged into the last 10 accepted nightly flat frames in the rolling average and used to calibrate the images. Our twilight sky flats were checked by comparing them with all sky flats. During a clear moonless night, a series of long exposure images were taken in a relatively star-poor part of the sky. Any particular pixel will be illuminated by the sky background and not a star on most of the images. Therefore, a true flat can be obtained from these observations. The median pixel value of the sky on all of the frames was used to calculate a median filtered all sky flat. The use of all sky flats is appealing because they are taken with the same spectral energy background distribution as is found in the data frames, and they may not suffer from gradients which can occur in the twilight sky. However, unless the frames used to calculate an all sky flat can be obtained during the regular science program, this method may not make the best use of the telescope time. The twilight sky flats and all sky flats methods produced flats which differed by less than 0.1% from each other. This comparison eliminates the possibility that the polarization of the twilight sky and/or the position of the optical tube assembly relative to the R.A. axis has a substantial effect on the determination of an accurate flat. Because twilight sky flats can be completed before the evening run, they are routinely used to correct the NFO WebScope data. NFO WebScope users have several choices for processing their images: 1.The user can obtain raw images and use the night s bias, dark, and flats to calibrate their images. 2.The user can have the processing done automatically using the available bias, dark, and flats. 3.The user can have the processing done automatically, plus a final photometric correction. Almost all of the users images have been processed with the third option, for reasons that will soon become apparent Indications of Trouble One of the primary objects of study with the NFO WebScope is eclipsing binary stars. In this kind of work there are, generally, three stars in the images that are measured. They are the variable star (Var), the comparison star (Comp), and the check star (Ck). These stars are typically separated in position by only several arcminutes. We initially assumed, because they were close together in the sky, that our calibration procedures would ensure that the differential magnitudes of these stars would be independent of their positions on the chip. Images are measured by using an application, NFO Measure, written for Macintosh computers especially for the study of variable stars (see for example, Lacy et al. 2005; Lacy et al. 2004). A pattern file is created for each eclipsing binary in which the Var, Comp, and Ck stars are identified. This file is used to match the pattern wherever it falls in the image. The NFO WebScope has a GEM. When the object being studied crosses the meridian, the telescope must execute a maneuver called the GEM-flip in which the body of the telescope moves from the west side of the R.A. axis to the east side. This maneuver rotates the image field by 180. The NFO Measure software compensates for the GEM-flip by rotating the image, if necessary, so that north is up and east is to the left in the working copy of the image. Therefore, the same pattern can be used on both sides of the R.A. axis. A problem showed up initially, in differential photometric measurements. A shift is seen in the Var Comp and the Comp Ck data immediately following the time of GEM-flip (Fig. 6). In addition, the field center after GEM-flip was observed to be shifted by several arcminutes. The size of the photometric discontinuity in the Comp Ck and Var Comp values varies with the geometry of the star pattern in the images Attempts to Correct the Problems Once problems were identified around the GEM-flip, a search for solutions began. Initially, the problem was thought to be related to skylight hitting the CCD directly. This situation could result from insufficient telescope baffling, which would allow rays to pass around the edge of the secondary to the CCD. In any event, the light used to produce the flat field must

6 AUTOMATED TELESCOPE WITH HIGH PHOTOMETRIC ACCURACY 997 ondary mirror cell and the baffle tube was extended (Fig. 7). The masks cut down the usable light from the mirror surfaces by 5%, but they were only able to reduce mean photometric error to between 0.03 and 0.05 mag. Our calibration procedures were insufficient to correct for this unknown source of error Photometric Flat Fielding FIG. 6. Photometric shift in the Comp Ck and Var Comp magnitudes for AQ Ser is seen just after the time of GEM-flip. This jump is due to the fact that the image is rotated 180 on the CCD after the flip, and there are photometric response variations across the image from the way in which the flat-fielding was done. be imaged by the optics exactly like the point sources which are to be measured photometrically. Our calibrated night sky images had a very uniform sky background. We had confirmed the flat fielding using two methods (all sky flats and twilight flats). However, the same stray light would even out the background during a starfield exposure as during a flat-fielding exposure. The telescope already had baffles and attention had been paid to reducing internal reflections, using low-reflection paints and black screen netting glued to the inside of the baffle tubes to roughen the surface. To further reduce any stray skylight, a mask was added to the sec- FIG. 7. Masking to reduce direct skylight from hitting the CCD. A is the secondary mask. B is