FIRE observing manual – local

Before your run 

You may prepare exposure time estimates using the zero point graphs located on the Specifications page 

Your other task is to prepare your observing catalog. The catalogs should be formatted in the standard Magellan fashion, which is documented here.

An always-cryptic field in the catalog is the instrument-specific rotator angle. The value you should use depends on whether you are observing in longslit or echelle mode, since changing modes requires rotating the slit by 90 degrees inside the instrument.

Most users will want to orient the slit along the parallactic angle. To achieve this, in fields 8 and 9 in the catalog file, you should use the following values:

Echelle mode: -30 HRZ

Longslit mode: 60 HRZ 

If you would like to orient the slit along a particular position angle (PA), replace these fields with the following

Echelle mode: [PA-30] EQU

Longslit mode: [PA+60] EQU

 Due to the move of the instrument from FP2 to FP1 there are some changes to the way the software is accessed and the GUI utilities: 

The offset A-> B feature in the Telescope-Control GUI does not work and ABBA offsets need to be performed by the command line in IDL. Details in the manual section "Dithering your target along the slit".

For offsets in the longslit configuration /ls should be added (fatv_abba,/b,/ls and fatv_abba,/a,/ls).

When using the command fatv_abba for offsets the position A/B is not updated in the header keyword.

To properly start fg (IDL slit viewer software) you must open a terminal and type "fatv". Typing fg in IDL is deprecated, as well as starting fatv under IDL.

Afternoon startup

When you arrive in the afternoon, the instrument specialists should have powered on the instrument and detectors, and all related software for the instrument. In case you need to do this yourself, log on to Llama  as obs1, and start the FIRE software by opening a terminal and typing "fire" at the command prompt. Likewise, from a Llama terminal you can start the slit viewer software by typing "fatv". 

The initial window will prompt you to enter the observers' names. It also allows you to change settings pertaining to what items are monitored and IP addresses of various hardware. Do not turn on temperature monitoring and control; this software is running offline 24/7 on a separate instrument-only workstation so you do not want to start a duplicate copy. Also, do not change the IP addresses from their default values. 

The check boxes at the right of the window are used to launch different modules of the software suite. Observers should only launch the camera control and telescope interfaces. Enter your name and click "OK" to launch the FIRE GUI. The telescope control window must be activated to ensure that time stamp and position information from the TCS is written into FITS headers. 

Instrument control GUI

The observer control GUI consists of an instrument control, a telescope monitor, and a log window. You only need to interact with the FireEngine control window. In addition you will launch a guider application, described below. The following subsections explain the various options and interactions required. 

Exposures are started with the "Go" button. A simple loop functionality is built into the instrument or calibration sequences, but not recommended for science frames. Macros are not currently supported. Please make sure to always update values pressing “Enter”, as it is necessary for them to be acknowledged.

Detector Gain

FIRE offers two different gain settings for the spectrograph detector, as shown in the table below. 

For most observations, users should select "High gain" mode. This offers the most sensitivity and lowest noise. However, for bright objects where dynamic range is a concern, or for certain cases in Long Slit mode, users may wish to operate in low gain mode to utilize the entire full well of the detector. Note that in high gain mode the electronics saturate well below the full 65,000 count range to which users may be accustomed. In low gain, the count rate saturates at 32,000 ADU, but at 3.8 e-/DN this captures the entire full well capacity of FIRE's HAWAII-2RG detector. In high gain the detector is completely linear until saturation, while in low gain a substantial non-linearity sets in above 28,000 counts. 

The noise values quoted are for a single double correlated sample read, equivalent to Fowler-1. 

Readout Modes: read noise vs. overhead 

The FIRE spectrograph detector may be read out in one of several modes. Unlike CCDs, IR arrays are read out non-destructively, so that the pixel charge may be sampled multiple times. This is generally desirable since multiple readouts reduce the read noise by ~1/sqrt(N), and read noise is significant for spectroscopy. However, there is an overhead penalty for large numbers of reads, since each individual read takes approximately 10 seconds. 

FIRE offers two main modes for taking science data: Fowler sampling, and Sample-Up-The-Ramp. 

