User's Guide to the Kast Double Spectrograph

Table of Contents

Quick Reference
Hardware Overview
Common Path
Blue Side
Red Side
Detector Characteristics
Kast Controller
Data Taking System
Position Angle
Arc and Flat-field Lamps
Diagonal Mirror
Kast Focus
Telescope Offset
Setup and Observing Hints
Setup Procedures
Observing Hints
Calibration Lamp Spectra
Exposure Time Calculator
Data Quicklook

Data Archive
Mt. Hamilton Homepage

Observing Hints

Direct Imaging | Flat Fields | Arc Lamp Calibrations | Overscan Subtraction | Acquisition and Guiding | Centering on very faint objects | Scripts | Polarimetry | Setting the Blue X-Stage | Setting the Grating Tilt

Direct Imaging

Direct images may be taken on both sides.

Use "Open" for a wide open decker, and "Open" for a wide open slit. The unvignetted field of view is about 145 arcsec square (337 pixels at ~0.43 arcsec/pixel).

A suggested blue-side window using the direct mirror is number of rows = 325, number of columns = 325; start row = 850, start column = 850. For the red side, with the tilted flat mirror in the grating tray at grating tilt 8800, use window 350, 390, 2020, 424. Check direct windows with the top lights, not the TUB lamps.

An obvious diffculty is that most of our filters are 2" square and must go in the user filter wheel. Their use instead in the red or blue side filter holders where the beam is 3.5" in diameter will reduce the effective aperture to about a meter; you may as well use the 40" and avoid the wrath of Practically Everybody. A more reasonable course is to put the 2" filters in the user filter wheel, which will utilize the full beam, and just use one side for the observation.

Flat Fields

In general, you will need a number of flats well exposed at the red end in order to get a reasonable total number of counts at the blue end. When you are calculating the statistcal accuracy of your flats, remember to convert from DNs to electrons.

In general to do dome flats you will want to configure the telescope and dome as follows:

  1. Open the mirror cover.
  2. Turn off lights in the dome.
  3. Set the diagonal position to either 3 or 4.
  4. Turn on the appropriate flat field lamp.
  5. Take test exposures to determine appropriate exposure time.
  6. Take flat fields (Note, in general the slit width and length for dome flats should match what you intend to use on the sky for data acquisition).
  7. When completed, please turn off the lamps.

Arc Lamp Calibrations

There are a large number of arc lamps available for calibrating Kast spectra. The following combinations of lamps are often used:

The Hg-Cd lamp does need three minutes to warm up to get all the Cd lines.

In general to do arc lamps you will want to configure the instrument as follows:

  1. Turn on appropriate arc lamps.
  2. Move the diagonal mirror to position 2.
  3. Set the slit to be 0.5 arcsec wide (Note, this may not be suitable for all programs or data reduction, but is the most commonly used slit width for arc lamp exposures).
  4. Take test exposures to determine appropriate integration time.
  5. Take arc lamp calibration frames.
  6. When done, please turn off the lamps.

See Calibration Lamp Spectra for sample Kast spectra.

Overscan Subraction

The number of overscan columns per amplifier is listed in the COVER keyword in the FITS header of Kast data. COVER should be 32 for Kast blue data and 100 for Kast red data (though it is possible for it to be some other value, it is very unlikely).

The both red and blue CCDs use two amplifiers for readout. Hence there are two overscan regions, each 32 (or 100 for red) columns wide. The all columns of overscan are at the righthand side of the image. The first 32 (or 100 for red) columns of overscan are for the left hand side of the image, the second 32 (or 100 for red) columns are for the right hand side of the image.

The IDL script will properly identify the data and overscan regions based on the FITS header information, subtract the overscan region(s) from the data and write out a new file containing the overscan subtracted data with an updated header.


InputFitsFile is the raw data file with overscan regions.
OutputFitsFile is the new data file containing the overscan subracted data.
/legendre is an optional keyword to do a third order legendre fit to the overscan region. If one does not use this keyword the overscan value subtracted from each row is simply the mean of that row's overscan values.


