The Palomar Transient Factory: System Overview, Performance, and First Results

The core of the PTF survey camera is a 12K × 8K mosaic made up of 12 2K × 4K MIT/LL CCID20 CCDs, arranged in a 6 × 2 array (Fig. 3). Three of the CCDs are high-resistivity bulk silicon (HiRho); the rest are standard epitaxial silicon (EPI). The EPI chips reach QEs of approximately 70% at 650 nm; the HiRho devices have somewhat higher QE, reaching ≈90% at 650 nm (Fig. 4). The HiRho chips also have lower fringing levels due to their increased thickness.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Depth of focus, and thus the flatness of the CCD array, is an important issue for the P48's f/2.5 beam. Approximately 92% of the PTF survey camera's CCD array is within 20μm of the reference focal plane position. Taking a detailed error budget into account, including terms for telescope jitter, atmospheric turbulence, and the optical quality, we predicted that 89% of the array would provide images meeting our 2.0'' FWHM image quality budget and that the remainder of the array would be within 0.2'' of the specification. On-sky tests confirm this (§ 2.9; the best images so far taken have a median FWHM of 1.5'' across the array). The instrument tip and tilt was adjusted to better than 0.003° tilt across the entire array using 20μm shims at the dewar mounting points and on-sky image tilt measurements from FWHM maps. The final full field-of-view image quality is better than 2.0'' FWHM in median seeing (§ 2.9).

The CCDs are 3-edge buttable and are separated by gaps of ≈500μm, corresponding to 35'' (see Fig. 3 for more details). The CCD array itself has a low number of bad pixels and offers excellent image quality across the field of view (Fig. 5).

Zoom In Zoom Out Reset image size

The full CCD array readout time is 30.7 s, using independent amplifiers for each chip and two Generation 3 Leach (SDSU) devices reading out 6 CCDs each. This readout time is approximately half that of the original camera electronics and was specified to match the expected telescope slew and settle time for PTF operations. The improved speed was reached at a small cost in readout noise; even so the final readout noise is < 15 e^- on all chips, well below the expected sky noise.

During the camera upgrade process, CCD03 became unresponsive. After diagnosing the fault as a problem inside the CCD package itself, we decided that attempting to repair the CCD would be both an unacceptable risk to the array and an unacceptable delay to the project. We are continuing to evaluate ways to repair or replace the malfunctioning CCD, but for at least the first year of science operations, PTF is accepting the 8% loss of imaging area.

We replaced the original flat dewar window with a field-flattening optic designed for P48's curved focal plane. The new optics are designed to provide better than 2'' image quality over the entire 3.4° span of the image in median seeing, including atmospheric, optical and mechanical tracking error terms. The optical distortion is 0.11% at the corners of the array relative to a uniform grid, a level comparable to differential atmospheric refraction over our field of view (Fig. 6 shows the distortion field).

Zoom In Zoom Out Reset image size

The telescope optics were also optimized: the primary mirror supports were tuned; all optical surfaces were cleaned or refurbished; baffles were improved to control stray light; the telescope tube was rendered light tight; and internal stabilized calibration sources were added to allow daytime quality control.

The original liquid nitrogen cooling system was replaced with a closed-cycle cooling system, implemented via a compact cold head located close to the CCDs. A Polycold Compact Cooler (formerly known as CryoTiger) with an expansion head with no moving parts was chosen to minimize vibration. We attached an H-shaped copper head spreader to the cold head to provide a nearly isothermal surface to which to attach flexible copper heat exchange straps leading to the CCD array assembly. A heating system attached to the dewar window actively maintains the CCD array temperature at 175 K.

The shutter is a commercial unit supplied by Scientific Instrument Technology, and employs dual split blades to avoid beam obscuration by the blades when the shutter is open. Only one blade covers the CCD array at any one time. An exposure sequence consists of stowing one blade, thus uncovering the CCD array, and then after an interval moving the second blade to cover the CCD array. The sequence is repeated in the opposite direction for the next exposure. The shutter unit is designed to provide minimal vignetting, and a frame-to-frame variation in exposure time of less than 2 ms, for all parts of the CCD array. The blade transit time is 1.300 s but the blades can be independently controlled to allow shorter exposure times in a slit scanning across the detector.