In "Fowler N" mode, the detector is read N times before the start of the exposure, then the software waits for the requested exposure time, and then it reads again N times again at the end of the exposure. The pre-integration and post-integration reads are averaged together respectively, and the image signal is the difference between these averaged groups. Since the detector is read 2N times, the total readout overhead is 2N * (10.6 seconds), in addition to the requested exposure time.

In SUTR or Sample-Up-The-Ramp mode, the detector is read out continuously during the exposure at regular cadence. At conclusion, the software fits for the slope of charge accumulated as a function of elapsed time to estimate the science signal. This provides the benefits of many reads (i.e. reduced read noise) with much smaller overhead at the start and end of the exposure. The drawback is that exposure times are limited to an integer multiple of the array readout time (10.6 seconds).

The graphs below show measurements made of the read noise in Fowler-N and SUTR read modes for the FIRE detector read out in the high gain mode (1.3 e-/DN). At left, we show the noise in electrons per pixel for SUTR integrations of various lengths. The overhead is fixed at roughly 30 seconds for any SUTR read. At right, we show the noise as a function of pair count for Fowler sampling. In this case, the noise goes down but the overhead time (shown at top of the graph) increases linearly with pair number. For exposure times >600 seconds, noise performance approaching that of Fowler-8 is achieved with much smaller overhead using SUTR, and most users will want to operate in this mode. 

The FIRE software only returns a single (slope * integration time) image for SUTR reads, with no cosmic ray rejection. Algorithms exist to reject cosmic rays given a full image cube. If users request this information we may provide utilities to access the raw detector reads but for now this is not provided as a default option. 

Several other readout modes are listed in the GUI pull-down menu, but these are for engineering purposes only and should not be used for science data acquisition. 

Exposure Times

Exposure times are entered into the GUI in the normal fashion; hit return to be sure it registers.

FIRE does not have a mechanical shutter, instead it starts exposures through an electronic reset of the detector. As a consequence, the exposure time you enter in the GUI is not necessarily the same as what you would get with a traditional shutter-based CCD. It also means that users cannot pause and resume integrations mid-exposure. 

When using Fowler sampling, the exposure time entered in the GUI is the time elapsed between the initial and final reads of the detector. Since light falls on the detector as it is being read, the effective exposure time is longer than the entered time by exactly one readout time, or 10.6 seconds. This also sets the effective minimum exposure time for the spectrograph to 10.6 seconds. This should be kept in mind when scaling exposure times for flat fields, standards, and other bright objects. The 10.6 second rule applies regardless of the number of Fowler samples, i.e. a Fowler-2 image also has a minimum exposure time of 10.6 seconds, not 21.2 seconds. 

For SUTR sampling, the exposure time must be a multiple of 10.6 seconds. The GUI will automatically force the exposure time entered into the box to the nearest legal SUTR integration that is >= the time requested. For very long exposures, the number of SUTR samples can become quite large with no significant improvement in noise performance. In this case, to keep the data flow manageable the software reads out in integer multiples of 10.6 seconds (i.e. every 21.2, 31.8 etc seconds) when it can do so without compromising noise performance. 

Slit Selection

FIRE has a slit wheel with 10 positions. One slot is used for a pinhole focus mask, and one contains an opaque blank. This leaves four slits apiece for echelle and longslit modes. The slits are oriented at right angles for the different modes. The echelle slits are 6 arcseconds long and oriented from left-to-right on the slit viewer. The long slits are 40 arcseconds in length (except for the spectrophotometry slit, which is 56 arcseconds) and oriented top-to-bottom.

A new 12 arcsec wide longslit was installed in 2015B as part of a major servicing of the instrument. This is in response to a demand for precise spectrophotometry as of transiting exoplanets, although potential users should be aware of its limitations. First, since the slit length is only 56 arcseconds, there will be a limited number of candidates with suitable comparison stars for differential photometry. Second, the combination of sensitivity and low resolution will result in rapid saturation of the sky in K.  

