JWST Observing Overheads and Time Accounting Overview
JWST proposers will use the Astronomer's Proposal Tool (APT) to estimate the total time required to achieve their observing program science goals, including total science time, direct observing overheads, and indirect overheads.
JWST operations will be event-driven, where events will occur sequentially in time as soon as the previous event is complete. This fact drives the JWST single stream proposal process. In contrast, HST events are driven by orbital viewing periods and are constrained to begin at a specific time.
Observations by JWST are broken into one or more visits. A visit is a grouping of activities scheduled together as a unit by APT. The JWST schedule is constructed from an optimized sequencing of visits from all JWST programs. Unless special requirements are specified for the observations, there is no guarantee that the visits in a given multi-visit observation (or all the observations in a given proposal) will be executed in a contiguous manner.
Details of the actual timelines that would execute for a proposed set of observations are not available at the time of proposal submission. Therefore, the estimate for the total time requested for a given proposal must be based on a statistical model of JWST operations. This model is used in the Astronomer's Proposal Tool (APT) where calculations for a proposal's total time allocation request is based on a set of assumptions, rules, and inputs for deterministic and statistical overheads for various activities. APT reports both the total science time and an estimate of the total time allocation request in the APT proposal's Proposal Information cover page, which, upon submission, becomes the official time allocation request for a proposal.
Different kinds of observing overheads
Overheads for JWST observations occur in 3 major categories:
- Observatory direct overheads are activities directly associated with a given observing program, such as major slews, mechanism motion times, guide star acquisition times, small angle maneuvers (SAMs; i.e., motions between dither points), and target acquisitions.
- Observatory indirect overheads are related to activities performed for the general support of science observations. This includes calibrations, momentum management, wavefront sensing and control, as well as other observatory maintenance. By NASA and STScI policy, a pro-rated fraction of indirect overhead time is assigned to each proposal. Indirect overheads are calculated statistically (currently assumed to be ~16% of JWST's time).
- Instrument overheads are activities directly associated with each instrument, such as filter changes, detector readout, and operations script compilation.
The time assumed for each overhead activity is calculated either deterministically or statistically.
Deterministic time estimates are those with times known a priori and typically directly associated with the sequence of activities within a visit. For each visit, the deterministic direct overheads include guide star acquisition, target acquisition (when appropriate), any mechanism motions, and SAMs.
- Statistical time estimates are those with times that depend on the exact sequence of JWST observations and cannot be deterministically calculated during the proposal submission process (the Science Observation Design Reference Mission; Gordon et al. 2012b; Tumlinson et al. 2013). Direct statistical overheads include the assumed major slew time from a previous target to the first visit of each target in the proposal. Because visits are the primary scheduling unit, and because a given proposal’s visits may be interleaved with those from other proposals and activities, the actual slew prior to each visit cannot be precisely known until the program executes.
Estimating total time allocation request for a JWST program
APT provides an estimate of the science time (time spent actually acquiring data) and total (including overheads) time charged for each observation. These values are estimated using the sequence of JWST activities for each observation, based on visit-breaking rules in APT, the APT pointing model to sequence visits within an observation, and the sequence of activities within each visit. In general, the largest source of overhead for a JWST observing program is the number of major slews executed, and the overheads associated with each visit (e.g., guide star acquisition, visit clean-up activities).
An APT observation is broken into visits according to the following rules:
- The total pointing change is large enough that it is no longer feasible to use a single guide star for all exposures. For moving targets, this distance is 30”. For fixed targets, this distance is a function of galactic latitude and is based on the density of available guide stars. The visit splitting distance is shown in each observation template in APT after a target is selected.
- The total duration of the visit exceeds a maximum of 24 hours. This limit is imposed for efficiency (because very long visits are hard to schedule efficiently) and to preserve flexibility to insert engineering visits where needed.
- Different instrument templates are used, and separate observations are required (which is automatically a new visit).
The visits are ordered by the APT pointing model, which uses the following strategy:
- For each target in the observation (supports target groups)
- For each mosaic tile in the mosaic specification
- For each filter/grating/exposure specification
- For each dither point (primary and secondary dithers are expanded first, then the pointing list is iterated. Visits will not be split between secondary dither points, but may be split at a primary dither point.
- Generate the pointing information for the visit (to be passed downstream to scheduling).
A visit's duration is the sum of the time it takes to execute all science exposures and the time for other associated activities (including initial slews, guide star acquisitions, mechanism motions, frame resets, small angle maneuvers, and visit clean-up activities).
Note that because of the APT pointing model's sequencing of activities, all pointings within a dither pattern will execute for a given filter, and then repeat after a filter wheel change. In contrast, a given mosaic tile pointing will be observed in all requested filters before the pointing is changed to the next mosaic pointing. Some primary dither patterns have pointing offsets large enough to require new guide star acquisitions that will "break visits" and incur additional overheads associated with a new guide star acquisition and visit clean-up activities. It is generally assumed that a new mosaic tile will require a new visit.