the extension to the baffle tube. See the electronic edition of the PASP for a color version of this figure. We, eventually, came to the conclusion that both the twilight sky and the night sky, which we were using to produce our flatfield images, could be illuminating the chip differently from the point sources being imaged by our telescope. Selman (2004) measured the response (photometric zero point) variations over the detector used in the Wide Field Imager at the MPG/ESO 2.2 m telescope at La Silla by using dithered exposures of dense star fields during clear nights. We have adopted his methods with good results. We observe open star clusters with a cross-shaped dither pattern. A central exposure and north, west, south, and east dithered exposures each with shifts of 5 are taken in rapid succession (our images are about 27 square). This 5-image sequence is repeated 6 times in succession over a period of about an hour, with the telescope on the same side of the R.A. axis for all exposures. Variations due to air mass differences are removed by using the mean extinction coefficient in the V filter (0:16 mag per airmass ). The V filter is used for all exposures. The Selman (2004) method uses a least-squares algorithm with singular-value FIG. 8. Responsivity variations across the image due to the optics as measured from dithered images of M52. Near the center of the twilight sky flat-fielded image, the response is less than near the edge of the field. The contours are at the 1=2% level.

7 998 GRAUER, NEELY, & LACY the latter reverses this correction by 3 to 5%. It is possible vignetting effects diffuse light and light being imaged from point sources differently, and the photometric flat actually corrects for this effect. Further, Selman (2004) and Anderson et al. (1995) suggest that gradients in the diffuse source, scattered light in the telescope and instrument, and sky concentration produced by the focal reducer are additional possible sources of error introduced in the traditional flat-fielding process NFO WebScope Scientific Results FIG. 9. Same data as presented in Figure 6 after the photometric flat-field correction has been applied. The photometric shift seen just after the time of GEM-flip is gone. decomposition to model the photometric zero-point variations across the field of view as a weighted sum of two-dimensional Cartesian Chebyshev basis functions (see his article for details). A typical fitted result is shown in Figure 8. Based on the fitted response function, we calculate a photometric flat image normalized to a value of 10,000 ADU. We use this photometric flat during initial reductions of NFO WebScope images before distribution to the investigators. Because the flat-field images are normalized to 10,000 ADU, each pixel location can be simply multiplied by the photometric flat and then divided by the twilight flat. After the images were corrected in this way, the photometric measurements showed dramatic improvement (Fig. 9). A carefully calculated flat-field image from an illuminated screen in the dome, the twilight sky, or the night sky does not guarantee a photometrically flat image. It should be emphasized that an image with a cosmetically flat sky may not produce accurate photometric results. To further test this hypothesis, data were taken on L107 (Landolt 1992) with the Steward Observatory 60 inch Telescope on Mt. Lemmon, Arizona. These images have one arcsecond pixels, a one degree field of view, and were carefully calibrated using an all sky flat obtained from hundreds of images of the night sky. These calibrated frames have a very uniform sky background (well under a 1% variation), and, yet, a constant star is found to vary by more than 0.15 mag as it is moved to different pixel locations on an image. The differences between a flat made using diffuse light and one made with light from point sources probably result from a combination of physical processes. The twilight sky flat and the photometric flat (Fig. 8) show similar features indicative of vignetting. The former corrects for a decline in sensitivity of approximately 25% from the center to the edge of the chip while The first scientific results from the NFO WebScope are now appearing in the astronomical literature. Times of minima of eclipsing binaries have been reported (Lacy 2006; Lacy 2007). NFO WebScope observations were used to rule out Delta Scuti-type pulsations in the eclipsing binary star EY Cep (Lacy et al. 2006). Figure 10 compares data obtained from three different observatories on the eclipsing binary star GX Gem. The URSA data were obtained at the Kimpel Observatory, which is described briefly here since this information does not appear elsewhere in the literature. This facility consists of a Meade 10 inch (25 cm) f=6:3 LX-200 telescope with a Santa Barbara Instruments Group ST8 CCD camera (binned 2 by 2 to produce 765 by 510 pixel images). The telescope is inside a Technical Innovations Robo-Dome and is controlled automatically by an Apple Macintosh G4 computer. The observatory is located on top of Kimpel Hall on the University of Arkansas Fayetteville campus. The control room is located on the floor below. This facility has been used to produce data on other eclipsing binary FIG. 10. Light curve of the primary eclipse of GX Gem from corrected URSA (squares) and NFO (plus signs) observations, and from the observations of SB (triangles) (Sánchez-Bajo et al. 2003). The O-C are the differences between the observations and the calculated eclipsing binary light curve. The NFO WebScope observations are seen to be of very high photometric accuracy.

