PaperlayerView exact revision on arXiv

Source-linked AI summary

ATLAS: A High-Cadence All-Sky Survey System

J. L. Tonry, L. Denneau, A. N. Heinze, B. Stalder, K. W. Smith, S. J. Smartt, C. W. Stubbs, H. J. Weiland, A. Rest

arXiv:1802.00879v2astro-ph.IM

TL;DR

ATLAS addresses the challenge of surveying the dynamic sky while finding dangerous near-Earth asteroids under resource constraints. It combines cost-optimized hardware with autonomous observing and low-latency processing, and reports strong asteroid performance alongside broad transient and variable-star science. The system is designed for extension, but wind and camera cross-talk remain operational constraints.

  • Problem

    Survey systems must monitor moving, variable, and transient objects across the sky while balancing facility, telescope, detector, computing, operations, and software resources.

  • Method

    ATLAS combines cost-optimized survey design with autonomous summit operations, image reduction, static-sky subtraction, and low-latency science processing.

  • Results

    ATLAS detects at least as many regionally dangerous asteroids during very close approaches as any other asteroid survey and reported 311 spectroscopically classified transients.

  • Takeaways & Limitations

    ATLAS’s low cost and reproducibility make its survey capability relatively easy to extend toward more frequent, broader sky coverage.

  • Takeaways & Limitations

    Wind can buffet the shutter and elongate images, while fast readout produces correctable bias patterns and cross-talk.

Abstract

from arXiv · show

Technology has advanced to the point that it is possible to image the entire sky every night and process the data in real time. The sky is hardly static: many interesting phenomena occur, including variable stationary objects such as stars or QSOs, transient stationary objects such as supernovae or M dwarf flares, and moving objects such as asteroids and the stars themselves. Funded by NASA, we have designed and built a sky survey system for the purpose of finding dangerous near-Earth asteroids (NEAs). This system, the "Asteroid Terrestrial-impact Last Alert System" (ATLAS), has been optimized to produce the best survey capability per unit cost, and therefore is an efficient and competitive system for finding potentially hazardous asteroids (PHAs) but also for tracking variables and finding transients. While carrying out its NASA mission, ATLAS now discovers more bright ($m < 19$) supernovae candidates than any ground based survey, frequently detecting very young explosions due to its 2 day cadence. ATLAS discovered the afterglow of a gamma-ray burst independent of the high energy trigger and has released a variable star catalogue of 5$\times10^{6}$ sources. This, the first of a series of articles describing ATLAS, is devoted to the design and performance of the ATLAS system. Subsequent articles will describe in more detail the software, the survey strategy, ATLAS-derived NEA population statistics, transient detections, and the first data release of variable stars and transient lightcurves.

1. INTRODUCTION

ATLAS applies cost-aware survey design to autonomous, high-cadence all-sky monitoring, targeting dangerous near-Earth asteroids while supporting transient, variable-star, and moving-object science. Its design balances facility, telescope, detector, computing, operations, and software resources against survey performance.

  • Survey motivation: ATLAS divides survey resources among infrastructure, optical hardware, detectors, computing, operations, and software because any component can limit performance.The system is designed around the tradeoff between survey capability and per-unit cost.
  • Survey optimization: Equation 1 separates survey design expectations from operational performance using collecting area, field of view, efficiency, cadence duty cycle, PSF footprint, sky brightness, and survey depth.The left-hand side describes how well a survey ought to perform, while the right-hand side describes how well it actually performs.
  • Survey optimization: $1M is the replication cost of an ATLAS unit, making survey speed divided by cost the system’s value metric.The funded implementation uses 0.5 m Schmidt telescopes to optimize telescope and detector cost.
  • Survey optimization: 15 deg^2/sec is the predicted speed for one ATLAS unit, while 10 deg^2/sec is its best-condition operational speed at m5σ ∼19.7 in 30-second exposures and 40-second cadence.The operational value is lower because the design equation omits read-noise and dark-current terms.
  • System operation: ATLAS operates autonomously with summit control, image reduction, static-sky subtraction, and science processing for moving objects.Autonomous operation supports low-latency processing and discovery for impending impacts.
  • Scientific capability: The system supports NEA discovery, supernova searches, gravitational-wave counterpart searches, stellar-variability catalogs, asteroid characterization, and satellite tracking.Reported outputs include 1175 candidate supernovae, 311 spectroscopically classified transients, variability in 5 million objects, and accurate satellite positions and velocities.
  • Scientific capability: Two ATLAS units provide an inexpensive, reproducible survey strategy that can cover the sky to m ∼20 with a 1-day multi-exposure cadence when deployed at separated sites.The paper presents the system hardware and software broadly, with specialized subsystem papers planned separately.

