User's Guide to the Gemini Twin-Arrays Infrared Camera

Table of Contents

Quick Reference
What is Gemini?
Summary Table
Quick start - for the expert
Not so quick start - set-up
Graphical User Interface
More About Gemini
Signal-to-Noise Estimates
Sampling Modes
Writing Scripts
Observing Recipes
Computer Setup
Testing the Arrays
Instrument Maintenance and Trouble-Shooting

Mt. Hamilton Homepage

Observing Recipes

Darks and Biases
Flat Fields
Spectroscopic Calibration - arc lamps and sky lines
Photometric Standard Stars
Spectroscopic Standard Stars
Narrow Band Imaging
Deep Imaging
Using the Grisms
Drift Scanning

Various observing "recipes" are described below. These are not meant to be strict rules of observing, but rather suggestions for using the tools provided by the Gemini system. Observers will, of course, have their own preferences and needs based on their science programs.

The most general approach used is to keep the star or object "on the frame" whenever possible and "dither" the telescope pointing between integrations. This technique eliminates bad pixels, and provides good flat-field response and sky subtraction and improves sampling. Obviously, for extended objects this may not be ideal.

Darks and Biases

Dark and bias frames should be taken at the beginning and end of each night.

The recipes below are for single sampling mode and very useful for checking the noise characteristics of the chip and to make sure the camera is in a steady state condition. If observing with a different sampling mode, you may want to change the recipe accordingly (the Gemini software will tell you the minimum Itime if what you set too short).

BIAS: Press F1, enter object name BIAS, set Itime = 0.07, Coadds = 1, select SampMode 1 (single sampling) and set the filter to Blank (B). Press F4 to take and image in both channels.

Typical ranges in single-sampled bias frames are:

  • NICMOS: -4,000 DN to +4,000 DN

  • SBRC: -7,000 DN to -1,000 DN (approx.)

    The dynamic range is about 6,000 DN for the SBRC at 24 electrons/DN giving 144,000 electrons full-well.

    The dynamic range is about 8,000 DN for the NICMOS at 20 electrons/DN giving 160,000 electrons full-well.

    Repeat this test in any sampling mode you plan to use. There should be very little bias structure remaining in any of the CDS modes.

    DARK CURRENT: Enter the Next Observing Setup and select Blank from the filter list and change the Itime to 100.0 sec with 1 co-add. This setup will provide a DARK frame.

    Examine the displayed STATISTICS for the MEAN value and divide by 100 to get DN/sec. In a steady state condition the nominal values are:

    NICMOS: 0.5 DN/sec = 10 electrons/sec
    SBRC: 2.0 DN/sec = 50 electrons/sec

    NOTE: These figures are usually dominated by the "hottest" pixels and the MODE is generally a bit smaller.

    Dark current can be scaled with time (if bias structure is subtracted away). However, it is also useful to take dark frames with exactly the exposure time with which observations were made.


    At the beginning of each night, it is necessary to optimize the focus of the telescope with the Gemini camera. Gemini is approximately parfocal with the Lick guide camera, but not exactly. To optimize the focus, you may choose either a relatively bright star, for example and Elias IR standard, or a fainter star, for example K = 11. In either case, integration time and co-adds must be chosen to be short enough not to saturate on the star, and long enough to average short timescale changes in the seeing. Focusing is best done with differences of mode 1 frames (since saturation is easy to determine), but this is time consuming. Most observers prefer to use CDS frames. Be aware, however, that a saturated star can resemble and out of focus star!

    Telescope focus may be changed using the telescope operator's POCO GUI. Focus encoder values are displayed to the telescope operator (TO), so you will have to ask the TO for the current focus value and the change the focus.

    1. Record the "nominal" focus units.

    2. Take and image of the star that is short enough not to saturate. Use enough co-adds to get 5 or 10 seconds of averaging the seeing.

    3. Record the FWHM, x, and y, of the star using the F9 profile.

    4. Change the focus by 3 or 5 units and take an exposure. Record the focus position and FWHM of the star. Repeat until you are sure you have gone through the best focus.