A python script to do overscan subtraction from a list of FITS files is also available: for Python 2.7 and for Python 3. This script will correctly identify the overscan and data regions regardless of binning or subregion of the Lick Observatory detector read out or whether one or two amplifiers are used. Each row of the overscan is median combined and optionally a third order legendre polynomial is fit to the overscan. This is then subtracted from the data region and written to a file.

Syntax: -f -i inputfilelist -o outputfilelist or -f -i inputfilelist -o outputfilelist

-f is an optional argument to indicate that the Legendre polynomial fit is desired (this is recommended for most data).
inputfilelist is a text file containing the names (one per line) of the FITS files for which you wish to do overscan subtraction.
outputfilelist is a text file containing the names (one per line) of the output FITS file names. This file should have the name number of lines as inputfilelist. outputfilelist may be the same as inputfilelist, which will overwrite the original files with the new overscan subtracted files.

Examples: -f -i rawdataList.txt -o overscanSubtractedList.txt -i inFitsList.txt -o outFitsList.txt requires python packages numpy and astropy. Also, this script assumes that the path to python is /usr/bin/python. If that is not the case on your machine, you may have to edit the first line of the program with the proper path. Bug reports, comments, and suggestions for should be directed to Elinor Gates (

Acquisition and Guiding

A very sensitive CCD camera is used for object acquisition and guiding. It is mounted on a remotely controlled x-y stage for offset guiding. The field of view is about 2 arcminutes with the diagonal mirror in position 2. Once the field is identified, the object will be positioned on the slit (either directly or with blind offsetting). If possible, the night assistant will guide off the science target on the slit. Otherwise, the NA will look for an off-axis guide star.

The night assistant will operate the camera for you. If you wish to see the guide camera image, you may start the guidercopy program on either shanevnc, gouda or karnak by clicking on its icon or selecting it from the Root Menu (depending on window manager). More information on the autoguider software, refer to the Lick Autoguider Manual.

The autoguider produces a reticle of on the guide camera image, which the night assistant will position over a guide star. The reticle is divided into four quadrants. The autoguider senses what fraction of the light from the guide star falls into each of the quadrants, and then guides the telescope so as to maintain that balance. The autoguider is a pretty good device, but not infallible, so you should watch if from time to time to make sure it's doing the job. Fall asleep at your own risk. Ask the night assistant to explain the autoguider history display to you; it's a useful check on performance.

There is a reticle which may be projected onto a pellicle and thereby mixed into the TV image. The pellicle is a highly stable with respect to the slit when mirror positions 2 or 3 are used, but in position 4 the periscope may cause some wander of the apparent slit image relative to the reticle. This is usually used by the telescope technicians to align the guide TV x-y stage and not during regular observations.

The most worry free situation is if your object is bright enough to see directly on the slit. If you can guide on that portion of the light that does not make it down through the slit, then at least you are assured of where the rest of the light is going. In many cases your object will be too faint to guie on the lost light, in which case you may be able to guide on some nearby object which happens to fall onto the slit jaws, or you may have to resort to offset guiding.

Offset guiding is done in mirror position three, which enables the camera to focus on the solid portion of the diagonal mirror, while an on-axis hole in the mirror passes the light from the object on down to the spectrograph.

The night assistant will operate the x-y stage for the camera for you, and help you find a guide star. Usually one just scans around randomly until one finds a suitable star, but in rare instances it may be helpful to know which way to look for a likely candidate which apears on your finding chart. The useful area of the mirror is approximately as shown below.

Centering on very faint objects (blind offsetting)

If the object is too faint to visually center it on the slit, then a major advantage of this spectrograph design becomes apparent. In almost any case one might imagine, you can dead reckon the object to within an arcminute or so of the slit center. Then, take a direct image of the object while offset guiding to prevent drift, identify your object (down to 23rd mag is not unusual), and use the telescope offset routine to move the telescope so as to center the object in the slit for a spectroscopic observation.

