MakePsfMatchedWarpTask¶
Convolve a direct warp by a kernel to produce a PSF-matched warp, whose PSF matches a desired target PSF.
This task separates the warp into a set of non-overlapping polygons corresponding to each detector.
The PSF-matching is done on each detector separately.
The subtask ModelPsfMatchTask is responsible for the PSF-Matching, and its config is accessed via config.psfMatch.
The optimal configuration depends on aspects of dataset: the pixel scale, average PSF FWHM and dimensions of the PSF kernel. These configs include the requested model PSF, the matching kernel size, padding of the science PSF thumbnail and spatial sampling frequency of the PSF.
Python API summary¶
from lsst.drp.tasks.make_psf_matched_warp import MakePsfMatchedWarpTask
-
class
MakePsfMatchedWarpTask(**kwargs) Base class for all pipeline tasks
...
- attributeconfig
Access configuration fields and retargetable subtasks.
See also
See the MakePsfMatchedWarpTask API reference for complete details.
Retargetable subtasks¶
modelPsf¶
- Default
lsst.meas.algorithms.gaussianPsfFactory.applyWrapper- Field type
Model Psf factory
psfMatch¶
- Default
lsst.ip.diffim.modelPsfMatch.ModelPsfMatchTask- Field type
Task to warp and PSF-match calexp
Configuration fields¶
connections¶
- Data type
lsst.pipe.base.config.MakePsfMatchedWarpConfigConnections- Field type
Configurations describing the connections of the PipelineTask to datatypes
saveLogOutput¶
Flag to enable/disable saving of log output for a task, enabled by default.
In Depth¶
Config Guidelines¶
The user must specify the size of the model PSF to which to match by setting config.modelPsf.defaultFwhm in units of pixels.
The appropriate values depend on science case.
In general, for a set of input images, this config should equal the FWHM of the visit with the worst seeing.
The smallest it should be set to is the median FWHM.
The defaults of the other config options offer a reasonable starting point.
The following list presents the most common problems that arise from a misconfigured ModelPsfMatchTask and corresponding solutions.
All assume the default Alard-Lupton kernel, with configs accessed via config.psfMatch.kernel['AL'].
Each item in the list is formatted as Problem, Explanation. Solution.
Troubleshooting PSF-Matching Configuration¶
Matched PSFs look boxy¶
The matching kernel is too small.
Solution¶
Increase the matching kernel size. For example:
config.psfMatch.kernel['AL'].kernelSize=27
# default 21
Note that increasing the kernel size also increases runtime.
Matched PSFs look ugly (dipoles, quadrupoles, donuts)¶
Unable to find good solution for matching kernel.
Solution¶
Provide the matcher with more data by either increasing the spatial sampling by decreasing the spatial cell size.
config.psfMatch.kernel['AL'].sizeCellX = 64
# default 128
config.psfMatch.kernel['AL'].sizeCellY = 64
# default 128
or increasing the padding around the Science PSF, for example:
config.psfMatch.autoPadPsfTo=1.6 # default 1.4
Increasing autoPadPsfTo increases the minimum ratio of input PSF dimensions to the matching kernel dimensions, thus increasing the number of pixels available to fit after convolving the PSF with the matching kernel.
Optionally, for debugging the effects of padding, the level of padding may be manually controlled by setting turning off the automatic padding and setting the number of pixels by which to pad the PSF:
config.psfMatch.doAutoPadPsf = False
# default True
config.psfMatch.padPsfBy = 6
# pixels. default 0
Ripple Noise Pattern¶
Matching a large PSF to a smaller PSF produces a telltale noise pattern which looks like ripples or a brain.
Solution¶
Increase the size of the requested model PSF. For example:
config.modelPsf.defaultFwhm = 11 # Gaussian sigma in units of pixels.
High frequency (sometimes checkered) noise¶
The matching basis functions are too small.
Solution¶
Increase the width of the Gaussian basis functions. For example:
config.psfMatch.kernel['AL'].alardSigGauss= [1.5, 3.0, 6.0] # from default [0.7, 1.5, 3.0]