2. ENCLOSURE

ATLAS evaluates enclosure designs for reliable autonomous operation and selects modified Ash domes. The enclosure integrates environmental protection, maintenance access, monitoring, fail-safe closure, and power continuity.

  • Enclosure choice: Ash domes were selected over clamshell, roll-off-roof, and truck-based designs because wind, leakage, ambient light, and reliability mattered for autonomous operation.The dome experiences unusual stress from rotating about every 40 seconds, so threadlocking adhesive was added to most fasteners.
  • Structure and access: The enclosure is a 16.5-foot-diameter half-sphere over an 8-foot cylinder, with an offset concrete pier, hoist mount points, and a motorized exhaust fan.The pier is offset 14 inches south and measures 41 inches tall by 30 inches in diameter.
  • Dome operation: Servo-controlled dome motion uses a stepper motor, absolute position sensing, and slip rings, allowing unrestricted rotation and movement during CCD readout.The dome moves about 25° during the roughly 9 seconds of CCD readout and shutter overhead.
  • Operations support: A mezzanine loft, stairs, network connections, industrial computers, and Raspberry Pi interfaces support access and control of the telescope, dome, camera, and observatory equipment.The equipment is distributed between on-site control and device-level interfaces.
  • Environmental protection: Cooling, dehumidification, webcams, audio communication, manual controls, rain sensing, emergency dome closure, UPS units, and environmental telemetry support safe autonomous operation.The fail-safe Raspberry Pi closes the dome immediately when power fails or rain or mist is detected.

3. MOUNT

ATLAS uses a German equatorial mount chosen for simplicity, cost, and all-sky performance, with fast slewing and scheduling that limits meridian-flip overhead. Mount and wind-related image elongation remain operational concerns.

  • 3. MOUNT: A German equatorial mount was selected because az-altitude systems degrade near zenith, while a fork mount is more complex and expensive at this scale.The mount carries seven counterweights of about 35 kg each.
  • 3. MOUNT: 15 deg/sec slew speed enables moves smaller than 45 deg to complete in 6.5 ± 0.8 sec, below the CCD readout time.A meridian flip typically takes 25 sec, so scheduling software minimizes such flips.
  • 3. MOUNT: Servo-loop instability occasionally elongates images, and retuning cures the issue although its recurrence is not fully understood.This is an explicitly unresolved operational limitation of the mount.
  • 3. MOUNT: The optical assembly is compact, with 1.9 m from shutter to mirror-cell back, a 0.4 m focus unit, and a 0.8 m mirror-cell diameter.The field corrector and camera housing has a 0.25 m diameter.
  • 3. MOUNT: Winds above 40 km h−1 can buffet the shutter, elongating images depending on dome position.ATLAS opens below 30 kph and must close above 60 kph.

4. TELESCOPE

ATLAS uses fast, wide-field Wright Schmidt telescopes with broad filters and corrected optics. Haleakala achieves sharper images than Mauna Loa, whose corrector retains residual astigmatism.

  • 4. TELESCOPE: Each telescope combines a 0.5 m Schmidt corrector, 0.65 m spherical primary mirror, three-element field corrector, filter, cryostat window, and detector.The optical train is a Wright Schmidt design.
  • 4. TELESCOPE: The 1.0 m focal length gives an f/2.0 system with a field diameter of about 7.5° and modest chromatic aberration across the broad c band.The c band spans 420–650 nm, while the redder o band spans 560–820 nm and produces sharper images.
  • 4. TELESCOPE: Corrector replacement improved Haleakala image quality from an initially delivered 3.8-pixel FWHM, while Mauna Loa retains residual astigmatism.The optics on Haleakala produce images slightly better than 2.0 pixels (3.5″), whereas Mauna Loa reaches about 2.8 pixels (5″).
  • 4. TELESCOPE: Collimation uses ray-traced out-of-focus donut mosaics and human judgment, while absolute-encoder focusing and detector tip-tilt adjustments support alignment.Once collimated, the telescopes seem to hold their adjustment well.
  • 4. TELESCOPE: ATLAS uses cyan, orange, and tomato filters spanning 420–650 nm, 560–820 nm, and 560–975 nm, respectively.The tomato band is intended to provide differential sensitivity to the silicate band of stony asteroids relative to orange.