8 AUTOMATED TELESCOPE WITH HIGH PHOTOMETRIC ACCURACY 999 stars (Lacy et al. 2005; Lacy et al. 2004). The NFO data were obtained with the NFO WebScope. The SB data were provided to us by F. Sánchez-Bajo, and discussed in the paper (Sánchez- Bajo et al. 2003). The top panel plots the light curve, and the bottom panel plots the difference between the observations and the model light curve around the primary eclipse (see Lacy et al. 2005; Lacy et al for model details). The residual standard error for the NFO results is less than mag. To achieve this accuracy, nightly corrections had to be applied, generally less than 0.01 mag, to achieve a common photometric zero point. 6. WEB INTERFACE The name NFO WebScope infers the importance of the World Wide Web to the operation of this facility. The Internet is used to: 1.Provide an astronomer interface. It enables users to input a target list and retrieve their images. The images are taken in New Mexico and stored in Arkansas. 2.Access the archive of images. A computer in Arkansas has all of the NFO WebScope images online, nightly weather data, and indexed thumbnail JPEG images of the FITS images. All the images are cross-referenced in a database. 3.Access image manipulation tools. Point-and-click photometry and image co-add software can be used with any of the images. 4.Access learning modules. The NFO WebScope is, normally, not a real-time remotely operated telescope. Image requests are done with interactive software at the UALR computer, which in turn sends an ATIS to the Observatory daily, in the late afternoon. Users are notified by when their images have been taken. Some of the users have their own local server accounts where their images are relayed from UALR automatically, each morning. The calibrated 8 Mbyte images are produced in FITS format at the observatory. An enhanced JPEG copy and thumbnail of each image, along with an HTML copy of its header are made. Then, the images are compressed with hcompress, a FITS compression routine written by Richard L. White for use at the Space Telescope Science Institute. 5 We use a four-to-one compression which maintains a high data quality. The compressed FITS image, plus the JPEGs, and HTML file are sent along with an index of the night s thumbnail images to UALR in Arkansas via a website. 6 At this main repository, the images are entered into a database with multiple cross indexes. All 120,000+ images are online along with 30,000 images taken by the first NFO telescope ( ). Users can look at a particular 5 For information on the hcompress software, see the website 6 The Web site for submitting images is night s operation with attached cloud data, or access any of the images in the data base with a variety of search options. The weather archive will supply complete weather information on a particular night if needed. There is also a link 7 to obtain current weather information, webcam views of the sky and telescope (daytime), and telescope telemetry. Using a combination of Sextractor (Bertin and Arnouts 1996) and IRAF 4 software and a GUI interface built with the PHP programming language, astronomers are able to perform online image analysis of any of the images. Several of the learning exercises described in the following section utilize these tools. 7. EDUCATIONAL ACTIVITIES In addition to the research opportunities afforded by observing sessions for college and graduate students, the NFO Web- Scope is engaged in a significant K 12 outreach. We are in the second year of NASA IDEAS Grant Number ID05-691, which was awarded to the University of Arkansas at Little Rock. 8 In the NASA IDEAS program, students complete interactive instructional units followed by short quizzes using a normal Web browser. Students must repeat quizzes until they achieve the score needed to obtain privileges on the telescope. The ATIS entries required to make the observations are made via a pointand-click interface (which is much simpler than the more