Table of contents:
Correlated Double Sampling (CDS)
The WIRCam images are correlated double sampling (CDS) images for the most part.
The LowOH1 filters images are now Fowler=4 images since January 19 2007.
Near-IR imagers are usually shutter-less instruments which means that
photons are detected immediately after the detector is RESET until it
saturates. A CDS image is constructed by subtracting a REF (a read of the
detector setting the start of the integration time, immediately after RESET)
from the RAW (a read of the detector after the integration time) - see figure
1.
Figure 1. Correlated Double Sampling (CDS)
Caption: A WIRCam image is RAW - REF.
Detector Architecture
Each WIRCam detector is read in parallel using 32 amplifiers (see figure 2).
Two electronic SDSU controllers are needed to read the whole 4-chip mosaic. A
controller for the east chips (part #54 and #60), a second controller for the
west chips (part #77 and #52). Each controller is equipped with 4 video boards
with 8 amplifiers per video boards. The numbers to keep in mind are 32
amplifiers per detector grouped in sets of 8 amplifiers per video board.
Figure 2. The WIRCam detector readout
Caption: Each detector is read in parallel using 32 amplifiers split on
4 video boards. Each amplifier is 64x2048 pixels.
Raw Images Construction
Early on, it was decided that only the CDS image be saved, the two
independant reads are lost. This was to save disk space and mostly to keep
things relatively simple for processing and displaying of raw images. In fact,
there is no compelling reason to keep both reads as even the non-linearity
correction can be effectively applied without them, with an iterative approach.
Two processing steps are performed on each CDS image before saving them as
16-bit images with the o.fits suffix (the o.fits are called our "raw" images
everywhere else in this documentation):
- Flag the pixels that are badly saturated to 65535. This is done by comparing
the REF and RAW individually to a reference "saturation map" that we produce in
engineering mode by looking at the pixels values on a single RAW/REF close to
saturation.
- Add a "chipbias", currently set at 7000 adu, before saving in 16 bit format.
This is to prevent negative values from wrapping to 65000 adu. It was found that
negative values down to ~5000 adu existed so we picked 7000 adu as a safe value
to make sure we never have wrapping.
|