The current release of STIPS appears to treat errors from dark current as proportional to the integration time, implying that they are caused by uncertainty in the dark current (thus dark current subtraction errors) rather than inflated Poisson noise created by the dark current.
A sample 100 second simulation in Roman WFI F062 with just dark current noise shows a standard deviation of 0.37 electrons, or 0.0037 electrons/sec. Given the predicted dark current is 0.028 electrons/sec (https://roman.gsfc.nasa.gov/science/WFI_technical.html), it seems somewhat surprising that dark current subtraction leaves pixel-to-pixel spatial residuals of 13% of the actual dark current. Is this correct?
More concerning is that, even accepting the 13% systematic error in dark current subtraction, the expected 2.8 electrons of dark current should create an additional sqrt(2.8)=1.7 electrons of Poisson noise. At this exposure time, the random Poisson error is a factor of 4.5 higher than the systematic error, and the two error sources would equate around 2000 seconds.
Perhaps this needs to be split into two issues - confirming what appears to be a large dark noise residual error (again, pixel-by-pixel, which should be smaller than global), and adding a dark noise Poisson error.
The current release of STIPS appears to treat errors from dark current as proportional to the integration time, implying that they are caused by uncertainty in the dark current (thus dark current subtraction errors) rather than inflated Poisson noise created by the dark current.
A sample 100 second simulation in Roman WFI F062 with just dark current noise shows a standard deviation of 0.37 electrons, or 0.0037 electrons/sec. Given the predicted dark current is 0.028 electrons/sec (https://roman.gsfc.nasa.gov/science/WFI_technical.html), it seems somewhat surprising that dark current subtraction leaves pixel-to-pixel spatial residuals of 13% of the actual dark current. Is this correct?
More concerning is that, even accepting the 13% systematic error in dark current subtraction, the expected 2.8 electrons of dark current should create an additional sqrt(2.8)=1.7 electrons of Poisson noise. At this exposure time, the random Poisson error is a factor of 4.5 higher than the systematic error, and the two error sources would equate around 2000 seconds.
Perhaps this needs to be split into two issues - confirming what appears to be a large dark noise residual error (again, pixel-by-pixel, which should be smaller than global), and adding a dark noise Poisson error.