Source code for hispecdrp.primitives.bias_subtraction.bias_primitive
from ..base_primitive import BasePrimitive
from ...calibrations import BiasSelector
from ... import datamodels
from ...utils.parallelization_utils import numba_thread_scope
from numba import njit, prange
import numpy as np
import logging
logger = logging.getLogger(__name__)
__all__ = ['BiasSubtractionPrimitive']
[docs]
class BiasSubtractionPrimitive(BasePrimitive):
"""
Subtract the bias level from the input data.
This is the base signal from which accumulated charges are measured.
The bias voltage is to prevent negative values to be read given the
unsigned nature of the values produced by the ADC. Bias subtraction
is useful to set the y-intercept term of the UTR data close to 0.
This primitive applies the appropriate bias calibration to the
input data product. The bias frame is subtracted from the science data,
the associated uncertainties are propagated by adding errors in
quadrature, and data quality (DQ) flags are combined using a bitwise OR.
Parameters
----------
dataset : str or RampModel
Input file path or data model to process. Currently, this is expected
to be a raw (ramp) data product.
suffix : str or None, optional
Suffix to append to the output filename. Default is ``'biassub'``.
max_cores : int, optional
Maximum number of CPU cores to use for parallel processing. Default is 1 (no parallelization).
Calibrations
------------
bias : BiasModel
- selector: :class:`~hispecdrp.calibrations.selectors.BiasSelector`
Returns
-------
output : DataModel
Bias-subtracted data model with updated error and data quality
arrays.
"""
PARAMETERS = {
'suffix' : 'biassub',
}
CALIBRATIONS = {
'bias': {
'selector': BiasSelector,
'selector_kwargs': {}
}
}
def _perform(self):
# Load the input model
input_model = datamodels.open(self.dataset)
# Get the bias model
bias, _ = self.get_calibration(input_model, 'bias')
bias_model = datamodels.open(bias)
# Output
result = self.get_output_model(input_model)
if not hasattr(result, 'DATA'):
raise Exception("DATA extension not found in input model for bias subtraction.")
if not hasattr(result, 'DQ'):
raise Exception("DQ extension not found in input model for bias subtraction.")
# Log
logger.info(f"Applying bias correction with {bias} to {input_model}")
if result.data.dtype != np.float32:
logger.info(f"Converting UTR data from {result.data.dtype} to float32")
result.data = result.data.astype(np.float32)
# Perform bias subtraction
max_cores = self.max_cores or 1
_bias_subtraction(
result.data, result.dq,
bias_model.bias, bias_model.dq,
max_cores=max_cores
)
# Output
return dict(output=result, bias=bias)
def _bias_subtraction(
input_data : np.ndarray,
input_dq : np.ndarray,
bias_data : np.ndarray,
bias_dq : np.ndarray,
max_cores : int = 1
):
"""
Subtract the bias from the ramps.
Parameters
----------
input_data
Input ramp data with shape (Nreads, Ny, Nx).
input_dq
Input DQ array with shape (Ny, Nx).
bias_data
Bias data with shape (Ny, Nx).
bias_dq
Bias DQ array with shape (Ny, Nx).
max_cores
Maximum number of CPU cores to use.
"""
with numba_thread_scope(max_cores):
_bias_subtraction_numba(input_data, input_dq, bias_data, bias_dq)
return input_data, input_dq
@njit(nogil=True, parallel=True, cache=True)
def _bias_subtraction_numba(data : np.ndarray, dq : np.ndarray, bias : np.ndarray, bias_dq : np.ndarray):
n_read, ny, nx = data.shape
for j in prange(ny):
for i in range(nx):
data[:, j, i] = data[:, j, i] - bias[j, i]
dq[j, i] |= bias_dq[j, i]