sophisticated text entry interface). The program is self-paced. Teacher interaction is encouraged, but is minimal in the majority of the exercises. Topics with relevant image requests include: general astronomy, the Moon and orbits, the scientific method and the planets, the atmosphere, variable stars, and near Earth objects (NEOs). There are online resources on the UALR computer to perform photometry, co-add images to make true color images from filtered images, and make movies of NEO s motion in the sky. 9 Photometry measurements are used to help the student learn graphing skills. Most of the units are completed by greater than 90% of students at the eighth grade level. Over 1400 K 12 students, from three states, have used the NFO WebScope to take images. Complete weather observations for the past several years are on our website and employed by science teachers and their students. 7 Weather information, webcam views, and telescope telemetry is available at 4 IRAF is written and supported by the IRAF programming group at the National Optical Astronomy Observatories (NOAO) in Tucson, Arizona. NOAO is operated by the Association of Universities for Research in Astronomy (AURA), Inc. under cooperative agreement with the National Science Foundation. 8 For more information, see StatsAbs.shtml. 9 For more information, see

9 1000 GRAUER, NEELY, & LACY 8. CONCLUSIONS It is possible, through careful attention to optical design and data reduction techniques, to produce high quality photometric data from a completely robotic telescope. Twilight or the night sky illuminates the CCD in a manner measurably different from the point sources imaged by the telescope. To obtain the highest quality photometric measurements, a correction must be applied to the flat-field image produced from these diffuse sources. A carefully crafted Web interface can be used by K 12 students to do astronomical projects with a robotic telescope, in concert with research observations. The authors wish to thank Ann Grauer for suggesting that we move the telescope to New Mexico and Alan Uomoto for his design work on the instrument package. Joel E. Anderson, Shannon R. Clardy, Steven A. Crawshaw, Rogers E. Davis, Albert E. Everett, R. Ben Gilbert, M. Keith Hudson, Lon Jones, Mike McCallister, Jimmy McGuire, Linda Musun, Dever Norman, Sandra L. Robertson, André Rollefson, Jeff Shaw, and Armand J. Tomany provided support to this project at UALR. Ed Beshore, Arlo Landolt, Steve Larson, and the referee provided suggestions for improvement. Ann Grauer is responsible for the final edit. We are grateful to the University of Arkansas at Little Rock, University of Arkansas at Fayetteville, AAS Small Grants Program, Arkansas Space Grant Consortium, and the UALR GIT machine shop for their contributions to this project. Internet services are provided by UALR and The Silver City School District. This work is supported in part by NASA IDEAS Grant Number ID05-691, a NASA grant from the Near Earth Objects Observation Program, and a NSF CCLI grant. REFERENCES Anderson, M. I., Freyhammer, L., & Storm, J. 1995, in ESO Conference and Workshop Proceeding, ESO/ST-ECF Workshop on Calibrating and Understanding HST and ESO Instruments, ed. P. Benvenuti, 87 Bertin, E., & Arnouts, S. 1996, A&AS., 117, 393 Boyd, L. J., et al. 1993, IAPPP, 52, 23 Calabretta, M. R., & Greisen, E. W., 2002, A&A, 395, 1077 Greisen, E. W., & Calabretta, M. R. 2002, A&A, 395, 1061 Lacy, C. H. S. 2006, IBVS, 5670, 1 Lacy, C. H. S. 2007, IBVS, 5764, 1 Lacy, C. H. S., Claret, A., Sabby, J. A., Hood, B., & Secosan, F. 2004, AJ, 128, 3005 Lacy, C. H. S., Torres, G., Claret, A., & Menke, J. L. 2006, AJ, 131, 2664 Lacy, C. H. S., Torres, G., Claret, A., & Vaz, L. P. R.2005, AJ, 130, 2838 Landolt, A. U. 1992, AJ, 104, 340 Neely, B., & Treasure, F. 1989, Remote-access Automatic Telescopes, ed. D. Hayes & R. Genet (Arizona: Fairborn Press) Sánchez-Bajo, F., et al. 2003, Astron. Nachr., 324, 511 Selman, F. J. 2004, Proc. SPIE,