    5. Choose the best value of the focus (typically 1.5-2 pix in each direction is optimal on an average photometric night).

    6. Move focus to the best value and take a test exposure to ensure that the focus is good.
    It may be necessary to refocus the telescope during the night. It is worthwhile to use the profile function periodically to check the focus.

    Flat Fields

    Dome flats often seem to work quite well with bright objects, so it is recommended to get a complete set of dome flats in all filters; daily dome flats should be obtained, although there is no real evidence of drift.

    To take dome flats, ask the telescope operator to open the mirror cover (and close the dome) and to put the diagonal mirror in position 3. There are three dome flat lights that may be used. The "red" lamp is faint and suitable for direct imaging flat fields. The "blue" lamp is brighter and useful for narrow band filters and grism flats. The "super blue" lamp has a rheostat voltage control and can be varied in brightness for any flat fields. When taking dome flats in 2 micron or longer filters, it is necessary to take frames with both the lamp on and the lamp off (dome dark) in order to remove thermal background. This is especially important for K grism flats.

    Sky flats should be obtained by dithering the telescope between exposures. A sequence of at least 5 positions is recommended for median filtering, with 7 to 9 being even better.

    Twilight flats are also possible, though they are somewhat tricky to obtain, as they are at any telescope. Beginning with J and proceeding through H and K and narrow band filters is the best strategy at the end of the night.

    Spectroscopic Calibration - arc lamps and sky lines

    Grism observations require relative and absolute wavelength calibration. Relative calibration is done using the OH sky emission lines. These lines are plentiful throughout the J, H, and K windows. There are fewer OH lines at the long end of the K window, but fortuitously, there are other sky emission lines which have been tabulated which fill this region; these lines have been observed to be repeatable over the course of the year. OH and other sky emission lines are tabulated in:

    Maillard, J.P. & Chauville, J. 1976, Journal of Molecular Spectroscopy 63, 120
    Oliva, E. & Origlia, L. 1992, A&A 254, 466
    Ramsay, S.K., Mountain, C.M., and Geballe, T.R. 1992 MNRAS 259, 751

    Absolute wavelength calibration is achieved using noble gas arc lamps. The Gemini equipment included the following lamps: Neon, Argon, Krypton, Hydrogen, Hg-Vapor, and CO2. These lamps are mounted on the inside of the telescope, on the side toward the control room door. Only one may be mounted at a time. The default installed lamp is Argon, if you require a different lamp contact a support astronomer in advance of your run to be sure the lamp is available. The lamps must be changed manually using the lift. The lamp is plugged into the first outlet of the lamp controller, Spare1_(110SSR). It is turned on by clicking the Spare1 button in slinelamp_fe. Also be sure that the diagonal mirror is in position 2 when doing arc lamp exposures. The lines from each lamp are tabulated in Tables of Spectral Lines of Neutral & Ionized Atoms, Stringov, A.R. & Sventitskii, N.S., IFI/Plenum Data Corp, 1968

    Argon line identifications for Gemini grisms are now available listing approximate pixel position with the wavelength of the line for each grism.

    Sky lines identifications are available at

    Photometric Standard Stars

    The usual method of observing standard stars is to use a script to move the standard star to two or more various positions on the chip.

    If seventh magnitude Elias standards are used with broad band filters, under good seeing conditions, it is essential to use the Single Sampling mode with the shortest allowed integration time. Even so, those brighter than K = 7.0 will saturate.

    For the fainter (9 to 13th magnitude) UKIRT standards, the CDS mode can be used, but care should be taken not to saturate on the brightest ones, and not to underestimate the required exposure time for good signal-to-noise ratio on the fainter stars (about 5 minutes).

    IR standard star lists are available both on-line and in the 120" control room.

    Spectroscopic Standard Stars

    This section is not intended to be a complete reference for spectroscopic standard stars. Most grism spectroscopy programs will choose appropriate standard stars in advance of arriving at the telescope. This section is meant simply as an introduction so that "spur of the moment" spectroscopy can be done if it is required.