Slits are selected from the pull-down menu in the GUI. We have observed positional repeatability of roughly 0.15", or one pixel in slit moves. Occasionally the wheel fails to reach its position detent, as can be seen from error messages and / or an uncentered slit image in the slit viewing camera. It is good practice to take a slit view image after slit moves to verify positioning. If a move fails, move to another slit position and then back to the desired position. To date, this has always yielded a successful move on the second attempt. If this fails, contact an instrument specialist for help. 

For echelle mode, the slit center should fall near the position x=168, y=192 on the slit view camera, within a few pixels. The x position should be perfectly repeatable, but the y position moves slightly on slit changes. 

Dispersers

The disperser mechanism is a binary unit which swaps between the echelle grating and the low-dispersion mirror. It is activated from a GUI pull-down menu. Disperser changes are slow and require approximately 90 seconds to complete. 

Calibration options and suggested sequences

A full sequence of spectral calibrations includes pixel flat fields, wavelength calibrations, and illumination (sky) flats. The instrument contains a set of internal calibration lamps, these provide good calibrations for echelle mode. For longslit observations, the internal lamp does not provide very uniform illumination; the dome screen is a better choice for flats and arcs in this mode.

It is advisable to take twilight flats to calculate the slit illumination function, for accurate sky subtraction. In a pinch, dome flats may be used in echelle mode for illumination correction. 

Echelle Calibrations 

To take internal calibrations, you must insert a mirror into the beam, and turn on your lamp of choice. These are two separate actions, and must be undone at the end of the sequence. The mirror is inserted via the "Calib" pull-down menu. A value of "mirror" inserts the calibration mirror, "Clear" removes the mirror for science observations. 

FIRE's footprint includes two orders with relatively few lines for wavelength calibration in echelle mode. It is therefore critical if your science exposures will be shorter than 8-10 minutes to obtain arcs of sufficient depth to bring out several lines. This is an important part of the pipeline since a proper fit to the tilt and curvature of the orders is needed to perform both wavelength calibration and sky subtraction. We have had good success with 10 second ThAr lamp exposures with a 0.6" slit; it should be OK to go even longer so long as most lines are not saturated. 

Wavelength: For exposures longer than 8-10 minutes, an accurate wavelength calibration may be obtained using OH lines from the night sky. For shorter exposures, particularly for flat fields and telluric standards, the ThAr lamp should be used to take a concurrent wavelength calibration. We recommend taking an arc at every pointing, since the instrument does flex at the 1 pixel level with different gravity orientations. Some orders do not have many lines from the ThAr lamp. When in doubt, err on the side of long exposures to bring out weak lines, even if the strong lines saturate. 

Flat fields: We recommend taking two types of flat fields: a set of 10 internal quartz flats to correct pixel gain, and a set of 10 flats on the dome screen to correct for slit illumination. Sky flats are not recommended because of their strong wavelength variations. 

Flats must be kept below 18,000-20,000 counts to avoid saturation effects in high gain mode. For a 0.6" slit, 1 second with the internal Qh lamp produces reasonable count rates. For wider slits, a Ql lamp is available which has lower flux levels to avoid saturation. Remember, when scaling flat exposure times in Fowler mode, the actual exposure time is (10.6 + texp) seconds, where texp is the value in the exposure time GUI. 

The following tables indicate recommended exposure times for flat fields in each of the operating modes. As always, we urge users to examine test exposures to ensure proper count rates.

Long slit calibrations

For longslit observations, dome calibrations with the flat field screen should be used rather than internal lamps. For flat fields, we use the variable quartz lamp whose power supply is in the dome. Ask the Instrument Specialist about the lamp operation, and please make sure to dial the voltage down and turn it off when you are done. Generally, two flats are needed at different voltages, to get adequate counts across the full spectral range. A high voltage flat (2.0-2.2V) is used to get counts in the z and J bands, but saturates in H/K. A low voltage flat (1.0-1.2V) is used to calibrate H and K, but is too faint for the blue end. The FIRE longslit reduction software takes these two flats and splices them together for a final superflat. 

For wavelength calibration, you may use either the night sky lines (which are substantially blended), or else arc lamps. After experimentation with various combinations of flat field lamps, we recommend using a combined NeNeAr setup (both Ne lamps, on Ar on the flat field screen). This balances line strengths nicely. 

