Current status of PFS
(As of 2025/07/30)
Status of instrument
All hardware is now online
Fiber throughput variation issue
We found that the relative fiber throughput changed in a time-scale of a few hours during a night due to the air gap between the fiber terminations in the fiber connectors. We decided to apply index-matching gel to fill the gap. Some tests in previous engineering runs show that the throughput stability has been significantly improved and the total throughput itself increased by several %.
Flux non-uniformity across FoV
In past engineering runs, we found that for some exposures, especially those taken at low elevations (EL <~ 35 deg) and high elevations (EL >~ 70 deg), the non-uniformity of flux across the FoV was not small, causing the worst fibers to suffer substantial flux loss. This is possibly due to inaccuracies in fiber positioning, as well as the effects of field acquisition. We are continuing to improve the performance during engineering nights.
Status of DRP development
PFS DRP Tutorial
PFS DRP Tutorial web page has been launched.
NIR data processing
The team is currently working on processing H4RG detector data while detrending various features incorporating the pixel-to-pixel and ramp-to-ramp analysis. We know that the strength and duration of the persistence depends strongly on the detectors, where the analysis is extremely difficult if the pixel traps many electrons in a level of saturation. We operate to avoid saturating the H4RG detectors and observe as less bright objects as possible.
Flux calibration and coadding exposures
As mentioned above, due to the issue on significant flux loss and variation, there is a large uncertainty in flux calibration in a part of observation. The current performance of flux calibration is about several % in rms in good cases. Partly due to this uncertainty and the lack of good/stable quality of data after applying index matching gel, we have not yet verified the coadding performance of multiple exposures and therefore the long integration for faint targets.
Scattered light correction
We know that there exists significant scattered light of several percent in amplitude. Currently, the intersection of fiber profile in the spatial direction (fiberProfiles) are generated without taking into account this effect and the wing of the profile is actually overestimated at a level of ~1%, therefore very bright fibers affect the spectral extraction of the neighboring fibers. We are currently planning to model the scattering kernel and apply in the processing, where we need to generate fiberProfiles in consistent way and apply both to the data in the extraction.
Sky subtraction accuracy
According to analysis of sky fibers, the chi distribution is close to the normal distribution N(0,1) in optical data. The quality does not significantly depend on the flux.
Wavelength calibration
The mapping between the possition of a fiber and the wavelength and the position on the detector (detectorMaps) is calibrated using quartz and arc lamps data. In order to correct the small change due to various changes of instrument condition, we adjust it using sky lines in each exposure. The typical uncerntainty of detectorMaps is at a level of 0.05 pixel in both spatial and wavelength direction.
Cosmic ray rejection
Currently it is hard to detect all CRs in a single exposure of CCDs. Unmasked CRs affect data reduction, particularly sky subtraction. For now, we use multiple frames (at least 2 frames) to detect almost all CRs and process further. Due to this requirement, at least 2 exposures must be taken for the same fiber design. The figures below show the expected degradtion on SNR due to splitting the exposure time (the left is for continuum and the right is for emission line).
Performance verification
Fiber configuration accuracy
The fiber configuration is carried out by comparing target and fiber positions on PFI using images taken by MCS in an iterative way. Based on results in previous engineering runs, the configuration performance achieves an accuracy of < 10 μm (75%-tile). The configuration time is approximately 120 sec in the best case. The final positional accuracy, including the uncertainty from coordinate transformation and field acquisition, have been measured using a method called raster scans and the typical accuracy is < 30 μm (95%-tile). The figure below shows the average residual for each visit.
System throughput
Based on data of bright Calspec stars and F-type stars for the flux calibration, we confirmed that the system throughput is basically consistent with the expectation integrating observed/laboratory data of each sub-system component, which is now incoporated in the current version of ETC. Note that the enhancement of throughput thanks to the index-matching gel is not yet included in ETC.
Signal-to-Noise Ratio
Based on the current capability of DRP, we confirmed that the signal-to-noise ratio against relatively bright galaxies is mostly consistent with the expectation of ETC for a single exposure and/or integration up to ~1 hour in optical arms. Regarding longer integrations of fainter objects, we have not verified that the signal-to-noise increases as coadding multiple exposures, since there is still uncertainty in the flux calibration and coadding process. The left figure below shows the SNR for 20.5 ABmag galaxies for a single 900 sec exposure and the right figure shows it for adding four 900 sec exposures. The vertical dashed line represents the expectation of ETC for similar observing conditions.