    The purpose of a spectroscopic standard is to provide a calibration for the effect of the atmosphere on the spectrum from an object. The "perfect" standard, would therefore be one with a featureless spectrum. The closest approximation to this in the astronomical world is to choose stars whose spectra have few features in the waveband of interest. Once a stellar type is identified, appropriate stars can be found in the SAO catalog or the Bright Star catalog. There are copies of both in the 120" control room. When choosing standards, it is important to avoid variable, peculiar, binary, and otherwise "unusual" stars.

    It is important to take spectroscopic standards at similar airmass and close in time to the objects they will be calibrating. As a rule, one should seek standard stars within 0.2 airmass and 30 minutes of the observation.

    J band: Late F and early G giant stars have few features in the J band. They are plentiful in most regions of the sky.

    H band: O stars are the best standards for the H band. Unfortunately, they are rare. A list of all the O stars in the Bright Star Catalog is below.

    Bright Star Catalog O Stars
    HRNameHDRA (2000)Dec (2000) VB-VU-BR-ISp. Type
    65---133700 17 43.051 25 59 6.14-0.13-0.970.00O9IIInn
    1209---2453403 55 23.031 02 45 6.100.29-0.820.00O9.5ep
    122846Xi Per2491203 58 57.9 35 47 284.040.01-0.92-0.01 O7.5III
    15429Alp Cam3061404 54 03.0 66 20 344.290.03-0.880.00 O9.5Ia
    1712---3407805 16 18.2 34 18 435.960.22-0.700.00 O9.5V
    185234Del Ori3648605 32 00.4 00 17 572.23-0.22-1.05-0.22 O9.5II
    187939Lam Ori3686105 35 08.3 09 56 033.54-0.18-1.03-0.17 O8III((f))
    2422---4712906 37 24.1 06 08 076.060.05-0.880.00 O8p
    2442---4743206 38 38.1 01 36 496.210.15-0.820.00 O9.5II
    245615 Mon4783906 40 58.7 09 53 444.66-0.25-1.07-0.22 O7V((f))
    2467---4809906 41 59.3 06 20 426.37-0.05-0.94-0.05 O6.5V
    75749 Sge18800119 52 21.8 18 40 196.230.01-0.920.00 O7.5Iaf
    7589---18820919 51 59.1 47 01 395.62-0.07-0.97-0.12 O9.5Ia
    7767---19332220 18 07.0 40 43 565.840.10-0.780.01 O9V
    8023---19957920 56 34.7 44 55 305.960.05-0.850.00 O6V((f))
    815468 Cyg20306421 18 27.2 43 56 455.00-0.01-0.940.00 O7.5III:n
    8281---20626721 38 57.6 57 29 215.620.21-0.740.00 O6.5V((f))
    8327---20719821 44 53.3 62 27 385.950.31-0.640.17 O9Ib-II
    840614 Cep20948122 02 04.6 58 00 025.560.06-0.860.00 O9Vn
    842819 Cep20997522 05 08.9 62 16 485.110.08-0.840.03 O9.5Ib
    846922Lam Cep21083922 11 30.7 59 24 525.040.25-0.740.15 O6I(n)fp
    866210 Lac21468022 39 15.7 39 03 014.88-0.20-1.04-0.22 O9V

    K band: G and K main sequence stars are featureless in the K window except for Br-gamma. If the 2.16 micron region is not of interest, these stars will suffice. If that region is necessary, the G or K star spectrum can be interpolated over (using a model atmosphere is possible). Alternately, a separate calibration can be taken for that part of the spectrum. Early to middle A stars have a feature-filled spectrum except in the region Br-gamma.

    Narrow Band Imaging

    Gemini has available more filters than can be installed at one time. While filters are not routinely changed, and it may be a year or more before the current setup is altered, this section will detail all the filters available. The User Interface Software will always have and accurate filter list.