The filter changer assembly is a custom-built mechanism that provides for motorized selection of one of two filters; one filter is always in the field. Filter exchanges are completed in 15 s, using a microstepper motor. Absolute filter positioning is registered by running the filter frame against a mechanical hard stop. Normal PTF operation is expected to use at most two filters in each night, but the system has been designed for easy manual swapping of filters as necessary.

The PTF survey camera array of 12 CCDs is divided into two banks of six CCDs, with each bank handled separately by its own detector controller. The controllers are Generation 3 Leach (SDSU) devices comprised of three 2 channel video boards, three clock boards, and a timing board. Communication between a controller and its respective host computer is conducted via a fiber optic serial link between the timing board and a corresponding PCI card in the host computer. The controllers share a common clock so that the exposures and readouts of the two controllers are synchronized with each other. The master camera computer also handles all the nondetector camera hardware—the shutter, filter changer, temperature sensors, and calibration LEDs. The camera control software is custom built for the PTF survey camera, and is based on panVIEW, the Pixel Acquisition Node (PAN) version of ArcVIEW (Ashe et al. 2002).

The survey operations are performed robotically by the P48 Observatory Control System (OCS), written in MATLAB. The OCS is responsible for controlling the camera, filter exchanger, shutter, telescope, dome, and focuser, on the basis of feedback information from the scheduling system, the camera and telescope, a weather station, and the data quality monitor (§ 2.8).

The OCS is responsible for sequencing all PTF observations, starting with the bias frames before sunset, through focus images, and finally the science images. The PTF scheduler is responsible for selecting the next target for observation. The PTF schedule is not predefined; the next target is selected 30–90 s before each exposure starts, during the previous exposure.

The criteria for selection of the next target include: (1) Sun altitude; (2) excess in sky brightness due to the Moon, using the algorithm of Krisciunas & Schaefer (1991); (3) Moon phase; (4) time needed to move telescope to the next target; (5) time needed to move dome to the next target; (6) airmass; and (7) time since the last sequence of observations of this field. In addition for some programs, such as the 5 day cadences (§ 3.2), the scheduler is required to optimize the field selection so that each field is observed twice during the night for asteroid detection.

These criteria are used to calculate weights for each field, and the scheduler selects the target with the highest weight for observation. In order to keep the operations of the scheduler as simple as possible, most of the weights follow a step function. An exception is the cadence (time since last observation) weight (W) which is of the form

where JD is the current Julian day, JD_l is the Julian day corresponding to the last successful sequence (pair of images) of observation, τ is the cadence, and s is a softening parameter that defines the time scale in which the weight goes from near 0 to near 1.

A software suite running at the mountain, the Data Quality Monitor (DQM), analyzes all the recorded images as they are taken. An initial check searches for problems with file corruption, truncation, or missing FITS headers. Next, the image is searched for stellar objects (using SExtractor; Bertin & Arnouts 1996), and the average FHWM and other image quality parameters are checked for obvious problems. Finally, if it has passed all the checks, the file is sent to the PTF pipelines.

To detect more subtle image quality problems such as increased electronic noise, the P48 twice daily automatically takes test exposures of the telescope tube cover. A band of temperature-stabilized LEDs provides illumination stable to < 0.1%. Because the illumination has passed through every optical element of the system, comparing the test exposures to reference exposures provides a simple test if anything in the telescope + camera + electronics systems is not operating as expected. Many problems can be found and corrected before the start of nighttime operations.

Finally, both the LNBL and IPAC pipelines produce image quality and sensitivity measures for all detected fields, providing a long-term monitoring facility to allow us to optimize our observing setup and equipment.

The PTF survey camera system achieved first light on 2008 December 13. Initial commissioning work was completed in 2009 January and the system has been used for initial science test operations since then (with small gaps for engineering). Here we describe on-sky performance results from the first 2 months of operation. Since the system continued to be commissioned during this time, the performance is likely to improve as the PTF survey continues.

A total of 91.7% of the pixels in the full 12K × 8K array are light sensitive and not dark, occulted, or otherwise bad (this increases to >99% good pixels if we exclude the unresponsive CCD). Most of the approximately 200 bad columns in the array are concentrated on CCD05 (at one corner of the field). CCD09 offers the best combination of PSF quality and CCD cosmetics. The chips are linear to better than 0.5% up to 60,000 ADUs (typical gain is 1.6 e^- ADU^-1).

