2024-02-21

Combined Bias+Dark

For PanSTARRS, the dark is stored as a multi-extension fits (one extension per amplifier), with the per-extension data stored as a (590, 598, 4) data cube.  The first two dimensions are the standard x/y coordinate on the amplifier, with the third dimension containing the dark components.  These components are then used to construct the dark model as:

model = component0 + CELL.EXPOSURE * component1 + (CHIP.TEMPERATURE * CELL.EXPOSURE) * component2 + (CHIP.TEMPERATURE * CHIP.TEMPERATURE * CELL.EXPOSURE) * component3

where component0 is essentially the bias frame, component1 is a standard exposure-time dependent dark frame, and the remaining components represent higher order temperature dependent terms.  The values listed in capital letters are header keywords, accessible from the appropriate header structure (CELL => amplifier, CHIP => detector).  This structure is flexible, with the configuration defining the components written to the calibration as an additional binary table (PS_DARK).

A similar model can be used for LSSTCam, although the new ISR Task assumes the bias correction is done in ADU and the dark in electrons, so such a calibration would have an extra dependence on the gain.

Persistence Correction

Persistence on GPC1 was significantly worse than we are expecting on the LSST Camera, with both on-frame and subsequent-frame components.  Figure 10 shows these two components.  During the exposure, the saturated core of the star changes the pixel, resulting in trapped charge that is not transferred out with the rest of the signal.  During the read out, this trapped charge leaks out, resulting in a "trail" pointing along the vertical axis away from the readout amplifiers.  On subsequent exposures, the pixels that fall along the vertical axis below the saturated source have previously been shifted through the contaminated region, and so show evidence of previously saturated star in a "burn" that extend towards the readout amplifier.

Correction of these features was done via the burntool program.  Sources are identified on the frame (with no sky/background subtraction), with saturated sources sent to the next step.  A power-law fit is applied to correct the "upward" pointing trails, and if a significant fit is found, the saturated core is saved in an output table containing the (x, y) position as well as the PONTIME ("power-on-time") of the camera.  This table is passed to subsequent exposures, and after correcting that exposure's trails, the historical information from the input table is used to check the locations for a "downward" exponential fit to correct the burns.  All positions that have valid fits are retained for the output table, along with all positions with PONTIME values  younger than a configurable expiration time.  The retention of positions without valid fits is important, because the fit quality is highly dependent on the astronomical scene:  A given exposure may have a bright source located near the top or bottom edge of the cell, and there may be additional astronomical sources that fall along the path of the trails/burns.

During operations, these tables are constructed on the fly during observing as part of the transfer from the summit, such that exposure (in full-focal plane terms) is not available for further processing until FITS files for all detectors have been downloaded and burntool successfully run on each.  This can create a backlog, as the burntool processes for a single detector are forced to run serially, and the execution time depends on source density.  Subsequent processing can use the tables generated the first time, so although there is a additional "calibration" that is needed for reprocessing, the serial/time issue should not be a problem.





  • No labels