    This section is more of a data sheet than a recipe as the procedure you will use ultimately depends on your program. The resolution numbers were measured from the manufacturer's traces, so they will be accurate to within a few percent. The other figures are much more variable, by perhaps as much as a factor of 2, due to the intrinsic variability of the backgrounds and sky transmissions. In addition, the observed transmission of filters in the K window will vary depending on which dichroic the light passes through; there may be as much as a 20% difference.

    Most of the narrow band filters exhibit "ghost" images; these are internal reflections off the filter and/or subsequent optics. The strength of these ghost images varies from filter to filter, but is typically on the order of 2% and always less than 5% of the signal from the "ghosted" object. Some "ghosts" move anti-parallel to the object when the telescope is dithered. Others move with the object and are thus harder to distinguish from real objects. It is very important to consider the possibility of ghosts in narrow band data. It is important to "dither" observations with narrow band filters; ghosts may then be removed in data reduction through appropriate combinations of pixel rejection and pixel masking.

    Often narrow band observations will require very long integration times to achieve desired signal to noise ratios. Typically, one desires to be background limited; that is to have the shot noise in the background signal be several times the read noise. This calculation can be done based on the background levels provided in Filters and the readnoise estimates provided for each sampling mode in Sampling Modes.

    Many narrow band filters also exhibit interference fringes due to reflections internal to the detector; these effects are reasonably stable and should flat-field or sky subtract away.

    Deep Imaging

    This section outlines the suggested method for performing deep near-infrared imaging with the Gemini camera on the Lick Observatory 3-m telescope. We have successful results using a dithering technique in which the total integration time is divided into numerous shorter integrations, displacing the telescope in a regular pattern between integrations. This way, the disregistered frames can be median filtered together to build a master "sky frame" which is used to flat-field the individual frames. Then the frames are registered and median filtered together to obtain a single deep image. Similarly, you may use the "sky frame" to subtract the sky background and use a dome or twilight flat for flat-fielding. Some observers prefer one method over the other.

    J, H, and K images may be obtained simultaneously in some programs, and each can reach comparable limiting magnitude by integrating J for 1/3 of the K integration and in H for the remaining 2/3 time. Overhead is greatly reduced by writing a script which will automate the telescope moves, filter changes, and integration times. Examples of scripts are found in the c:\gemini\script directory of the Gemini 486 PC.

    If you know what resides in your field, it is simple to calculate the necessary parameters in this observing strategy. However, if the purpose of the deep imaging is to find out what is there, reasonable guestimates are required. The parameters are outlined below.

    Integration times. Deep imaging is broken down into three different integration times: time per co-add, time per frame, and total integration time. Of these, the time per frame is the only one not normally encountered with infrared array observing.

    The time per co-add must be as large as possible without the brightest interesting objects (usually the background level) reaching the non-linear part of the wells in the array. This also ensures that the frames are background-limited. Total integration time depends on how deep the observer wishes to reach. Time per frame is a bit more tricky. If is important to have a large number of frames taken close in time for purposed of constructing a good sky frame. Optimally, we tend to use frame times of about 2 minutes; any longer than this, and changes in the sky may prevent construction of an accurate flat-field. In addition, if objects in the field get too bright in the individual frames, a "shadowing" effect may appear in the final image which surrounds bright objects with inverse images of themselves.

    Dither pattern. Once the total integration time and time per frame are determined, the required number of individual frames is set. A dither pattern should be constructed that is centered on the field of interest and symmetric around it, so that there is sufficient overlap of frames in all directions. We tend to use a square 3 x 3 pattern of nine frames.

    Telescope displacement. The greater the telescope move, the smaller the deep central overlapping region becomes. However, the displacement must be great enough for extended objects to "clear" themselves, such that they will not fall onto common pixels from frame to frame. A displacement of 14" is a good default value.

    Once the observing parameters have been chosen, some final issues must be considered.

    Using the Grisms

    Observing procedures for the grisms are still being optimized, but our experience in the July 1994 run was very efficient and fruitful. This section will not address the necessary preparation which precedes the actual observing, such as finding suitable standard stars and calculating the required exposure times to obtain the desired signal-to-noise ratios; however, these tasks are usually more important for spectroscopy than the analogous tasks are for other types of observing. For more information about spectroscopic observing and data reduction, please see Don Figer's 1995 Ph.D. thesis and the IRAF documentation on long slit spectra.