5. CAMERA

The custom ATLAS camera combines large, fast-readout CCDs with deep cooling and a vacuum cryostat. Its performance is shaped by read noise, dark-current behavior, detector flatness, and corrected electronic artifacts.

  • 5. CAMERA: The in-house camera was designed for a 130 mm optical field, pixels no larger than 10 µm, sub-10-sec readout, and a 0.75 duty cycle for 30-sec exposures.Requirements also included remote detector tip-tilt and a detector temperature below −50 °C.
  • 5. CAMERA: The STA-1600 detector has 10560×10560 9 µm pixels, while the cryostat must maintain vacuum at 1 mtorr or less for at least a year.The detector and cryostat were selected to meet ATLAS's custom mechanical and operational constraints.
  • 5. CAMERA: Sixteen-amplifier readout takes about 9 sec, giving a 75% shutter-open duty cycle for 30-sec exposures, with about 11 e− read noise.In a 30-sec exposure, typical moonless backgrounds are about 300 e− in c and 350 e− in o, and read noise degrades SNR by about 16–18%.
  • 5. CAMERA: At −50 °C, dark current averages about 0.8 e−/pix/sec and doubles every 5 K, but is not significant in the wide c and o filters at roughly −53 °C.It remains a concern for 10 nm narrow-band and UV filters; cooling can lower it to about 0.2 e−/pix/sec.
  • 5. CAMERA: Fast readout produces plaid bias structure and cross-talk, which are corrected with overclocks, masking, and subtraction using a measured cross-talk matrix.The cross-talk is reported as not a serious issue after these corrections.

6. COMPUTERS AND SOFTWARE

ATLAS combines autonomous summit computing, distributed reduction, and specialized science clients to deliver low-latency moving- and stationary-transient processing. The system handles substantial nightly data while retaining operation during poor connectivity.

  • 6. COMPUTERS AND SOFTWARE: Each summit maintains six 1U servers for gateway, camera administration, and general-purpose computation, supporting autonomous operation without an internet connection.The computers reside in a separate summit computer room.
  • 6. COMPUTERS AND SOFTWARE: A normal night produces about 900 images per camera type and approximately 150 GB of raw compressed data.The data volume includes main, auxiliary 35mm, and fisheye 35mm cameras.
  • 6. COMPUTERS AND SOFTWARE: ATLAS performs initial reductions at observatories and image subtraction and science processing at a base cluster, targeting less than one hour from shutter close to final results.This distribution doubles stored and copied data but supports poor-bandwidth operation and future unit proliferation.
  • 6. COMPUTERS AND SOFTWARE: The moving-object client adapts Pan-STARRS MOPS to link detections and report observations to the Minor Planet Center, requiring about 25 minutes of real time for a typical 10k observation.The full process takes approximately 40 minutes of CPU time.
  • 6. COMPUTERS AND SOFTWARE: A separate ATLAS Transient Server links stationary detections into objects and reports supernova candidates to the IAU Transient Name Server using the moving-object pipeline's input files.It runs after the final detection table is created.

7. SYSTEM FEATURES

ATLAS combines autonomous weather protection, calibrated fisheye monitoring, and synchronized auxiliary imaging to support reliable, high-quality all-sky operations. These instruments quantify clouds and extinction while extending monitoring across a broad magnitude range.

