Stage 1 steps

The stage 1 steps perform basic detector-level corrections at the group level, followed by ramp fitting. Below we detail the workings of several steps that you may find useful for improving the quality of your reduction beyond the default JWST stage 1. For a complete list of ExoTiC-MIRI steps and their workings, see the API.

The following examples assume you have set up the pipelines and loaded in an _uncal.fits data segment. For example:

import os

os.environ["CRDS_SERVER_URL"] = "https://jwst-crds.stsci.edu"
os.environ["CRDS_PATH"] = "path/to/your/local/crds_dir"
os.environ["CRDS_CONTEXT"] = "jwst_1100.pmap"

from jwst import datamodels
from jwst.pipeline import calwebb_detector1

# Load data segment.
proc = datamodels.RampModel("path/to/your/data/jwxyz-segnnn_mirimage_uncal.fits")

Drop groups

The MIRI detector is adversely affected by several effects such as Reset Switch Charge Decay and dragging down of the final frame. As such, you may wish to mask certain groups from being included in subsequent steps, such as jump detection or ramp fitting. This can be achieved with DropGroupsStep.

from exotic_miri.stage_1 import DropGroupsStep


custom_drop_groups = DropGroupsStep()
proc = custom_drop_groups.call(proc, drop_groups=[0, 1, 2, 3, 4, -1])

Here, we have masked the first 5 groups and the final group. The number of groups to mask will depend primarily on the brightness of your target, but in general at least the final group should always be masked (assuming FASTR1 readout mode).

Group-level background subtraction

If you would like to experiment with subtracting the background/sky at the group level, to clean up the data before ramp fitting, then you can make use of GroupBackgroundSubtractStep.

from exotic_miri.stage_1 import GroupBackgroundSubtractStep


custom_group_bkg_subtract = GroupBackgroundSubtractStep()
proc = custom_group_bkg_subtract.call(proc, method="row_wise")

Here, we have used the default background regions either side of the spectral trace and applied a row-wise background subtraction. You can adjust the method and background region as required.

Custom gain

The default gain value for MIRI may not be the most accurate. This can have effects on other steps which require the gain to estimate statistical information. You can create a custom gain model with SetCustomGain.

from exotic_miri.reference import SetCustomGain


custom_set_gain = SetCustomGain()
stsci_jump = calwebb_detector1.jump_step.JumpStep()
stsci_ramp_fit = calwebb_detector1.ramp_fit_step.RampFitStep()
stsci_gain_scale = calwebb_detector1.gain_scale_step.GainScaleStep()

gain_model = custom_set_gain.call(proc, gain_value=3.1)
proc = stsci_jump.call(proc, override_gain=gain_model)
_, proc = stsci_ramp_fit.call(proc, override_gain=gain_model)
proc = stsci_gain_scale.call(proc, override_gain=gain_model)

Here we have created a custom gain model with a value of 3.1 (see Bell et al. 2023) and then passed this to the jump detection, ramp fitting, and gain scale steps.

Custom linearity correction

Determining a linearity correction, the model which accounts for the decrease in gain as pixels become increasingly full, is challenging for MIRI given all the nuances to this Si:As detector. It may be worth generating a custom linearity correction which is self-calibrated from your dataset, if you have a sufficient number of groups, using SetCustomLinearity.

from exotic_miri.reference import SetCustomLinearity


custom_set_linearity = SetCustomLinearity()
stsci_linearity = calwebb_detector1.linearity_step.LinearityStep()

linearity_model = custom_linearity.call(proc, group_idx_start_fit=5, group_idx_end_fit=40,
                                        group_idx_start_derive=5, group_idx_end_derive=100,
                                        row_idx_start_used=350, row_idx_end_used=386)
proc = stsci_linearity.call(proc, override_linearity=linearity_model)

This correction involves extrapolating a linear fit to an assumed linear, or well-behaved section of the ramps. In this case, this is between groups 5 and 40. A polynomial is then fit to the ramps for data between groups 5 and 100 and for rows 350 to 386. The polynomial has the constant- and linear-term coefficients fixed at 0 and 1, respectively. This polynomial may then serve as the correction for all ramps in your dataset.