    The following steps, taken from experience in the July 1994 run, are a good first start for most spectroscopy programs. Before beginning observations, check which column the slit is centered on (nominally 128, but it may vary slightly depending on mechanical alignment). For each observation, you will need to decide a "start spot" along the slit to place the object; which row this spot is on depends on how many nod positions you plan to use. You will need to guide during the observations. If you are only using two nod positions, ask the telescope operator about using the alternating guide box positions ("A" and "B" on the guider).

    1. Have the operator move the telescope to the target field; it doesn't matter if the object is off-center up to a minute or so as the telescope will have to be moved again in step 4. Start guiding.

    2. Take a quick exposure in order to identify the object in the field.

    3. Press F7 or F8 to activate the cursor in the graphics display.

    4. Put the cross-hair cursor on the object and press F11, the astrometry key. Position the cursor on the "start spot" and press return; press the "y" key to have the software move the telescope so that the object is on that spot. Stop guiding while the telescope is moved.

    5. Take another quick exposure to verify that the object is on the "start spot". Begin guiding.

    6. Use Setup-Motor Move to put the slit in place.

    7. Choose the appropriate grism from the filter list in the observing setup at the bottom of the text screen.

    8. Set the integration time to the desired value; we obtained S/N > 10 for K = 11.0 in 3 minutes.

    9. Set the sampling mode to 3, and set the number of reads (MREADS) to 16.

    10. Press the Go key to initiate the exposure.

    11. If the image looks ok, stop guiding.

    12. Move the telescope to the next nod position.

    13. Begin guiding.

    14. Press Go again.

    15. Examine the spectrum to verify that the object has not fallen out of the slit. If it has, you will need to move a filter into place instead of the grism, and set the focal plane aperture to open instead of the slit. Then take another quick image to try and rectify the situation.

    16. Continue with this procedure until you are confident that enough spectra have been taken. You may want to automate it with a script, but you will need to be vigilant that the object does not fall out of the slit.

    Note: To get Arclamps

    1. Put mirror to position 2.

    2. Switch on "Spare1" in slinelamp_fe gui to turn on Gemini's installed arc lamp (usually Argon, unless otherwise specified).

    3. Do a "Lamp Off" of same position.
    Arc light is reflected into the instrument from white card on the back of the acquisition mirror.

    Column # to Wavelength Conversions for some Popular Lines
    BandColumn #Wavelength Line
    K1472.121 H2


    The Gemini polarimetry mode allows polarimetry data to be taken with a minimum of observer involvement after the initial setup phase. To access this mode, go into "General Setup" under the "Setup" menu item; then select the "polarimetry" option under "observing mode". This sets up the software for polarimetric observations. The software is now configured so that the waveplate and polarizer are automatically placed into the beam. Check that this is so by looking at the display parameters in the upper part of the screen. Assuming all is correct, you are ready to begin observations.

    There are two basic differences when observing in polarimetry mode vs. imaging mode. The first is the software now automatically takes 4 images each time the "GO" command is issued. These are taken in the order of 0, 45, 22.5, 67.5 degree positions of the waveplate. The second, more visible change, is in the TV display area.

    On the display, all four waveplate position images are shown in smaller versions from the normal display. Each frame is displayed as the data is received. Also, there are 4 new command functions, F1 thru F4. These allow you to select and individual image with which to interact on the TV display. Setting the offset and gain for one image will set it for all 4, though you must manually go to each image (via the F1 thru F4 commands) for the screen to redisplay. [Bug: it doesn't always automatically display when the new image is selected. Easiest way around this is to press PgUp followed by PgDn which will restore the original stretch set up on another image.] All other commands are the same.

    Test frames are still available in the polarimetry mode and they are a single frame. This image is displayed in the upper left image spot and is generally taken in the waveplate angle position of 0.