During the design of the PTF survey camera we specified a FWHM of 2.0'' during median Palomar conditions (1.1'' seeing). This requirement was based on a number of factors—the minimum PSF size that is reasonably sampled by our 1''pixel^-1 plate scale; what could be reasonably achieved with the P48 optical setup; optimization of the survey limiting depth; and the requirement to detect supernovae in the cores of bright nearby galaxies. Analysis of the first two months of PTF on-sky data shows that we have met or exceeded this requirement. Typical images taken in good seeing show both sub-2.0'' FWHMs and very small variations in image quality across the entire field of view (Figs. 7 and 8).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The survey depth is correspondingly encouraging (as shown in Fig. 9). Based on all the data taken during the last 2 months of commissioning, the system provides median 5σ, 60 s limiting dark-time magnitudes of m_g^' = 21.3 (25th percentile m_g^' = 21.7; 75th percentile m_g^' = 21.0) and m_R = 20.6 (25th percentile m_R = 20.8; 75th percentile m_R = 20.3).

Zoom In Zoom Out Reset image size

The entire PTF survey system (the P48 telescope, PTF survey camera, OCS, and the DQM) has a current uptime of ≈90%, excluding losses due to weather. As initial science operations proceed, many issues are being addressed and the uptime is continually improving. The survey efficiency is currently approximately 50% open-shutter time, compared to a goal (and theoretical maximum) of 66%. This is due to the cumulative effect of many small inefficiencies in the system operation and is continually improving.

The three major PTF experiments, together accounting for 92% of the project's P48 time, are based on two filters: SDSS g^' and Mould-R (Fig. 4). These filters were chosen to optimize the survey detection limits in full moon (R-band) and dark-sky (g^'-band) conditions for a variety of different transient classes (see Rau et al. 2009). During dark-sky conditions, the reduced quantum efficiency in the g^' filter is more than compensated for by the darker sky and the blue colors of many of our targeted transients. Mould-R is very similar to SDSS r^' and allowed us to reuse an existing CFH12K filter; it is the standard filter for programs that are expected to spend at least part of their time operating while significant moonlight is present. Because no follow-up is done with P48, we elected to use standard passbands for the survey to allow a meaningful comparison of PTF detection magnitudes to follow-up telescope results.

The initial standard exposure time for the 5DC and DyC surveys is 60 s, based on the readout time of the camera, expected system efficiency, and the expected sky noise and dark current in the exposures. In later stages of the PTF project this exposure time will be optimized for the different programs expected to be pursued.

The 5 day cadence experiment (5DC) will run yearly from March 2 until October 30 using 65% of the time in that period. The main goal of this experiment is to construct large samples of type-Ia SNe and core-collapse SNe. It will also study AGNs, quasars, blazars, CVs, extragalactic novae, luminous red novae, and tip of the AGB variability. The experiment will observe a footprint of approximately 8000 square degrees, with |b| > 30°, ecliptic latitude |β| > 10°, and with a median cadence of 5 days. At a given time, the active area for the 5DC search (including all overheads) will be 2700 square degrees. The 5DC observations will be conducted in almost all lunar phases in R band. In each epoch we will obtain two 60 s exposures separated by 60 minutes. The two images will be used to remove cosmetic artifacts and cosmic rays and to find solar system objects.

The dynamic cadence experiment (DyC) is designed to explore transient phenomena on time scales shorter than ∼5 days and longer than 1 minute. The experiment will explore unknown territories in the transients phase space—mainly fast transients with time scales of less than a day. Therefore, the dynamic cadence will be an evolving experiment. Every 6 months, the PTF collaboration will review the results of the dynamic cadence experiment and decide how to continue.

Motivated by a search for short-lived transients in the brightness gap between novae and supernovae (-10 < M < -16), the DyC experiment is targeted at fields that are luminosity concentrations in the local universe. Since these transients are expected to be fainter than M ∼ -16, the search volume is limited to DM (distance modulus) < 36.5, i.e., less than 200 Mpc.

The total luminosity of nearby galaxies in the PTF FoV was computed on a fine grid of possible pointings (limited by declination >-30). This was done 7 times with 7 different maximum distances (in steps of 0.5 between 30.5 ≤ DM ≤ 36.5). We assigned a weight for each pointing based on the fraction of maximum luminosity in that list. A total weight for each pointing was computed by averaging the individual weights. Finally, pointings with the highest weights were chosen for PTF. This list naturally includes the nearest, brightest galaxies (e.g., M31, M51, M81) and nearest galaxy clusters (e.g., Virgo, Perseus, Coma), but also covers the other accessible nearby mass concentrations from physical and chance galaxy alignments.

