Week 12 of the Early Operations system optimization period
The seasonal high humidity weather pattern of recent weeks continued at Cerro Pachón, leading to delayed start on several nights, and no on-sky observations being acquired on one night. The team has installed thermal insulation around utility lines that provide coolant to the LSST Camera, secondary mirror (M2), and telescope top end assembly to reduce the accumulation of water condensation at the top end assembly of the Simonyi telescope, and thereby increase the range of acceptable dewpoint temperatures for nighttime operation.
From the ensemble of on-sky tests over the past months, there is increasing evidence for additional physical variables in the system that perturb the optics beyond the wavefront measurement uncertainty, in addition to the expected effects related to telescope elevation, camera rotator angle, and filter-specific focus offsets that regularly occur in the course of survey operations, and that are incorporated into the open-loop look-up tables (LUT) of the Active Optics System (AOS) to provide first-order corrections for predictable patterns. Additional physical variables are expected at some level, e.g., slowly changing temperatures of the telescope components, and can be mitigated with the near-realtime feedback provided by the AOS closed-loop control system. However, the presence of unidentified variables can pose challenges for the optimization of the LUTs, which often incorporate observations from time periods that are large relative to the timescales over which the optical perturbations occur.
Such effects currently manifest at Rubin Observatory as a drifting optimal focus that varies on hour timescales within a night, and from night-to-night, as observed during AOS stability tests with fixed telescope pointing, dedicated LUT tests, and the pre-LSST observations. There is also some evidence for anomalous focus offsets and tilts that occur around filter changes that are repeatable on hour-timescales, but that are non-repeatable on longer timescales. While there appears to be some (expected) correlation of the focus offsets with the telescope truss temperature, a more complete understanding of the physical origin for these effects is still under investigation. Initial analysis of the observations acquired in recent weeks to refine the combined elevation + rotator angle LUT shows no strong dependence on elevation or rotator angle, but a larger scatter than would be expected from the statistical uncertainty of the wavefront measurements alone. Continued analysis of both dedicated engineering data and survey-mode observations sampling a range of environmental conditions will likely be needed to refine the LUT performance. Targeted on-sky engineering tests are planned for the next week.
In parallel, the team is continuing efforts to optimize the summit computing infrastructure and wavefront measurement algorithms to compute and apply AOS closed loop corrections as quickly as possible. Currently, the typical correction lag is 3 visits for pre-LSST observations, meaning that corrections determined from wavefront measurements on visit index N have been applied prior to starting the exposure for visit index N + 3. There is active testing to achieve the design specification of N + 2 corrections, and thus provide more responsive optical adjustments during survey-mode observations.
Sustained pre-LSST observations with the Feature Based Scheduler (FBS) continued as the primary on-sky activity this week. As a near-term mitigation for the effects described above, the team has inserted AOS closed loop iterations to refocus the optics after filter changes. The team experimented with different values for the gain of these corrections over the past week to optimize for robust and efficient optical convergence. Currently, these inserted AOS closed loop iterations add approximately 3 minutes overhead for each filter change.
Intra-night filter changes to measure the colors of transient and variable objects, as well as regular observations of the LSST Deep Drilling Fields (DDFs), which typically involve large slews, are both considered to be essential elements of the LSST survey strategy. These requirements are a current stressor for the system. The team is exploring variants of the baseline LSST cadence that would deliver reliable image quality in the near term, provide acceptable temporal sampling in multiple filters, and support ongoing refinement of the AOS open-loop LUTs. On the timescale of the next week, the team plans to test FBS configurations with constraints on the telescope elevation and camera rotator angle range to reduce stress on the AOS, working towards a candidate survey strategy that could consistently achieve the start-of-LSST criteria (RTN-093) given the current performance capabilities. On a longer timescale, further variants on the LSST survey strategy implementation could be explored, for example, minimizing the number of intra-night filter changes, or avoiding simultaneous large slews and filter changes.
A second area of emphasis is increasing the effective survey speed, both through gains in the system availability for sustained survey observations, and the rate of acquiring visits during regular nighttime operations (i.e., open shutter efficiency as determined by slew and settle times together with survey strategy). This week, updates for the timing controls and interactions of telescope subsystems helped reduce the median gap time between visits from ~10 seconds to ~8 seconds during the pre-LSST observations. The sustained pre-LSST observations of the past 2 weeks have also provided a more representative view of the system availability. On a typical recent night, the fraction of active observing time has been 80-90% when not limited by the environmental conditions, and there have been some nights of continuous, uninterrupted pre-LSST operations. The team has also been improving the start-of-night efficiency this week by reducing the number of iterations for the initial optical focus and alignment and by evaluating an earlier start to initial focus and alignment during evening twilight.
A filter swap to remove y and install u occurred as planned on 13 January, such that the currently available filter set is ugriz.