Here are two important hints: 1) be sure to turn off the autoguider during moves, and 2) all of the experienced observers take another direct image after the move to verify that the telescope moved as desired.


There are currently two scripts available to take loops of exposures on the red and blue sides of kast. These scripts are run from the command line on gouda, karnak, or shanevnc. As scripts are added they will be documented here. If you need a specialized script, please contact a support astronomer ( in advance of your run to see if your needs can be accomodated.


To do polarimetry the polarimeter module must be installed in Kast (currently this is the default). Put the waveplate into the light path by selecting a rotation angle for the waveplate from the kast controller software. This will give you a split spectrum (one polarization on the top, the other on the bottom). For certain calibrations you will also have to select the 'filter' - which is actually a polarizing filter (not be confused with the 'polaroid' filter, which is only good for wavelengths < 7300 Angstroms) - in the Upper Filter Wheel.

The waveplate is rotated to any one of four positions (or 16 if you choose "More Options") from the Kast motor control GUI.

Using the dichroic with the polarimeter is not recommended. It introduces uncalibrateable wiggles over the several hundred Angstroms in the vicinity of the dichroic crossover.

The polarimeter shifts the spectrum on the CCD, so you will have to define a new Window to get all the data.

Focusing in polarimetry mode is nearly the same as for regular spectroscopy. However, you should use the centerline option in kastfocus to choose the center row of the top or bottom spectrum for focusing otherwise the kastfocus program will assume the center row, which lies between the two polarization spectra.

Data-taking proceeds as in regular observing, (including TUB rotation as necessary), except that you will want to take exposures with the waveplate in each of its four rotations (0, 22.5, 45, and 67.5 degrees). Most observers take data with the waveplate rotation in the following order: 0.0, 45, 22.5, 67.5 degrees.

It is helpful to note that when using the polarizing filter and a waveplate rotation 0 degrees puts all the light of a calibration lamp in the upper spectrum, 45 degrees in the lower spectrum, and in 22.5 and 67.5 degrees the spectra are evenly split between the two.

Additional calibrations are required for polarimetry: Polarizance test, Polarization standard star, and Null standards. (Descriptions courtesy of Ryan Chornock, UC-Berkeley)

Note that direct imaging polarimetry is also possible with Kast, though the field of view is reduced to about 40 arcseconds by the polarimeter. The necessary standards and calibration procedures are very similar to that for spectropolarimetry.

Further reading on polarimetry:

Miller, J. S., Robinson, L. B., & Goodrich, R. W. 1988, in Instrumentation for Ground-Based Astronomy, ed. L. B. Robinson (New York: Springer-Verlag), 157 "A CCD Spectropolarimeter for the Lick Observatory 3-Meter Telescope" The basic reference for the instrument design and data reduction strategy. A couple of the equations (particularly for the errors) have typos in them.

Goodrich, R. W. 1991, PASP, 103, 1314 "High-efficiency 'superachromatic' polarimetry optics for use in optical astronomical spectrographs" A good description of the design of similar polarimeters. The LRIS polarimeter manual by Marshall Cohen (and updated by Aaron Barth). A description of a similar instrument that also describes the data reduction process.

Schmidt, G. D., Elston, D, & Lupie, O. L. 1992, AJ, 104, 1563 The best polarization standards, if you throw out the ones they mark as variable (!!). Used to calibrate HST.

Setting the Blue X-Stage

Here is an example of how to calculate how much the blue x-stage needs to be moved to get the desired wavelength coverage.

