A guided tour to the scheduling of an XMM-Newton orbit
(or: How we do our daily revolution)
How do we do it?
A sequence of XMM-Newton observations during an orbit is the result of a long and complex process. The goal of this process is to define an observing plan, which maximizes simultaneously the scientific return and the observation efficiency. This plan needs to take in proper consideration all the constraints imposed by observation strategies of some specific programs (e.g.: coordination between XMM-Newton and other satellites and/or ground-based observatories, monitoring programs, etc.), as well as the global safety of the spacecraft and of the scientific payload. Last, but not least, the observing plan must be ready enough in advance, to allow sufficient time to efficiently react to operational contingencies.
The planning of an XMM-Newton revolution is articulated in three main steps:
- the Advanced Plan
- the Short-Term plan
- the orbit-by-orbit scheduling
The goal of the Advanced Plan is to provide the community with mid-term planning information. The current baseline is a monthly updated, three-months long plan. This plan actually fills only 50% of the potentially available observing time. The Mission Planning Team needs the remaining part to optimize the observation efficiency during the preparation of the orbit-by-orbit scheduling.
During the scheduling process, the XMM-Newton Mission Planning Team prepares a detailed plan for a specific revolution, filling the whole available science time and optimizing the spacecraft slews and exposure times with respect to the instrumental overheads (which are a complex function of the instrumental modes). Once this process is completed, the scheduling plan is sent to the Mission Operation Center (MOC), at ESOC (Darmstadt, Germany), where it is transformed in telecommands and eventually sent to the spacecraft
Step 1: the Advanced Plan and the concept of "critical targets"
At the end of every month, the Mission Planning Team must release an Advanced Plan covering up to the next third month. The definition of this Plan starts from the sample of targets, which are visible during the 15 revolutions covering one month. Observable targets are automatically ranked according to a "criticality parameter", which is mainly based on the length and the time distribution pattern of the visibility window still available within the current observing cycle. However, other constraints can have higher priority and drastically alter the criticality rank. Fixed-time observations (e.g.: coordinated pointings) need in most cases to be placed in a determined position of the plan. The plan must foresee the need for routine calibration, or engeenering test observations. For the whole first observing cycle at least, the distribution of the observed targets must accurately reflect the balance between the Guaranteed Time and Guest Observed programs, and of course the priorities as defined by the Time Allocation Committee. All in all, these operations are performed manually.
The main goal of this first phase is to determine critical targets for each revolution. If two or more "critical targets" are present in the same revolution (e.g.: because two coordinated observations coincide within 48 hours), planning can become fairly cumbersome.
Step 2: optimizing the observation efficiency (the Short-Term Plan)
Once critical targets have been allocated to each revolution, the whole schedule must be filled, using the targets which are potentially visible in each observation. The main guideline is to choose (once again, manually) those targets, which are the closest to the critical ones, in order to minimize slew times. This is far from being trivial. The area of the sky visible by XMM-Newton at a given time is strongly limited by the solar or Earth aspect constraints (it is rarely larger than 10% of the entire sky). Pointing maneuvers within the allowed visibility regions can be forbidden, because they would cross the apparent path of the Earth and the Moon. In some cases the slew from/to the perigee attitude maybe be inhibited for the same reasons. In this case, the spacecraft must be tricked, and "dummy" pointing created to triangulate to the final attitude. Bright planets can be also a problem, if the OM is operated in a science mode. The possible paths are studied by eye with the help of plots as that shown in Fig.1.
Figure 1. - Visibility plot for an XMM-Newton revolution. Symbols are explained in text
The big blue blobs, dominating the vast majority of the sky, are the avoidance areas due to the solar constraints. The red big circles represent the apparent path of the Earth avoidance disk over the sky with increasing time (in hours) from the perigee passage. The blue smaller circles represent an analogous Moon path.
After several trials and errors, an acceptable solution is eventually found, and a time ordered list of candidate targets is produced. This list corresponds to a Short-Term Plan, whose feasibility still needs to be verified before implementation in the operational system.
Step 3: the orbit-by-orbit scheduling
One of the subsystems composing the XMM-Newton ground segment is the Schedule Generator System (SGS). Via a graphic interface, SGS allows to:
- generate a candidate list of observations, which are observable during a given revolution
- include them in the orbit-by-orbit XMM-Newton schedule
- generate - once the schedule is ready - all the files needed for the production of the actual timeline at the MOC.