This experiment will detect RR Lyr stars in the Galactic halo, CVs, AM CVn stars, flare stars, SNe, eclipsing binaries, and solar system objects. Moreover, this experiment will put strong constraints on the existence of any optical counterparts that may be associated with radio transients.

The Orion project has been assigned 40 consecutive nights year^-1 for 3 yr to perform intensive time-series observations of a single field in the Orion star forming region, with the aim of detecting close-in Jupiter-sized planets transiting young stars. The observations will be acquired in R band with no dithering in a continuous sequence of 30 s images, with the goal of producing high-precision (better than 1%) differential photometry. The field will be chosen to represent a stellar age range where protostellar disks should be on the point of dissipation (5–10 Myr)—giving important clues about planet formation and migration—and to optimize for source density and avoid regions where the stellar variability in the ensemble is severe enough to prohibit precise differential photometry. The IPAC PTF pipeline will be used for standard image reduction, source identification, and astrometry, but a dedicated photometric pipeline is being developed to produce the high-precision differential photometry.

For 3 nights each lunation during full Moon we will carry out a 3π sr sky survey in Hα. The survey area will include all the sky with δ > -25°. Each field will be observed twice in four narrowband (approximately 10 nm width) filters around the Hα line covering the redshift range of about zero to 0.05, corresponding to distances of 0 to 200 Mpc. Narrowband filters that give a consistent bandpass across the entire field of view of the camera are difficult to produce, especially in the fast beam of the P48. However, a description of usable Hα filters for a Schmidt telescope with the same optical design is given in Parker & Bland-Hawthorn (1998), and our industrial partners are confident that these filters can be constructed.

We estimate that the Hα survey sensitivity limit will be ≈0.6R in a 2.5'' aperture¹³; a detailed comparison of this sensitivity to other Hα surveys can be found in Rau et al. (2009). We are currently designing a scheme that will allow us to photometrically calibrate these data to about 10% accuracy.

The PTF quick-reduction and subtraction pipeline was designed from the start to take full advantage of the parallel high performance computing facilities at the National Energy Research Scientific Computing Center (NERSC).¹⁴ The major improvement to the hardware used in this pipeline over previous versions designed by the Supernovae Cosmology Project (Perlmutter et al. 1999) and Nearby Supernovae Factory (Aldering et al. 2002) at LBNL is the use of the NERSC Global Filesystem (NGF). NGF is a 300 terabyte shared file system with over 1 gigabyte s^-1 bandwidth for streaming I/O which can be seen by all of the high-performance computers. On the software front, the improvements include a tight coupling of a PostgreSQL database involved in tracking every facet of the image processing, reference building, image subtraction and candidate detection along with a complete rewrite of the processing and subtraction codes, described as follows.

Incoming P48 images are immediately backed up on the High Performance Storage System, a 3 petabyte tape archive system at NERSC. Processing begins by splitting the packed images apart by chip, applying crosstalk corrections, and performing standard bias/overscan subtraction and flatfielding. After this point, all operations are performed on a chip-by-chip basis in parallel. Catalogs of sources are created for each image via the Terapix SExtractor code (Bertin & Arnouts 1996), which then is fed to the astrometry.net¹⁵ code to perform an astrometric solution. At this point a comparison to the USNO-B1 (Monet et al. 2003) catalog is made to determine the zeropoint and 3σ limiting magnitude of the image along with a calculation of the seeing. The image is then loaded into the processed image database, storing all relevant information from the FITS headers.

After several images are obtained for a given PTF pointing, we are able to create a reference image. The reference images are created via the Terapix software Scamp and Swarp (Bertin 2006) after querying the processed image database in order to obtain the highest quality input images for a given pointing/chip combination. The references are assigned version numbers and (along with their relevant information) are stored in the reference database.

Subsequent new images are processed as just described, and if a reference image exists for them, a subtraction is performed. First, the new image is astrometrically aligned to the reference by utilizing Scamp and the new and reference catalogs. The reference is then Swarped to the size and scale of the new image and a subtraction is performed using HOTPAnTS¹⁶. The subtraction is performed in both the positive and negative direction (reference minus new and new minus reference) to detect both positive and negative flux changes. Candidate transient sources are detected via SExtractor and along with other relevant parameters (location with respect to bad pixels, etc.) are stored in a candidate and subtraction database. The subtraction image that led to the first PTF supernova discovery is shown in Figure 10.

