7.1 Introduction

In this chapter, we assess the instrumental requirements (telescopes, mosaics of CCD chips, computers, etc.) imposed by the observing strategy and follow-up research outlined in Chapters 5 and 6, and we comment on observational techniques and observing network operation. In order to cover the requisite volume of search space, the survey must achieve a stellar magnitude limit of at least V = 22, dictating telescopes of 2-3 m aperture equipped with CCD detectors. The most efficient use of CCD detectors is achieved if the pixel size is matched to the apparent stellar image size of about 1 arcsec, thus defining the effective focal length for the telescopes at about 5 m. The area of sky to be searched is about 6,000 square degrees per month, centered on opposition, and extending to +/- 30 deg in celestial longitude and +/- 60 deg celestial latitude. These considerations lead us to a requirement for multiple telescopes with moderately wide fields of view (at least 2 degrees) and mosaics of large-format CCD detectors. We develop these ideas in this chapter to derive a proposed search program. This program is not unique (that is, an equivalent result could be obtained with other appropriate choices of telescope optics, focal-plane detectors, and locations), but it is representative of the type of international network required to carry out our proposed survey.

7.2 Lessons From the Spacewatch Program

The Spacewatch Telescope, operated at the University of Arizona (see Chapters 3 and 4), is the first telescope and digital detector system devised to carry out a semi-automated search for NEOs. As such, the lessons learned from its development and operation are invaluable when considering a future generation of scanning instruments. The Spacewatch system comprises a single 2048x2048-pixel CCD chip at the f/5 Newtonian focus of an equatorially mounted 0.91-m telescope. Each pixel covers 1.2 x .2 arcsec on the sky. With the telescope drive turned off, the camera scans the sky at the sidereal rate, and achieves detection of celestial bodies to a limiting magnitude V = 20.5.

One of the important demonstrations provided by the Spacewatch Telescope team is that image-recognition algorithms such as their Moving Object Detection Program (MODP) are successful in making near-real-time discoveries of moving objects (asteroids and comets). False detections are almost eliminated by comparing images from three scans obtained one after the other. At present, the Spacewatch system makes detections by virtue of the signal present in individual pixels. With the incorporation of higher-speed computers, near-real time comparison of individual pixels to measure actual image profiles would lead to a great reduction in the most frequent sources of noise, cosmic ray hits and spurious electrical noise events.

In light of the successful performance of Spacewatch, we have rejected a photographic survey. Even though sufficiently deep exposures and rapid areal coverage could be attained to fulfill the survey requirements using a small number of meter-class Schmidt telescopes (similar to the Oschin and U.K. Schmidts), there is no feasible way, either by visual inspection or digitization of the films, to identify and measure the images in step with the search. A photographic survey would fail for lack of adequate data reduction and follow-up. Future developments in electronics and data processing will further enhance the advantages of digital searches over the older analog methods using photography.

7.3 Detector and Telescope Systems

The largest CCD chips readily available today contain 2048x2048 pixels, each about 25 micrometers on a side. Thus, the chips are about 5x5 cm in size. Quantum efficiencies have attained a peak near 80 percent, and useful sensitivity is achievable from the near-ultraviolet to the near-infrared. To reach a limiting stellar magnitude of V = 22, we require the use of these CCDs at the focal plane of a telescope with an aperture of 2 m or larger, operated during the half of the month when no bright moonlight is present in the sky (from last quarter to first quarter phase).

In the coming decade, we envisage a trend toward smaller and more numerous CCD pixels covering the same maximum chip area as at present. No great increase in spectral sensitivity can be expected. At the telescope, the pixel scale must be matched to the image scale (the apparent angular size of a stellar image) in good or adequate atmospheric (seeing) conditions. In what follows, we assume a pixel scale of 1 arcsec/pixel (25-micrometer/arcsec, or 40 arcsec/mm), which implies a telescope of 5.2-m focal length. For a telescope of 2 m aperture, the focal ratio is f/2.6; for a 2.5-m, f/2.1; and for a 3-m, f/1.7.