The main window of the SGS appears as in Fig.2. It represents the span of an XMM-Newton revolution (about 48 hours). Each big rectangle bar represents one XMM-Newton observation (different colors code different proposal categories: "Calibration" in green, "Engeenering" in pink, "Science" in blue). Small yellow lines represent the spacecraft slews. They are automatically calculated by the SGS, once the Mission Planner includes two adjacent observations in the schedule. The SGS system has not been designed to support the generation of the Advanced or Short-Term Plans, or the optimization of the observation efficiency, due to the intrinsic complexity of the tasks.
Fig.2 - Graphic interface of the SGS
The best scenario at this stage of the planning process arises when the feasibility of the Short-Term Plan - manually elaborated during the previous steps - is actually confirmed by the SGS system as well. In this case the only task which is left to the Mission Planner is to optimize the exposure times for each exposure of each observation. The duration of an exposure is in facts set by the time requested by the proposal Principal Investigator, plus a quantity ("instrumental overhead"), which is a function of the instrumental mode. The total elapsed time of an observation is therefore the longest of these sums. The Mission Planner needs to compensate the exposure time allocated to instruments with shorter overheads, to avoid loss of science time. This operation, albeit conceptually simple, is fairly long, due to the intrinsic complexity of the XMM-Newton ground segment software. In practise, even in the most favorable cases, the scheduling of an XMM-Newton observation requires no less than a half working day.
Part of the XMM-Newton orbit cannot be fully used for scientific observations. At low altitude along the orbit, where the radiation environment background is high, the optical filters of the EPIC cameras must be kept closed, to prevent high energy protons from being focused onto the detector focal planes. Such particles can in facts seriously compromise the charge transfer efficiency of the CCDs. The EPIC observations during the first seven hours of each revolution therefore only contain internal calibration exposures. As far as possible, the first observation of a revolution has RGS as prime instrument, to maximize the scientific efficiency. This represents an additional complexity since EPIC is far more often requested as prime instrument than RGS.
In some cases more than one iteration is needed to complete successfully the schedule. Some subtle constraints appear in facts only once the final telecommands are generated at the MOC, i.e. at the very end of the overall process, and cannot be guessed with global plots such as those shown in Figure 1 or 2. One example is represented by the "handover gaps". The control of XMM-Newton makes use of three antennas: Perth (Australia), Kourou (French Guyana) and Santiago (Chile). When the control of the spacecraft passes from one antenna to the next, telecommanding capabilities are inhibited for about half an hour. These "black-out" phases, not only need to be void of any telecommand or slew, but one must also ensure that all preceding commands have finished executing before the gap starts. Four such "gaps" exist during each revolution. XMM-Newton possesses nine scientific subsystems to telecommand. One can easily realize that the likelihood of violation of this constraint is not negligible! In this case, the Mission Planner needs to move the position of the observations in the SGS to avoid such a condition. This may imply further modifications of the exposure times, and additional iterations.
Last, but not least, the scheduling process is subject to a rather long list of manual checks, which are not automatically performed by the SGS. These verifications concern - among other things - the set-up of the EPIC cameras instrument configuration below the science window start/end altitude, the verification of the RGS readout sequences, the optimization of BLOCKED filter OM observations, checking for the possible presence of bright planets within the OM field of view, etc ...
Finally, more than often the scheduling of an XMM-Newton revolution must be revised, or even completely changed at a later stage. Such changes can be prompted by science contingencies (e.g.: Targets of Opportunities; please refer to a specific paper, Santos-Lleo et al. 2001, ESA Bulletin 107 p.54, on this subject), last-minute-corrections of PI errors (e.g.: wrong coordinates, whose correctness is, however, exclusive responsibility of the PI), or operational contingencies (database updates, last minute change in the antenna availability etc.).
In spite of all the difficulties inherent in the process described above, we are proud to say that the scheduling process is now routinarily completed at least four weeks in advance the start of a revolution. This allows a time span wide enough, for scientific or operational contingencies to be tackled with a reasonably safe margin. The scheduling efficiency (i.e.: total scheduled time against the total duration of an orbit outside the radiation belts) is steadily increasing with time, and is now well above 80%.