The Goodman Spectroscopic Pipeline is designed to be simple to use, however simple does not always is the best case for everyone, thus The Goodman Pipeline is also flexible.

Getting Help.

This manual is intended to be the prefered method to get help. However the quickest option is using -h or --help

redccd --help

Will print the list of arguments along with a quick explanation and default values.

It is the same for redspec

redspec --help

Observing Guidelines#

In order to be able to process your data with the Goodman Spectroscopic Pipeline you need to follow some guidelines, we do not intend to tell you how to do your science. Here are some basic hints.

  • Make sure you have a good observing plan as well as a good backup plan too.

  • Put special attention to the calibration files that are needed for the data that you are planning to obtain, for instance, you can process your spectroscopic data without bias because using overscan will give you a good enough approximation, but Imaging does not have overscan therefore you MUST obtain bias frames.

  • Keep a detailed log of things that happened while you were observing, mistakes that you made, exposures repeated, etc. An observing log is not an extraction of header information. Well, it can be, but it will be useless.

  • If you are unsure about the required steps to achieve your science goals ask your PI, not the support scientist, Her/His job is to assist you on how to get good quality data not what data you need in order to achieve your scientific goals.

For using the pipeline you don’t need to use any special file naming convention, in fact all the information is obtained from the headers. As of version 1.2.0 you need to use a reference lamp naming convention though. Not the file but the field that goes into OBJECT. It is actually very simple:

Convention names for comparison lamps#

Lamp name










Mercury Argon


Mercury Argon Neon


This is to ensure the pipeline is able to recognize them. This will no be the case in future versions but for now this is how it works.

Observing for Radial Velocity#

Radial velocity measurements are possible with the Goodman High Throughput Spectrograph but you have to be careful. A very detailed description of the procedures and what you can expect was prepared and is available here and here .

Please read it carefully so you don’t find any surprises when trying to reduce your data.

Prepare Data for Reduction#

If you did a good job preparing and doing the observation this should be an easy step, either way, keep in mind the following steps.

  • Remove all focus sequence.

  • Remove all target acquisition or test frames.

  • Using your observation’s log remove all unwanted files.

  • Make sure all data has the same gain (GAIN) and readout noise (RDNOISE)

  • Make sure all data has the same Region Of Interest or ROI (ROI).

The pipeline does not modify the original files unless there are problems with fits compliance, is never a bad idea to keep copies of your original data in a safe place.

Updating Keywords#

Since version 1.3.0 if your data is older than August 6, 2019, you will need to change the following keywords.

  • SLIT: Replace whitespaces with underscore, remove “ and all letters are uppercase. For instance 0.45" long slit becomes 0.45_LONG_SLIT.

  • GRATING: Grating’s lines/mm goes first and then the manufacturer. For instance. SYZY_400 becomes 400_SYZY.

  • WAVMODE: Replace whitespace with underscore and all letters are capitalized. For instance. 400 m1 becomes 400_M1.

  • INSTRUME: Instead of using the classical keywords ‘Goodman Spectro’ and ‘Goodman Imaging’, the AEON standard keywords ghts_red and ghts_blue will be used for spectroscopy, and ghts_red_imager and ghts_blue_imager for imaging. This is an exception of the upper case rule.


General rules are: Underscore is the only accepted separator. All letter must be upper case. Remove any character that need escaping.

Processing your 2D images#

It is the first step in the reduction process, the main tasks are listed below.

  • Create master bias

  • Create master flats

  • Apply Corrections:

    • Overscan

    • Trim image

    • Detect slit and trim out non-illuminated areas

    • Bias correction

    • Normalized flat field correction

    • Cosmic ray rejection


Some older Goodman HTS data has headers that are not FITS compliant, In such cases the headers are fixed and that is the only modification done to raw data.

The 2D images are initially reduced using redccd. You can simply move to the directory where your raw data is located and do:


Though you can modify the behavior in several ways.

Running redccd will create a directory called RED where it will put your reduced data. If you want to run it again it will prevent you from accidentally removing your already reduced data unless you use --auto-clean this will tell the pipeline to delete the RED directory and start over.

redccd --auto-clean

A summary of the most important command line arguments are presented below.

  • --cosmic <method> Let you select the method to do Cosmic Ray Removal.

  • --debug Show extended messages and plots of intermediate steps.

  • --flat-normalize <method> Let you select the method to do Flat Normalization.

  • --flat-norm-order <order> Set order for the model used to do Flat Normalization. Default 15.

  • --ignore-bias Ignores the existence or lack of BIAS data.

  • --ignore-flats Ignores the existence or lack of FLAT data.

  • --raw-path <path> Set the directory where the raw data is located, can be relative.

  • --red-path <path> Set the directory where the reduced data will be stored. Default RED.

  • --saturation <saturation> Set the saturation threshold in percentage. There is a table with all the readout modes and their values at which saturation is reached, then all the pixels exceeding that value are counted. If the percentage is larger that the threshold defined with this argument the flat is marked as saturated. The default value is 1 percent.

This is intended to work with spectroscopic and imaging data, that it is why the process is split in two.