Zoom In Zoom Out Reset image size

In addition, PTF takes advantage of DeepSky.¹⁷ DeepSky was started in response to the needs of several astrophysics projects hosted at NERSC. It is a repository of digital images taken with the point-and-stare observations by the Palomar-QUEST Consortium (Djorgovski et al. 2008) and the Near Earth Asteroid Team (Bambery et al. 2007). This data spans nine years and more than 15,000 square degrees, with 20–200 pointings on a particular part of the sky and cadences from minutes to years. For a large fraction of the survey, DeepSky achieves depths of m_R > 23 magnitude. In total there are more than 11 million images in DeepSky. This historical data set complements the Palomar Transient Factory survey by identifying known variable stars, AGN, and the detection of low surface-brightness host galaxies of supernovae.

During commissioning, human scanners were utilized to reject "junk" transient detections via a web interface. The subtraction system and rejection algorithms are currently being improved with an aim of full automation by making use of machine learning techniques. The results of human scanners are being used to train an automatic classifier; a more detailed account of this classification technique is in preparation.

Any statistical inference of PTF transient rates will require a careful measurement of the survey detection efficiency. We have constructed a pipeline that produces realistic fake variable objects and adds them into PTF images, on which we apply the standard PTF search algorithms to determine the detection efficiency for any specific data subset. We currently model three generic classes of transients: variable stars, SN/AGN-like sources (i.e., point sources in host galaxies), and "rogue" transients (point sources on random positions in the image). The transients are directly injected into PTF images using point-spread functions (PSFs) modeled from nearby stars in the fields.

The Transients Classification Pipeline (TCP; Bloom et al. 2008; Starr et al. 2008a,2008b) is a parallelized, Python-based framework created to identify and classify transient sources in the real-time PTF differencing pipeline (§ 4.1). The TCP polls the candidate database from that pipeline and retrieves all available metadata about recently extracted sources. Using the locations and uncertainties of the transient candidate objects, the TCP either associates an object with existing known sources in the TCP database or, after passing several filters to exclude nonstellar (e.g., known minor planets) and nonastrophysical events, generates a new source. The Bayesian framework for event clustering is given in Bloom et al. (2008). Another classification engine will also be used on PTF transients (Mahabal et al. 2008a,2008b).

Once a transient source has been identified, the TCP generates "features" which map contextual and time-domain properties to a large dimensional real-number space. After generating a set of features for a transient source candidate, the TCP then applies several science classification tools to determine the most likely science class of that source. For rapid-response transient science, a subset of features—such as those related to rise times and the distance to nearby galaxies—is most useful. As light curves are better sampled and colors are obtained, more features are used in the classification. The resulting science class probabilities are stored in a database for further data-mining applications. Sources with high probabilities of belonging to a science class of interest to the PTF group will be broadcast to the PTF's "Follow-up Marshal" (§ 5) for scheduling of follow-up observations. As more observations are made for a known source, the TCP will autonomously regenerate features and potentially reclassify that source.

While the transient pipeline will be optimized for rapid turn-around, the IPAC pipeline will be geared toward providing the best possible photometry consistent with hardware capabilities and data rate. To enable the generation of optimal calibration files (primarily flat fields), the IPAC pipeline will commence image processing only after all of a given night's observations have been received. Additional source association operations, database updates, and data transfer to the Infrared Science Archive (IRSA)¹⁸ will be carried out over a period of several days. Processed images and incremental source lists will be available through IRSA within 2 days from the time the observations are taken.

After a given night's observations have been completed and received at IPAC (as indicated by receipt of end-of-night image manifest), the image processing pipeline is initiated. Parallel processing on 11 multiprocessor Linux machines is employed to simultaneously process the PTF mosaic's 11 working detectors and meet the data rate requirement. Key pipeline-processes are multithreaded to take full advantage of the 8 core CPU in each of the Linux machines.