The following tables indicated suggested exposure times for flat fields and arcs in each of the operating modes:

Selecting Telluric Standards

It is best practice to observe A0V telluric standard stars between science objects, to calibrate atmospheric absorption features. These observations are also used to perform flux calibration by the reduction software. We have installed a tool at the telescope (on the workstation llama) which assists users in selecting tellurics from a large list of A0V standards. 

To run the tool, go to the directory /Users/fire/python and enter "source python_setup", just once to get the session started. Then, for each field required, run the following:

find_tellurics --RA 14:47:17 --DEC +04:01:12 --UO 08:50 --UT 09:25

The program, find_tellurics, takes the following inputs:

-RA, DEC - The right ascension and declination of your science target

-UO - The central UT time of your science target observation (note this is the letter "O", not a zero)

-UT - The expected UT central time of observation for the telluric standard

The software will output a list of stars which are the closest match in airmass and sky angle:

Object: RA= 14:47:17, Dec= +04:01:12 airmass= 1.372 at lst 16:43:14

Best telluric matches:

(1): HD123309 ----- Magellan Catalog Number: 10215

Mag= 9.40 (Band= V) RA= 14:07:24 Dec= -23:28:29 Anguler Offset= 29.16 deg

This telluric: airmass = 1.364 at lst 17:18:14 airmass diff from source = 0.007. Match rating = 54.18

(2): HD125062 ----- Magellan Catalog Number: 10249

Mag= 9.98 (Band= V) RA= 14:17:29 Dec= -19:29:07 Anguler Offset= 24.62 deg

This telluric: airmass = 1.346 at lst 17:18:14 airmass diff from source = 0.025. Match rating = 30.37

(3): HD123426 ----- Magellan Catalog Number: 10217

Mag= 9.66 (Band= V) RA= 14:08:10 Dec= -24:33:51 Anguler Offset= 30.12 deg

This telluric: airmass = 1.354 at lst 17:18:14 airmass diff from source = 0.018. Match rating = 27.47

The "Match rating" field is a figure of merit for the quality of telluric match, and should be as large as possible with maximum 100.

Once you've selected from the list presented, the telescope operators have the catalog file of all FIRE A0V calibrators available in FIRE_tellurics.cat. It is a long list, so you should specify the star by its catalog object number rather than name. The object number is listed above in the field "Magellan Catalog Number." 

Note that the Telescope Operator has two catalogs available, one for echelle and one for long slit, taking into consideration the 90 degree offset for slit orientation, so that the slit is in the parallactic angle.. Please make sure to tell the TO your observing mode. 

The FIRE slit viewer

FIRE's slit viewer is controlled by the observer. It has a field-of-view of approximately 1 arcminute with 0.147" pixels, and utilizes a fixed MKO J band filter.

To start the slit viewing software, from a terminal on Llama, execute "fatv". This should bring up the window shown below. The software is IDL-based, and is derived from the atv software package distributed freely by Aaron Barth. 

The user's main interactions are to change the exposure time, take frames, set the slit location, and move objects onto the slit. Generally the slit viewer is used for object acquisition, and the observatory guiders perform actual guiding functions. However, the slit viewer may be operated independently during long integrations with the spectrograph, to check the location of the object on the slit and perform adjustments if needed. 

To take a slit image, press the "Acquire" button. The detector will expose for the requested duration and auto-display the image. Press the auto-scale button to scale the image stretch, and use the mouse (left button) to achieve the desired contrast. Though not shown in the image above, at the telescope the display will include a compass rose and scalebar to aid in orienting the field. 

Specifying the nominal slit position: Before starting your spectroscopic observations, you should specify the pixel that will be the defined "A" dither position on the slit. It should be offset by 1-1.5" to the left of the slit center in echelle mode to allow dithering by ~2". In longslit mode, 5" below the slit center is appropriate. Under "MouseMode," choose the "Slit" option and left-click on the pixel where you'd like to place the slit. A crosshair will indicate the location of your selection. 