Special observations that incur additional time to configure have additional overheads. These include the time lost in order to disrupt the nominal JWST schedule for a disruptive target of opportunity observation, and any "fixed time" observation requests that place hard stakes in the scheduling timeline that can cause forced dead time in the schedule.
See also: APT Smart Accounting
APT initially assumes one major slew for every new observation specified in a proposal. In some cases, this will be approximately correct, but in others, it may significantly overestimate or underestimate the overall resources needed. For example, proposals that include targets with largely overlapping visibility windows (or even multiple observations of the same target with different position angles or instruments) may be observed in temporal proximity, and hence not require a large “average” slew for each new observation.
The goal of the total time request estimation is to produce realistic and fair overhead assessments for the whole range of potential proposals and observation types that will be received. The way APT accomplishes this goal is the following: once a proposal’s full complement of observations has been specified and the visit planner has been run to demonstrate schedulability, the user will invoke the proposal planning step called "smart accounting," which looks through the targets and observations specified in the proposal and formulates what are called same scheduling sets, that is, subsets of the requested observations that will likely be schedulable together. APT then reduces the number of major slews and other overheads it was initially charging to the proposal, thus reducing the total overhead charges for the proposal to a fair and equitable value.
It should be noted that these same scheduling sets are non-binding; that is, there is no guarantee that the observations of a given same scheduling set will actually be scheduled together unless appropriate special requirements have been specified that tell APT to do so. Rather, the same scheduling sets assess the probability that observations will be grouped if and when they are ultimately accepted and put into the scheduling system. After executing the smart accounting step, the revised total time allocation request is finalized and reported on the proposal information cover page by APT.
Modifying APT time allocation requests
The APT time estimates will be the official time allocation requests shown to the TAC panels and ultimately used in planning and scheduling for accepted proposals. The only exception identified to date is that certain target of opportunity proposals may need to specify proposed observations where the details are unknown at the time of the initial submission. In this or any other special case that may arise, APT provides the ability to specify a different proposed allocation for the proposal on the proposal information cover page, followed by the opening of a text box for the user to provide an explanation. It is expected that this option will be used extremely infrequently and only for such special cases.
How to improve the efficiency of JWST programs
Because of the pre-configured sequences of activities associated with APT observation templates for a given instrument/observing modes, users have limited options for modifying the sequence of observing activities and total overheads. The largest contribution to JWST observing overheads are the number of major slews associated with a JWST program, and the activities associated with each visit (including initial slews, guide star acquisitions, mechanism motions, frame resets, small angle maneuvers, and visit clean-up activities). Programs that minimize the number of major slew and the number of visits will typically achieve a higher efficiency (science time/total time) than programs with large numbers of slews and visits.
To the extent possible, select targets that can schedule together. Smart accounting tries to gather observations that can schedule together and thus reduce the major slew charges. This is the largest overhead reduction that APT can assess, and so the more grouped the proposed observations are, the more the reduction in overhead gained.
Pay attention to the visit splitting distance reported by APT. For the most part, users specify observations in APT, and APT then decides on the splitting of the observations into visits (i.e., the “scheduling units”). The visit splitting distance assumed by APT is tied to the Galactic latitude of each target because the density of potential guide stars drops off away from the Galactic plane.
APT reports the visit splitting distance along with the assessment of the number of visits required for the observation on the instrument/observing mode observation template GUI where one enters the exposure and dither specifications. Many observations are accomplished with a single visit, but other common observations may involve many visits. Each visit of an observation requires a guide star acquisition, which incurs an assumed charge of 284 s. Hence, anything a user can do that reduces the number of visits will reduce total guide star acquisition overheads.
A specific example where this might come into consideration would be NIRCam or MIRI imaging mosaics. The fields of view of these instruments are large enough that each new tile of the mosaic may be a separate visit. Depending on the size of the mosaic (i.e., the number of tiles), the size of the object or field to be observed, and the visit splitting distance reported for a given observation, it may be possible to adjust the amount of overlap in the mosaic tiles so that their separation is less than the visit splitting distance. If so, then APT can observe more than one tile in a given visit and thus reduce the total number of visits (and guide star acquisitions) required.
A special case of this idea would be an observation of a target group. If one has a set of target positions that are so close together that they could be scheduled within a single visit, the overheads for that set of observations can be greatly reduced from what would otherwise be estimated. Since a typical visit splitting distance is of order 30"–80", this is indeed a special case (perhaps a set of closely spaced IFU pointings, for example). Not only would such a set only require a single major slew, but a single guide star acquisition (and possibly target acquisition) would be needed. See the APT Targets article for more details on target groups.
Gordon, K., et al. 2012a, JWST-STScI002564
JWST Observatory and Instrument Overheads
Gordon, K., et al. 2012b, JWST-STScI-000045, Rev. C
Science Operations Design Reference Mission
Lee, J., Reid, I.N., Lotz, J.M., 2014, JWST-STScI-003986, SM-12
Overhead Accounting Policy for JWST Science Users
Tumlinson, J., et al. 2014, JWST- STScI-003350
JWST Observing Efficiency: Baseline Results from SODRM 2012