5.6	WFC3 IR Readout Formats

5.6.1 Full-Frame Readouts and Reference Pixels

5.6.2 Subarrays

5.6.1 Full-Frame Readouts and Reference Pixels

The WFC3 IR detector contains 1024×1024 square pixels of 18×18 micron physical size. The detector is divided into four quadrants of 512×512 pixels, each of which is read out independently from its outer corner, as illustrated in Figure 5.16. The outermost rows are read first, proceeding along each row from the outermost column to the horizontal mid-point of the detector, and then continuing inwards on subsequent rows to the vertical mid-point.

Figure 5.16: Schematic layout of the WFC3 IR detector. The long (red) and short (blue) arrows indicate the direction of the fast and slow multiplexer clocking, respectively. In contrast to CCD “bucket-brigade” image-shifting to the output amplifier, the IR detector pixels are selected for readout in a raster pattern by multiplexer circuits.

A major effort has been made to eliminate both the amplifier glow and bias drifts that have affected the NICMOS detectors.

To eliminate the amplifier glow entirely, WFC3 uses external amplifiers located in the immediate vicinity of the detector, rather than those directly on the multiplexer (which are also present, but are not activated in the WFC3 implementation).

In regard to bias drifts, the WFC3 IR class of detectors is the first to use reference pixels, configured as follows (see Figure 5.17). Of the 1024×1024 pixels, only the inner 1014×1014 pixels are light-sensitive. The five outer rows and columns of pixels all around the array use fixed capacitances to provide constant-voltage reference values. There are actually two types of reference pixels: (1) the pixels on the outermost columns/rows are connected to capacitors located outside of the unit cells. Their values follow a 4× periodic pattern, providing 4 sequentially increasing voltage levels all within the range of the detector output signal; (2) the 4 inner rows/columns are instead connected to capacitors created within their unit cells. These on-board capacitors are identical by design and all provide nearly the same reference signal. The current version of the WFC3/IR data reduction pipeline uses only the inner reference pixels, as they provide a more robust statistical estimate of the variable detector bias.

The reference pixels track the low-frequency drift of the readout electronics and efficiently remove the “pedestal” variations that affected, for example, NICMOS. Analysis of ground test data has shown that the reference pixel signal is also sensitive to the detector temperature and may therefore be used to assess the expected level of dark current during an exposure, independently from a reading of the detector temperature itself. Actual on-orbit experience indicates that detector temperature is very stable.

Full-frame exposures result in one raw 1024×1024 pixel image for each readout, which includes the 5 rows and columns of reference pixels on the periphery. After calibration, the reference pixels are trimmed off, leaving only the 1014×1014 arrays of light-gathering pixels.

Figure 5.17: Schematic layout of the active pixels (dark shading) and of the reference pixels at a corner of the WFC3/IR detector. The color coding represents different values of the reference pixel capacitance.

5.6.2 Subarrays

The default IR exposure mode is to read out the entire detector. It is also possible, however, to read out only a portion of the detector. WFC3 IR subarrays are implemented in four user-selectable sizes: 64×64, 128×128, 256×256, and 512×512 pixels. All subarrays are centered on the detector with an equal number of pixels in each quadrant, using each of the 4 detector amplifiers to read the subarray pixels contained in its quadrant (as with full-frame readouts).

The 5-pixel wide bands of reference pixels that share rows or columns with the subarray are also included in subarray readouts. The reference pixels therefore come from the same detector rows and columns as the “live” portion of the subarray, with the 5×5 pixels at the subarray corners filled by the reference pixels at the corresponding corner of the detector.

Certain combinations of IR subarrays and sample sequences give rise to images containing a sudden low-level jump in the overall background level of the image. The cause of the artifact is under investigation. The use of IR subarrays is discussed in more detail in Section 7.4.4.