  • Autonomous operations: The metfish combines cloud, weather, GPS, and fisheye sensors, while dome systems use rain and power monitoring to close shutters autonomously.The Boltwood system samples sky clouds every second; a Raspberry Pi can close the dome during rain or power failure.
  • Calibration: Galaxy-15 calibration measures shutter movement latency at 0.281 ± 0.017 sec.The calibration uses the satellite's precisely tracked position to determine the interval from command initiation to the shutter blade reaching halfway across the aperture.
  • Fisheye monitoring: The fisheye cameras operate on a 40 second nighttime cadence, with the Hawaii units staggered by 20 seconds so one shutter remains open.Night exposures last 32 seconds, and the images receive the same flattening, astrometric, and photometric calibration as scientific data.
  • Fisheye monitoring: Fisheye data achieve about 0.1-pixel astrometric residuals and 0.02 mag photometric accuracy for suitably bright stars.The astrometric precision is primarily limited by the undersampled PSF and combining color pixels into monochrome super-pixels.
  • Fisheye monitoring: A single fisheye image reaches about m = 7 at 5σ, while stacking improves sensitivity as N^1/2 for N much larger than 1000.The stated examples are m = 9.5 over an hour and m = 12 over an entire night, although faint stacked sources can be confused by the 4.3-arcmin pixels.
  • Auxiliary imaging: A synchronized Canon 5DIII and 135mm lens monitors 0 < m < 20 alongside the main science camera.Its 15°×10° field reaches approximately m ∼14 at 5σ in a 25 sec exposure, while fisheye images provide extinction measurements for up to 10,000 stars and quantify cloud opacity and motion.

8. PERFORMANCE RESULTS

ATLAS combines wide, rapid coverage with autonomous operations to detect near-Earth asteroids and support transient, variable-star, and gravitational-wave counterpart studies. Its performance includes broad sky access, substantial asteroid and variable-star yields, and strong detection rates for close or bright NEAs.

  • Sky coverage: A single ATLAS unit can cover 24,500 deg2 of accessible sky in one night with one 30-second exposure per pointing.The accessible region excludes sky near the Sun.
  • Cadence and depth: Two ATLAS units cover the entire accessible sky every two days, using four exposures per field and reaching o ∼19.5 per frame or o ∼20.2 in nightly co-adds.The four exposures are distributed over approximately one hour.
  • Operational availability: During June–September 2017, 92% of nights were at least half workable for one of the two summits, reflecting partially decorrelated weather.For each individual summit, about 80% of nights were at least half productive.
  • NEA performance: By January 2018, ATLAS had discovered 125 NEAs, 16 PHAs, and 9 comets, while submitting 5.5 million observations of 128,000 distinct asteroids.The system's trailed-asteroid sensitivity and characterization supported this moving-object capability.
  • NEA performance: ATLAS detected 75% of 287 NEAs that brightened past magnitude 19.0 under the evaluated visibility conditions; 19% of those detections were ATLAS discoveries.The sample included objects visible for less than one day, near full Moon, or moving as fast as 50 deg/day.
  • Variables and transients: ATLAS analyzed two-year lightcurves for about 20,000 asteroids and 140 million stars, detecting variability in 5 million stars.The analyzed sample covered declinations −30°<δ<+60° and magnitudes m<18.

9. CONCLUSIONS

ATLAS provides automated, low-cost, reproducible time-domain coverage while pursuing dangerous near-Earth asteroid discovery. Planned southern units would improve cadence and extend its benefits for asteroids, transients, variables, and gravitational-wave follow-up.

  • ATLAS now observes the complete northern sky every two days to fainter than m = 19 while routinely surveying for dangerous NEAs.
  • ATLAS data are available to any institution able to receive them, with no proprietary period.
  • Two planned southern units would enable the entire system to re-observe the sky every 24 hours, and some regions every 12 hours.
  • Tightening the net for NEA discovery: A full-sky, nightly ATLAS system would reduce the number of detectable NEAs that currently remain undiscovered each lunation.
  • Denser coverage for transients: Denser coverage could observe detectable supernovae within 12–24 hours of explosion, improving access to shock breakout and early follow-up spectra.
  • New variable stars in challenging classes: The survey could reveal challenging variable stars, including very-low-amplitude, alias-prone, and extremely long-period variables.
  • Immediate followup of any LIGO/Virgo transients: Its flexible target list could point an ATLAS telescope at a candidate LIGO/Virgo event within 60 seconds.
Loading 1802.00879v2…