Keck-DataReductionPipelines / MosfireDRP

http://keck-datareductionpipelines.github.io/MosfireDRP
10 stars 13 forks source link

Fix 2D Wavelength Fitting #118

Open joshwalawender opened 6 years ago

joshwalawender commented 6 years ago

Several significant changes to the Wavelength code were made:

First, the function to determine the estimated offset between rows of pixels was changed to use the new find_shift method rather than the xcor and xcor_peak methods. This new method is similar, but works out to be about 2x faster when timed using the %timeit tool in ipython.

Second, that shift determination was constrained to be within 3 pixels of the neighboring line, rather than within 30 pixels of the starting line. This ensures a much more continuous wavelength solution in the Y direction.

Third, when fitting Chebyshev polynomials to the wavelength solution for each line, an extra step is added (the smooth_solution method) to median filter each Chebyshev coefficient in the Y direction to limit the changes in wavelength solution from pixel line to adjacent pixel line.

This appears to have fixed the problem in #115, though further testing to ensure that it works in all cases is still pending.