Extracting the spectra#

After you are done Processing your 2D images it is time to extract the spectrum into a wavelength-calibrated 1D file.

The script is called redspec. The tasks performed are the following:

  • Classifies data and creates the match of OBJECT and COMP if it exists.

  • Identifies targets

  • Extracts targets

  • Saves extracted targets to 1D spectrum

  • Finds wavelength solution automatically

  • Linearizes data

  • Saves wavelength calibrated file

First you have to move into the RED directory, this is a precautionary method to avoid unintended deletion of your raw data. Then you can simply do:


And the pipeline should work its magic, though this might not be the desired behavior for every user or science case, we have implemented a set of command line arguments which are listed below.

  • --data-path <path> Folder were data to be processed is located. Default is current working directory.

  • --proc-path <path> Folder were processed data will be stored. Default is current working directory.

  • --search-pattern <pattern> Prefix for picking up files. Default is cfzst-. See File Prefixes.

  • --output-prefix <prefix> Prefix to be added to calibrated spectrum. Default is w-. See File Prefixes.

  • --extraction <method> Select the Extraction Methods. The only one implemented at the moment is fractional .

  • --fit-targets-with {moffat, gaussian} Model to fit peaks on spatial profile while searching for spectroscopic targets. Default moffat.

  • --target-min-width <target min width> Minimum profile width for fitting the spatial axis of spectroscopic targets. If fitting a Moffat it will be reflected as the FWHM attribute of the fitted model and if fitting a Gaussian it will be reflected as the STDDEV attribute of the Gaussian model.

  • --target-max-width <target max width> Maximum profile width for fitting the spatial axis of spectroscopic targets. If fitting a Moffat it will be reflected as the FWHM attribute of the fitted model and if fitting a Gaussian it will be reflected as the STDDEV attribute of the Gaussian model.

  • --reference-files <path> Folder where to find the reference lamps.

  • --debug Shows extended and more messages.

  • --debug-plot Shows plots of intermediate steps.

  • --max-targets <value> Maximum number of targets to detect in a single image. Default is 3.

  • --background-threshold <background threshold> Multiplier for background level used to discriminate usable targets. Default 3 times background level.

  • --save-plots Save plots.

  • --plot-results Show plots during execution.

The mathematical model used to define the wavelength solution is recorded in the header even though the data has been linearized for record purpose.

Description of custom keywords#

The pipeline adds several keywords to keep track of the process and in general for keeping important information available. The following table gives a description of all the keywords added by The Goodman Pipeline, though not all of them are added to all the images.

General Purpose Keywords#

These keywords are used for record purpose, except for GSP_FNAM which is used to keep track of the file name.

General purpose keywords, added to all images at the moment of the first read.#




Pipeline version.


Original file name, first read.


Parent file name.


Current file name.


Path from where the file was read.


Observing technique. Imaging or Spectroscopy.


Date of processing.


Overscan region.


Trim section.


Slit trim section. From slit-illuminated area.


Master bias file used.


Master flat file used.


Science target file name (for lamps only)


Reference lamp used to obtain wavelength solution


Master flat normalization method.


Cosmic ray rejection method.


RMS error of target trace


Extraction window at first column


First background extraction zone


Second background extraction zone


Wavelength solution RMS Error.


Number of points used to calculate RMS Error.


Number of points rejected from RMS Error Calculation.


Reference paper for DCR software (cosmic ray rejection).

Target Trace Model#

Keywords used to describe the model used to fit the target’s trace.#




Name of mathematical model from astropy’s modeling


Order of the model used.


Value of parameter c0.


Value of parameter c1.


Value of parameter c2. This goes on depending the order.

Non-linear wavelength solution#

Since writing non-linear wavelength solutions to the headers using the FITS standard (reference) is extremely complex and not necessarily well documented, we came up with the solution of simply describing the mathematical model from astropy’s modeling. This allows for maintaining the data untouched while keeping a reliable description of the wavelength solution.

The current implementation will work for writting any polinomial model. Reading is implemented only for Chebyshev1D which is the model by default.

Keywords used to describe a non-linear wavelength solution.#




Name of mathematical model from astropy’s modeling


Order of the model used.


Number of pixels.


Value of parameter c0.


Value of parameter c1.


Value of parameter c2. This goes on depending the order.

Combined Images#

Every image used in a combination of images is recorded in the header of the resulting one. The order does not have importance but most likely the header of the first one will be used.

The combination is made using the combine() method with the following parameters

  • method='median'

  • sigma_clip=True

  • sigma_clip_low_thresh=1.0

  • sigma_clip_high_thresh=1.0

At this moment these parameters are not user-configurable.

Keywords that list all the images used to produce a combined image.#




First image used to create combined.


Second image used to create combined.

Detected lines#

The reference lamp library maintains the lamps non-linearized and also they get a record of the pixel value and its equivalent in angstrom. In the following table a three-line lamp is shown.

Description of all the keywords used to list lines in lamps in Pixel and Angstrom.#




Pixel value for the first line detected.


