Differences from the Commissioning Pipeline

Much of the processing done by BANZAI-NRES is similar to the original pipeline used for commissioning NRES. We note a few of the differences here.

The main difference is that the commissioning pipeline (available here) is written in IDL, while BANZAI-NRES is primarily written in Python.

Both pipelines perform optimal extraction as per Horne 1986 DOI: 10.1086/131801. The profile is estimated by fitting polynomials to the quartz lamp flats in the commissioning pipeline, iterating to reject cosmic rays, while BANZAI-NRES uses the lamp flat pixels directly only normalizing in the spatial direction. Cosmic ray rejection is an early planned improvement for BANZAI-NRES.

BANZAI-NRES currently only uses the ThAr arc frames taken in during the day to obtain its wavelength solution. The commissioning pipeline applies an additional shift to the wavelength calibration derived from the simultaneous ThAr exposure from the calibration fiber.

BANZAI-NRES and the commissioning pipeline estimate the continuum in different ways. The commissioning pipeline subtracted the an estimate of the blaze function (using a polynomial fit from the stacked quartz lamp flat) and then applied a high pass filter to the science spectrum before calculating the radial velocity. BANZAI-NRES divides out an estimate of the blaze function from the stacked quartz lamp flat. Any residual continuum is estimated by smoothing using a median filter after masking likely absorption lines and is divided out. The continuum normalized spectrum including both steps in included in the ‘normflux’ column of the extracted-spectrum data product.

The commissioning pipeline measured the radial velocity using ~12 blocks across each order. We tested this approach but did not find it improved our radial velocity precision in the current version of BANZAI-NRES. We plan to reexamine this as we continue to improve the radial velocity precision of BANZAI-NRES.

Currently, the commissioning pipeline does not estimate the stellar parameters of a given target. BANZAI-NRES estimates the stellar classification by cross-correlating with the PHOENIX stellar atmosphere models from Husser et al. 2013, DOI: 10.1051/0004-6361/201219058.

BANZAI-NRES propagates an estimate of the formal uncertainties from all of the data processing stages and include these in the output data products. These are used as weights in the cross correlation function to measure the radial velocity. We adopt an optimal weighting scheme based on Zackay & Ofek 2017 DOI: 10.3847/1538-4357/836/2/187.

Data Product Differences from the Commissioning Pipeline

The main difference between the 1-D data products produced by the commissioning pipeline and BANZAI-NRES is that we have formatted the data into a FITS Binary table instead of using multiple FITS extensions. This makes the association between the different pieces of data more explicit and alleviates some issues when compressing the files with fpack.

Below are the extension names from the FITS extensions from the previous data products and how they map to the columns in the extracted spectra produced by BANZAI-NRES. Rather than splitting the ThAr fiber and the science fiber into separate extensions, BANZAI-NRES interleaves the orders in the FITS binary table.

  • ‘SPECRAW’ Extension: ‘flux’ column.

  • ‘SPECFLAT’ Extension: ‘normflux’ column is the closest analog. The ‘normflux’ column includes extra continuum normalization compared to the commissioning pipeline.

  • ‘SPECBLAZE’ Extension: This is the blaze subtracted spectrum. We do not provide this explicitly, but BANZAI-NRES does provide the ‘blaze’ column and the ‘flux’ columns to calculate this.

  • ‘THARRAW’ Extension: This is the flux of the ThAr fiber orders. These are interleaved with the science fiber in the ‘flux’ column .

  • ‘THARFLAT’ Extension: Flat-field divided ThAr fiber spectrum. These are interleaved with the science fiber in the ‘normflux’ column.

  • ‘WAVESPEC’ Extension: Wavelengths solutions per pixel for the science fiber. These are stored in the ‘wavelength’ column in BANZAI-NRES data products. Note that the commissioning pipeline stored these in values in nanometers. BANZAI-NRES stores wavelengths in Angstroms.

  • ‘WAVETHAR’ Extension: Wavelengths solutions per pixel for the calibration ThAr fiber. These are stored in the ‘wavelength’ column in BANZAI-NRES data products and are interleaved with the science spectrum orders. Note that the commissioning pipeline stored these values in nanometers. BANZAI-NRES stores wavelengths in Angstroms.

  • ‘SPECXCOR’ Extension: Summed cross correlation function (CCF) for a range of velocities. The velocity for a given pixel is saved as the WCS in the header. BANZAI-NRES stores its cross-correlation results in the ‘CCF’ extension as a FITS binary table. Rather than saving the summed CCF, we store the CCF per order and corresponding velocity coordinates for clarity for the user.

  • ‘RVBLOCKFIT’ Extension: Per-order/per-block RVs in a FITS Binary table. As we do not calculate radial velocities using blocks, BANZAI-NRES does not include this extension.