The JWST NIRISS detector is read out non-destructively using 4 outputs for the full-array and 1 output for the subarrays.
The NIRISS detector is read out in MULTIACCUM mode, where charge is continuously measured "non-destructively" rather than transferred across pixels as in CCDs. The detector is reset at the beginning of every NIRISS integration. The full specification of a readout involves defining the readout pattern and readout format.
The readout pattern denotes the (temporal) sampling cadence for non-destructive "up-the-ramp" reads of the detector. The readout pattern determines whether each frame is saved or whether they are coadded. There are two NIRISS detector readout patterns available, NISRAPID, where each frame is saved and is considered a "group", and NIS, where four frames are coadded within each group. NISRAPID will primarily be used for bright objects or observations requiring a high cadence, where data volume limits is not a concern, and mostly when using a subarray for readout. NIS is preferred for longer integrations of faint sources. An integration is specified by the number of groups between detector resets. There can be multiple integrations in an exposure.
The readout format denotes the spatial region of the detector to be read and is specified by the detector subarray. Fewer pixels are read out when using a NIRISS subarray compared with full-frame readout, which results in a shorter frame time. The full-frame readout mode is the only option available for WFSS and Imaging science observations. The SOSS and AMI modes mostly use the subarray readout formats.
A schematic of the NIRISS detector array is illustrated in Figure 1. In full-frame mode, four parallel output channels, 512x2048 pixels, are read out simultaneously. Subarrays are read through a single output channel.