In the current lvmdrp, ra's and dec's of the individual fibers are given in the SFrame data, so presumably either we are guessing that the PA is 0, or we have some other way of determining this. If we know the PA then we should create and populate the keyword. But even if is a guess, we should create the keyword (possibly with a warning in the header), and populate it with our best guess value, as this keyword is often used to make a WCS for the data in order to create images, and forcing every downstream program to account for this is time consuming.
ndrory
commented
1 week ago
An example of this is exposure 3471.
In the current lvmdrp, ra's and dec's of the individual fibers are given in the SFrame data, so presumably either we are guessing that the PA is 0, or we have some other way of determining this. If we know the PA then we should create and populate the keyword. But even if is a guess, we should create the keyword (possibly with a warning in the header), and populate it with our best guess value, as this keyword is often used to make a WCS for the data in order to create images, and forcing every downstream program to account for this is time consuming.