A single 2048x2048 CCD chip simultaneously detects the signals from more than 4 million individual pixels. This is a very powerful data-gathering device, but it still falls short of the requirements for wide-field scanning imposed by the proposed NEO survey. At the prime focus of a telescope of 5.2 m focal length, such a CCD covers a field of view on the sky about one half degree on a side. However, we wish to scan an area at least 2 degrees across. Therefore, we require that several CCD chips be mounted together (mosaicked) in the focal plane. The mosaicking of CCD chips is not a simple process, but it is one that is being vigorously pursued today by astronomers. At Princeton University, for example, a focal plane with 32 CCDs is under development. Mosaicking of 4 to 10 CCD chips into a single focal plane should not be a problem for the proposed survey telescopes by the time they are ready to receive their detector systems.

Studies and planning are underway at the University of Arizona for a modern 1.8-m Spacewatch telescope. The new telescope will be an excellent instrument to test and develop some of the necessary instrumental and strategic considerations outlined in this report. From the Spacewatch design considerations, it is safe to assume that 2- to 3-m-class telescopes can be built having focal lengths near 5 m and usable fields of view between 2 and 3 deg. Refractive-optics field correction is probably required, and it appears advantageous to locate CCD mosaics at the prime focus of such instruments. Here, we indicate telescope functional requirements but do not exactly specify the size or design of the proposed survey telescopes.

7.4 Magnitude Limit and Observing Time

Exceptionally fine astronomical sites have more than 1,000 hr/yr of clear, moonless observing conditions, during most of which good to adequate seeing prevails. More typically, 700 hr/yr of observing time is usable. We assume that a region of 6,000 square degrees is to be searched each month and that initial NEO detection is made by two or three scans on the first night. Parallactic information is derived by four scans on a subsequent night, and an orbit is calculated from observations on a third night. Thus, nine or more scans of the search region are needed each month. In a given month, follow-up will be attempted for some of the NEOs that have moved out of the search region (mainly to the west). As a working value, we assume that 40 hr/month/telescope are available for searching.

The limiting (faintest) stellar magnitude that can be observed by a telescope can be determined as a function of the ratio of the source brightness to that of the sky, the number of pixels occupied by a star image, the pixel area, the light-collecting area of the telescope, and the effective integration time (Rabinowitz 1991). For certain detection, the source brightness must be at least six times that of the sky noise. We have normalized to the performance of the Spacewatch Telescope, which achieves a stellar limit of V = 20.5 using an unfiltered 165-s scan at sidereal rate, and we have allowed for an improvement over the performance of that system arising from improved detector quantum efficiency and improved image-recognition algorithms. We find for the survey telescopes that a single CCD should be able to achieve the survey requirement of V = 22 with the following combinations of telescope aperture and scan speed:

7.5 Number of CCD Chips and Telescopes Required

A single 2048x2048-pixel CCD chip, having an image scale of 1 arcsec/pixel, can scan at0.14 square degrees per minute at the sidereal rate. If 40 hr/month/telescope can be allotted to searching for NEOs over 6,000 square degrees to a limiting stellar magnitude of V = 22, and ten scans per sky region are required for detection and rough orbital characterization of an NEO, then telescopes of the apertures considered above have the following performance capabilities:

Primary Diameter (m)	Exposure Time (s)	Scan Rate (x sidereal)
2.0	260	28
2.5	420	18
3.0	500	13

In computing values for the total number of CCD chips required in the worldwide network of telescopes we assume that no two CCD chips together scan the same region of the sky. These are minimum requirements for the telescopes; in practice more scans may be needed for reliable automatic detection, and probably there will be some overlap of coverage between telescopes.

Searching to +/- 60 deg celestial latitude implies sky coverage, over the course of a year, at almost all declinations. Thus telescopes must be located in both hemispheres. Usable fields of view of between 2 and 3 deg probably limit the number of CCD chips in a telescope's focal plane to about ten at the scales we have been considering. However, real-time image processing is simplified if each chip independently samples the sky. Most likely, four CCDs chips/telescope can be accommodated in a linear array in the focal plane. Thus, it appears that seven 2.0-m telescopes, five 2.5-m telescopes, or four 3-m telescopes suffice to fulfill the search, follow-up, and physical observations requirements of the idealized 6,000-square degree survey. Most likely, there would remain extra observational capability to enhance the detection rates of Atens and LPCs by scanning a few times per month outside the standard region. We note that each telescope must be equipped with a minimum of four 2048x2048 CCD chips or their equivalent in light-collecting ability. If space remains in the focal plane, additional filtered CCD chips could be inserted to undertake colorimetry, which would give a first-order compositional characterization of some of the NEOs discovered while scanning.