Pixel value for the second line detected.


Pixel value for the third line detected.


Angstrom value for the first line detected.


Angstrom value for the second line detected.


Angstrom value for the third line detected.

Cosmic Ray Removal#


The parameters for either cosmic ray removal method are not fully understood neither tuned but they work for most common instrument configurations. If your extracted spectrum shows weird features, specially if you use a custom mode, the most likely culprit are the parameters of the method you chose. Please let us know.

The argument --cosmic <method> has four options but there are only two real methods.

default (default):

Different methods work different for different binning. So if <method> is set to default the pipeline will decide as follows:

dcr for binning 1x1

lacosmic for binning 2x2 and 3x3 though binning 3x3 has not being tested.


It was already said that this method work better for binning 1x1. More information can be found on Installing DCR. The disadvantages of this method is that is a program written in C and it is required to write the file to the disk, process it and read it back again. Still is faster than lacosmic.

The parameters for running dcr are written in a file called dcr.par a lookup table and a file generator have been implemented but you can parse custom parameters by placing a dcr.par file in a different directory and point it using --dcr-par-file <path>.


This is the preferred method for files with binning 2x2 and 3x3. This is the Astroscrappy’s implementation and is run with the default parameters. Future versions might include some parameter adjustment.


Skips the cosmic ray removal process.

Asymetric binnings have not been tested but the pipeline only takes in consideration the dispersion axis to decide. This does not mean that the spatial binning does not impact the performance of any of the methods, we just don’t know it yet.


The prefix c is added to all the comparison lamps, despite they not being affected by cosmic rays.

Flat Normalization#

There are three possible <method> (s) to do the normalization of master flats. For the method using a model the default model’s order is 15. It can be set using --flat-norm-order <order>.


Calculates the mean of the image using numpy’s mean() and divide the image by it.

simple (default):

Collapses the master flat across the spatial direction, fits a Chebyshev1D model of order 15 and divide the full image by this fitted model.


Fits a Chebyshev1D model to every line/column (dispersion axis) and divides it by the fitted model. This method takes too much to process and it has been left in the code for experimentation purposes only.

Extraction Methods#

The argument --extraction <method> has two options but only fractional is implemented.


Fractional pixel extraction differs from a simple and rough extraction in how it deals with the edges of the region. goodman_pipeline.core.core.extract_fractional_pixel()


Unfortunately this method has not been implemented yet.

File Prefixes#


Overscan is no longer performed by default. Since it caused a double bias level substraction. Fixed since release V1.3.3 19-04-2021.

There are several ways one can do this but we selected adding prefixes to the file name because is easier to add and also easy to filter using a terminal, for instance.

ls cfzst*fits

or in python

import glob

file_list = glob.glob('cfzst*fits')

So what does all those letter mean? Here is a table to explain it.

Characters and meaning of prefixes#




Overscan Correction Applied


Trim Correction Applied


Slit trim correction applied


Bias correction applied


Flat correction applied


Cosmic rays removed


Spectrum extracted to 1D


1D Spectrum wavelength calibrated

So, for an original file named file.fits:


Means the spectrum has been extracted to a 1D file but the file has not been flat fielded (f missing).

Ideally after running redccd the file should be named:


And after running redspec:


File Suffixes#

After extraction, suffixes may appear in new files created. There are two scenarios where this can happen:

  • More than one spectroscopic target for extraction. *target_X.

  • More than one comparison lamp. *ws_Y.

  • Both above

Let’s consider the following scenario: We start with 3 reduced files.

Sample files in example.#

File Name





Science file with two spectra.



Reference lamp valid for sci_file.fits



Another valid reference lamp

Assuming the two targets in sci_file.fits are extracted and they are approximately at the position 400 and 600 (pixels in spatial axis), after extraction we’ll end up with:


The default prefix for extraction is e and does not have an underscore to separate it from the file name.

After wavelength calibration, since there are two suitable lamps and due to the fact that the pipeline does not combine solutions, it will save two wavelength calibrated files with each one solved by the respective lamp. Then:


Common issues#

No comparison lamps were found#

The latest version may introduce changes to wavelength solutions. To view the current list of available modes and lamps, please refer to the GitHub repository.

If your lamp, filter, or mode is not included in the repository mentioned above, redspec will not work as expected. This is particularly common with custom modes of the Goodman Spectroscopic Pipeline. You may encounter the following error after running redspec:

$ redspec

[15:02:49][W]: No comparison lamps were provided for file ecfzst_0001_science.fits

This error occurs because the Goodman Spectroscopic Pipeline relies on specific keywords to extract spectra, such as:

Keywords that are used to produce a wavelength solution.#




Indicates if HgAr lamp is used.


Indicates if Ne lamp is used.


Indicates if Ar lamp is used.


Indicates if Fe lamp is used.


Indicates if Cu lamp is used.


Slit and mode configuration.

Multiple spectra output#

When taking multiple ARC images during your observation run, they will be linked with your science data. This means that if you capture several lamp files, they will be processed alongside the science images, potentially resulting in multiple outputs of the same spectrum.