The pipeline initially generates night-specific bias frames, and trims and bias-subtracts all images in the usual manner. Nightly flat fields are generated for each detector by source-masking and median-combining all of the night's sky observations; although twilight flats are taken each night, PTF data reductions use by preference these "superflats" generated from the night's observations. This is an involved and memory-intensive process and, given the number of exposures involved, takes 3–4 hr with the processing hardware currently in place. On completion, these flat fields are used to process the entire night's science observations. Raw and processed images as well as all ancillary and calibration frames are subsequently passed to IRSA for archiving and retrieval on demand.

Source detection is carried out using SExtractor (Bertin & Arnouts 1996). SExtractor is applied to each image a number of times in the pipeline to mask sources prior to flat-field generation, measure the effective seeing, generate pixel weights and masks, and finally to generate photometry for all detected sources. Pixel masks are generated to flag bad rows and columns, cosmic rays, saturated pixels, and charge bleeding. The final application of SExtractor is carried out after multiplying the processed image by a pixel area map to preserve point-source photometry.

For each image, the SExtractor output list is then astrometrically calibrated using the Terapix Scamp software (Bertin 2006). Photometric zeropoints are determined by calibrating against stars in the Sloan Digital Sky Survey (York et al. 2000), taking color terms into account. With appropriate airmass corrections, these zeropoints are then applied to all sources outside the SDSS footprint. Ultimately photometric calibration may also be related to flux measures from (saturated) Tycho stars (Ofek 2008).

Currently in development is a source association pipeline that will match, on a nightly basis, each detected source with all previous detections of that source. Appropriate pointers will be incorporated into the IRSA database, enabling users to query the photometric time history of every object ever detected by PTF.

The automated P60 observation queue (Cenko et al. 2006) is sent all candidates flagged as bona fide astrophysical transients (i.e., a new point source detected at least twice in P48 imaging and which either lacks a counterpart or has a galaxy host in the reference imaging). A 120 s snapshot in g^', r^', i^', z^' is obtained for quick classification. On a typical night with median weather conditions, this reaches a limiting magnitude of , , , and . Our efficiency currently allows 20 2 minute exposures hr^-1.

As soon as the 60 inch data are obtained, an automated pipeline detrends the data and solves for an astrometric and zeropoint solution. Next, if the location is in SDSS, the pipeline downloads the corresponding SDSS image (mosaiced to fit the P60 FOV) to serve as a template. A convolution kernel is measured and image subtraction performed. On the subtracted image, PSF photometry is used to measure the magnitude at the source location and the result is sent to the PTF database. If the source is not detected, an upper limit is returned. Currently, if the location is outside SDSS, direct photometry is performed. We note that host galaxy light can be a significant contaminant which is not properly corrected for in case of direct photometry. Hence, even for fields outside SDSS, we are investigating other options for template imaging for image subtraction.

The 60 inch snapshot observation provides independent confirmation that the candidate is a real transient (and not a subtraction artifact), a color for quick classification (for example, m_g - m_r ≈ 0 suggests that it is a SN Ia), and a second-epoch magnitude to assess photometric evolution. All this information helps direct further spectroscopic and multiband photometric follow-up with P60 and other telescopes.

We have designed a centralized follow-up Marshal to ensure that our discoveries are efficiently distributed to our follow-up resources and that the temporal coverage of a given target meets the requirements of the appropriate science program. The design of this system is complete, but the automated follow-up program has not yet begun. A later paper will detail the performance of this novel system.

The first duty of the follow-up Marshal will be to monitor the output of the transient classifier to identify any new sources detected by P48 that warrant follow-up observations. For this purpose, the Marshal will maintain a list of target descriptions (in terms of the classifier output) for each science program to define trigger conditions for same night and long-term monitoring. For example, a supernova program may define same night triggers for all objects with at least a 25% chance of being a supernova but no more than a 1% chance of being an asteroid. Discoveries worthy of same-night follow-up will be sent primarily to P60, although sources identified at the end of the night may fall to Faulkes Telescope North (FTN; Haleakala, Hawaii) in rare cases where twilight preempts the required observation. These requests will be issued as ToO triggers, similar to the gamma-ray burst triggers already handled by both P60 and FTN except that interruption of integrations or read outs will not be required.

After completion of an observation, the data must be analyzed to extract the target parameters (e.g., the magnitude and position in photometric data or the spectral classification, redshift, etc., for spectral observations). The responsibility for this data reduction lies with the partner in charge of the instrument, and the high-level data will be sent back to the Marshal. The Marshal will then relay this information back to the classifier to refine the classification, and it will archive the observation in its own database for future use.