The Automated Planet Finder (APF) at Lick Observatory on Mount Hamilton is a modern 2.4 meter computer controlled telescope. At one Nasmyth focus is the Levy Spectrometer, at present the sole instrument used with the APF. The primary research mission of the APF and the Levy Spectrometer is high-precision Doppler spectroscopy. Observing at the APF is unattended; custom software written by diverse authors in diverse languages manage all aspects of a night’s observing.
This paper will cover some of the key software architecture decisions made in the development of autonomous observing at the APF. The relevance to future projects of these decisions will be emphasized throughout.
The Automated Planet Finder (APF) was originally designed as a single purpose facility to search for exoplanets. The APF, however, has become a general use observatory that is used by astronomers the world over. We describe the improvements to our software for operations that both optimize finding planets with known periods and supporting a much broader community of astronomers with a variety of interests and requirements. These include a variety of observing modes beyond the originally envisioned fixed target lists, such as time dependent priorities to meet the needs of rapid varying targets, and improved tools for simulating observing cadence for the planet hunting teams. We discuss the underlying software for the APF, illustrating why its simplicity of use allows users to write software that focuses on scientific productivity. Because of this simplicity, we can then develop scheduling software, which is easily integrated into the APF operations suite. We test these new scheduling modes using a nightly simulator which uses historical weather and seeing data. After discussing this new simulation tool, we measure how well the methods work after a 36 month simulated campaign to follow-up transiting targets. We find that the data yield of each of the tested schemes is similar. Therefore, we can focus on the best potential scientific return with little concern about the impact on the number or duration of observations.
We report initial performance results emerging from 600 h of observations with the Automated Planet Finder (APF) telescope and Levy spectrometer located at UCO/Lick Observatory. We have obtained multiple spectra of 80 G, K, and M-type stars, which comprise 4954 individual Doppler radial velocity (RV) measurements with a median internal uncertainty of 1.35 ms−1. We find a strong, expected correlation between the number of photons accumulated in the 5000 to 6200 Å iodine region of the spectrum and the resulting internal uncertainty estimates. Additionally, we find an offset between the population of G and K stars and the M stars within the dataset when comparing these parameters. As a consequence of their increased spectral line densities, M-type stars permit the same level of internal uncertainty with 2× fewer photons than G-type and K-type stars. When observing M stars, we show that the APF/Levy has essentially the same speed-on-sky as Keck/high resolution echelle spectrometer (HIRES) for precision RVs. In the interest of using the APF for long-duration RV surveys, we have designed and implemented a dynamic scheduling algorithm. We discuss the operation of the scheduler, which monitors ambient conditions and combines on-sky information with a database of survey targets to make intelligent, real-time targeting decisions.
By July 2014, the Automated Planet Finder (APF) at Lick Observatory on Mount Hamilton will have completed its first year of operation. This facility combines a modern 2.4m computer-controlled telescope with a flexible development environment that enables efficient use of the Levy Spectrometer for high cadence observations. The Levy provides both sub-meter per second radial velocity precision and high efficiency, with a peak total system throughput of 24%. The modern telescope combined with efficient spectrometer routinely yields over 100 observations of 40 stars in a single night, each of which has velocity errors of 0.7 to 1.4 meters per second, all with typical seeing of < 1 arc second full-width-half-maximum (FWHM). The whole observing process is automated using a common application programming interface (API) for inter-process communication which allows scripting to be done in a variety of languages (Python, Tcl, bash, csh, etc.) The flexibility and ease-of-use of the common API allowed the science teams to be directly involved in the automation of the observing process, ensuring that the facility met their requirements. Since November 2013, the APF has been routinely conducting autonomous observations without human intervention.
The Automated Planet Finder (APF) is a dedicated, ground-based precision radial velocity facility located at Lick Observatory, operated by University of California Observatories (UCO), atop Mt. Hamilton in California. The 2.4-m telescope and accompanying high-resolution echelle spectrograph were specifically designed for the purpose of detecting planets in the liquid water habitable zone of low-mass stars. The telescope is operated every night (weather permitting) to achieve meaningful signal-to-noise gains from high cadence observing and to avoid the aliasing problems inherent to planets whose periods are close to the lunar month.
To take full advantage of the consistent influx of data it is necessary to analyze each night's results before
designing the next evening's target list. To address this requirement, we are in the process of developing a fully automated reduction pipeline that will take each night's data from raw FITS files to final radial velocity values and integrate those values into a master database. The database is then accessed by the publicly available Systemic console, a general-purpose software package for the analysis and combined multiparameter fitting of Doppler radial velocity observations. As each stellar system is updated, Systemic evaluates the probability that a planetary signal is present in the data, and uses this value, along with other considerations such as the star's brightness and chromospheric activity level, to assign it a priority rating for future observations. When the telescope is once again on sky it determines the optimal targets to observe in real time using an in-house dynamic scheduler.