  1. Take an arc lamp exposure with the desired grism (recommended lamps are usually the He and HgCd to get good coverage on the blue side).
  2. Identify a line and determine what column it is on. For example, the 5460A Hg line is at column 1724.
  3. Look up the dispersion of the grism in use. For this example, we'll use the 600/4310 grism, which has a dispersion of 1.02 A/pix.
  4. Determine the wavelength at the extreme end of the spectrum. The blue detector has 2048 columns and in normal use all columns are read out to get the full wavelength coverage. Hence, in this example the reddest wavelength on the detector is 5460A + (2048pix-1724pix)*1.02A/pix = 5790A. The manual states the wavelength coverage range is ~2090A, so the blue end would be 3700. (Note you can do a similar calculation with a line at the other end of the spectrum with respect to column 0. We recommend doing the calculation at the end of the spectrum that is most important for your science.)
  5. For the sake of this example, let us say that the bluest wavelength needs to be 3450A. Hence, the detector needs to shift 250A. Blue x-stage moves are done in millimeters of motion of the detector, so we first need to convert Angstroms to pixels, then to millimeters. We know that the motion of the stage is 67pix/mm, so the conversion is 250A / (1.02A/pix) / (67pix/mm) = 3.66mm. Because we want the spectrum to be bluer, you would tell the telescope operator or support astronomer to move the stage +3.66mm.
  6. Take another exposure after the stage is moved to verify the motion was correct. If not, iterate to the proper position of the stage.

Setting the Grating Tilt

Here is an example of setting the grating tilt to get the desired wavelength coverage.

  1. Look up the tilt formula for the grating. For this example we'll use the 600/7500 grating and its tilt formula is 2.23*c-4687. c is the desired central wavelength of the spectrum. Note that the tilt formulae are approximate, but do give a good starting point for determining the actual wavelength coverage. In this case we have a desired central wavelength of 6563A (H-alpha), so the tilt is 2.23*6563-4687 = 9948.
  2. Take an arc lamp exposure with the desired grating (recommended lamps for the red side are the Neon, Spare Ar, and Hg-A).
  3. Identify an arc line and determine which row it is on. For this example, we'll use the 6402A line at row 2323.
  4. The Kast red detector is much larger than the area illuminated by the spectrum, so the only rows guaranteed to be fully illuminated and not vignetted are rows 760 to 3310. This means the central row of the spectrum is 2035. Hence we want to determine what wavelength is at row 2035. Note that bluer wavelengths are at higher numbered rows.
  5. Look up the dispersion of the grating. For this example with the 600/7500 grating, the dispersion is 1.30A/pix. If 6402A is at row 2323, the wavelength at row 2035 is 6402A + (2323pix - 2035pix) * 1.30A/pix = 6776.4A.
  6. It is also helpful to determine the wavelengths at the extreme ends of the spectrum. In this example, the bluest wavelength will be 6402A - (3310pix - 2323pix) * 1.30A/pix = 5118A. The reddest wavelength is 6402A + (2323pix - 760pix) * 1.30A/pix = 8434A.
  7. With this information you can decide if you need to shift the wavelength coverage so you go redder, bluer, or have the exact desired central wavelength. In this case we'll opt for the exact desired central wavelength. Hence we want to adjust the grating tilt to move the central wavelength from the measured 6776A to 6563A, a difference of 213A.
  8. Calculate the change in tilt to move the spectrum, in this case by 213A. The tilt formula has a slope of 2.23, so simply multiply 213A by 2.23 to get 475. Since we are moving to a bluer central wavelength, this corresponds to a negative tilt move. Hence, change the tilt from 9948 to 9473 (note that the tilt stage only accepts integer values).
  9. Take another arc lamp exposure to check that the move was in the correct direction and amount. Since moves are not exact, you may need to iterate on this procedure.
Note that in many cases, the bluest wavelength is set to match the dichroic beamsplitter, rather than setting for a particular central wavelength. E.g. the bluest wavelength is desired to be 5700A to match the D5700 dichroic, or maybe even 5600A to have some overlap with the blue side when using the D5700 dichroic.

Support Astronomers (
Last modified: Sat Mar 6 16:17:37 PST 2021