If a single-point failure due to weather or other adverse factors is not to hamper effective operation of the survey network, we conclude that three telescopes are required in each hemisphere. With fewer telescopes, orbital, and perhaps parallactic, information on NEOs would be sacrificed. The desirability of searching near the celestial poles calls for at least one telescope at moderate latitude in each hemisphere. In summary, we propose a network of six 2-m or larger telescopes distributed in longitude and at various latitudes between, say, 20 deg and 40 deg north and south of the equator.

7.6 Scanning Regime

At high declinations, scanning along small circles of declination results in curvature in the plane of the CCD chip, so star images do not trail along a single row of pixels. The problem can be avoided by scanning along a great circle. A good strategy would be to scan in great circles of which the ecliptic is a meridian (the pole being located on the ecliptic 90 deg from the Sun). Such scanning can be achieved using either equatorial or altazimuth telescope mounts, but is probably more easily and cheaply accomplished using an altazimuth mount. In either case, field rotation is required, as is currently routinely used at the Multiple-Mirror Telescope in Arizona and other installations.

At the proposed 1.8-m Spacewatch telescope, it is planned to make three scans of each region of the sky (as is currently done at the 0.9-m Spacewatch telescope). Each scan would cover 10 deg in 26 min, so the interval between the first and third scans is sufficiently long that objects moving as slowly as 1 arcmin/day can be detected. For the proposed NEO survey, we envisage two or three longitudinal scans per sky region, about an hour apart. Thus, at a scan rate of 10 times sidereal, each scan could cover an entire strip of the 60-deg-wide search region, with a second search strip being interposed before the first was repeated. We assume that false positive detections, being somewhat rare, will not survive scrutiny on the second night of observation, and thus will not significantly corrupt the detection database.

7.7 Computer and Communications Requirements

Near real-time detection of faint NEOs requires that prodigious amounts of data processing be accomplished at the telescope. The image processing rate scales linearly with the number of objects (NEOs, stars, galaxies, noise, etc.) recorded per second. The number of objects detected per second (the "object rate"), and therefore computer requirements of the NEO survey outlined above, can be estimated from the current performance of the Spacewatch Telescope. The computer system in use at the Spacewatch Telescope can detect up to 10,000 objects in a 165-s exposure. Thus, its object rate is 60/sec. Scanning to V = 22 requires detection of about 30,000 objects/square degree. For an image scale of 1 arcsec/pixel, using the scanning rates tabulated above, and allowing a ten-fold increase in computing requirements to perform real-time image profile analysis, we calculate the total network computer requirement to be 2,000 to 3,000 times that at the Spacewatch Telescope. Therefore at each of six telescopes, it would be 300 to 500 times that at Spacewatch. Such requirements, although not easy to achieve, are possible using parallel processing with relatively modest modern workstations.

There are at least three levels of observational data storage that can be envisaged: (1) preservation of image-parameter or pixel data only for the moving objects detected; (2) preservation of image-parameter or pixel data for all sources detected (mostly stars); (3) storage of all pixel data. The first option is clearly undesirable, because data for slow-moving NEOs mistaken as stars would be lost. The first two options have the disadvantage that there would be no way to search the database, after the event, for sources whose brightnesses are close to the limiting magnitude and that would therefore have been discarded. The third option---the most attractive scientifically---may appear to result in serious problems of data storage and retrieval. However, we anticipate that, using technology shortly to be available, the third option is tractable.

On the order of one thousand NEOs and one million main-belt asteroids could be detected each month---about ten detections per second of observing time. Therefore, only moderate-speed data communication is needed between observing sites and a central-processing facility. Careful observational planning will be required to ensure efficient coverage of pre-programmed scan patterns, to avoid unintentional duplication of observations, to schedule the necessary parallactic and follow-up observations, and to optimize program changes so as to maintain robustness of the survey in response to shutdowns. Successful operation of this survey system will also require the coordination and orbital computation capabilities of a modern central data clearinghouse as described in Chapter 6.

Contents

Next Chapter