    To improve the duty cycle, polarimetry mode will begin the next integration in the series of four before saving the first frame to disk. If the exposure time is short enough that the next frame arrives before the first is saved the software will beep while it finishes saving the first integration. In most cases (exposures longer than about 10 seconds), this will not be a problem.

    The observing setup for the next set of polarimetric observations (four frames) may be input at any time during an observation. The changes will not be made until the current suite of four positions is finished.

    Warning: Exposure setups with exposures less than 1.5 seconds will fail in polarimetry mode, requiring a reboot of the computer. If you need to do short exposure polarimetry, you will have to do it in Imaging mode, put in the Polarizer, Waveplate, and manually rotate the waveplate through the four positions.


    This mode works analogously to the polarimetry mode. Selecting "SpecPol" in the "General Setup" menu moves the waveplate, the calcite polarizer and the double slit aperture into position. The image display is again split into four images.

    Drift Scanning

    1. To run the drift scanning software, on the Gemini computer set the directory to c:\Gemini\PC and type "drift", which will run the most recent Gemsa and Confsa files.

    2. This will bring up the Gemini user interface.

    3. Go to the setup menus for each chip and choose the "drift" (no. 5) option from the type of integration menu.

    4. To set the exposure time the following formula is helpful though the final adjustments have to be done by trial and error. The primary determinant of the rate is the declination and the max drift rate of the telescope.

      Since we are limited to a maximum drift rate of 1"/sec, that number will be fixed and just the declination comes into play:

      cos(51degrees) * 1"/sec = 0.629"/sec

      The Gemini pixel scale is 0.7"/pix, thus if we want the time per pixel we need to divide the rate by the pixel scale:

      0.7"/pix / 0.629"/sec = 1.11sec/pix

      This number does not take into account the overhead of reading out the array which cuts the sec/pix by a factor of ~2. So the maximum possible integration time is 0.55sec. This should be adjusted by a standard star or other point source at the same Dec as the object that is being imaged. Then the number can be split into an actual exposure time and co-adds.

    5. The amount of sky to be images is determined by the multiread function. The value entered in the "multiread" is the number of columns that the final strip will be (Actually it's the number minus 16). So if you put in 240, that will give the same Gemini Field as if it were done in standard imaging: 256x256. If you do 241 columns then it will display the first 240 (as a 256x256 picture) and then display the 241st column as a separate frame.

    6. Now the integration on each array can be started but there should be a 30 second separation between the start of integration between each array so they do not try to write to disk at the same time. The "Integration Status" displays do not function correctly. Co-adds count down but the bar is not displayed. The channels beep when finished.

    7. For L band observations you need to set the dichroic to K@SW. Remember that the program comes up with K@LW, so if you left the dichroic at K@SW be sure to enter General Setup and change the K@LW line.

      Detector A starts at pixel 129 and next 16 (to 144)
      Detector B starts at pixel 113 and next 16 (to 128)

    Drift scanning is like single sampling. Typical dark/bias levels are:

    A: -4,500 in a single co-add.
    B: -7,200 in a single co-add.

    In drift scanning you get 16 columns worth of these values. Therefore typical dark/bias levels are:

    A: -4,500 x 16 = -72,000
    B: -7,200 x 16 = -115,200

    You can plot these best by using the following gains and offset settings:

    A: Gain = -0.0125 Offset = -65,000
    B: Gain = -0.0625 Offset = -115,000

    Both plots show horizontal "banding" structure. F10 (spPlot) can give a scan (inverted) along any line to show the ramp up at both sides.

    If multireads = 128, the "right hand" half of each display is shown.

    Saturation: Drift Scanning is like single sampling x 16. Therefore if bias is:

    A: -72,000 then the saturation is ~+500 x 16 = 8,000
    B: -115,200 then the saturation is ~-3,200 x 16 = -51,200

    Note: These saturation levels are really 2/3 full well:
    Max for A: ~+1-,400 DN
    Max for B: ~-35,800 DN

    L exposures must be less than 0.010sec (good clear skies) and probably more like 0.007sec.

    Last modified: Wed Oct 20 14:20:26 PDT 2010 by
    Elinor Gates