The mouse can only enter integer pixel positions for the slit. If you want finer control over position, you may enter any value by hand. To do this, from the IDL command line, type "fatv_slit, <x>, <y>" where <x> and <y> are the desired positions, which may be fractional pixels. 

Putting objects on the slit: When your slow completes, take a test image. Locate your object, select the "TCS" mouse mode, and left-click on the object to move it to the slit. You will be prompted for approval before performing a telescope move. It is usually the case that an additional small move (0.1-0.5") is needed to precisely position your target on the slit. See the "handpadde" instructions below for information about how to perform fine moves. 

Saving images from the slit viewer: It is often useful to save images from the slit viewer for later analysis. To do this, click the "Save" button. This will open a dialog box which prompts you for where to write a fits image containing header information about the pointing center and rotation angle. 

Displaying sky subtracted images: For faint objects it is often helpful to take images at two dither positions and difference them to suppress the sky. The guider has a built in function to do this, operated as follows. First, take an exposure that will serve as the sky reference. When it appears on the screen, click "SetSky." This will load that frame into a separate image buffer reserved for the background frame. Then, move the telescope to the target position and take another frame. When the new frame appears, go to the "Display" pull-down menu and select "Raw-Sky." The display window will now show the raw image just obtained minus the sky buffer you saved previously. To switch back to displaying the raw read, simply change the display pull down menu value. An example of a sky-subtracted image is shown below; note that objects on the sky frame will appear as negative images on the Raw-Sky display mode.

There is a known bug in which the compass rose sometimes disappears from the frame in Raw-Sky mode. Operating in "Raw" mode should work if your observations can support this. 

Dithering your target along the slit: For several years, dithering along the FIRE slit was done in IDL. More recently, this capability was added to the FireEngine control software. This is now the suggested procedure as it captures to the FITS header whether your target is in the A or B slit position. Simply enter the magnitude of the offset into the GUI (default 2.0"), and click ""A->B" to move from A to B position, and vice versa. The software is not intelligent enough to tell whether you are in the A or B position to start a sequence, so users may need to use the "set" button to tell the software the current position. 

For those preferring the old-fashioned way, a simple idl script may also be executed from the IDL command line to dither objects on the slit. After completing your first observation, at the prompt, type:

IDL> fatv_abba, /b

to move the object 2.5" to the right in slit in slit viewer coordinates. To move from the "B" position back to "A", type

IDL> fatv_abba, /a

If you wish to move by an amount other than 2.5", you may do so. For example, for a 2.0" dither length, type:

IDL> fatv_abba, /a, dist=2.0 

By default the dither macro assumes you are in echelle mode. If you are observing in long slit mode, add the "/ls" keyword to dither up and down in slit viewer coordinates, rather than right to left as you would for echelle slits. 

Measuring seeing or counts on the slit: The ImExam Mouse Mode may be used to perform rudimentary analysis of image quality. Select this using the pull down menu and then left click on a star to measure its FWHM in pixels. Remember that the pixel scale is approximately 0.147". 

Performing small dithers by hand: Occasionally users may want to perform small offsets or corrections to the pointing in the vicinity of the slit. There is a Virtual Handpaddle widget in fatv built for this purpose. From the top menu bar, select "HandPaddle" to bring up the widget. Offsets are sent by default in the coordinate system of the guide camera, with the software calculating the correct rotation into RA and DEC. If you wish to send offsets directly in RA and DEC, change the "Basis" menu from Guider to Equatorial. For guider offsets, we move the telescope, not the universe. So, clicking "down" on the handpaddle moves the telescope/slit down, or the object up in guider coordinates. 

Blind Offset Pointing with FIRE

For objects not visible in the slit viewer, you must perform blind offset pointing from nearby bright stars. In 30 second integrations on the guider in 0.6-0.7" seeing it is possible to see point sources with J>20 on the slit viewer. Fainter targets may need a blind offset.

To perform a blind offset, first place the reference star on the slit as you would any bright target.

True Blind Offsets: The simplest form of offset is then a true blind offset, where the telescope is sent from the reference star to the target location by dead reckoning. This can be accomplished by simply asking the TO to perform the offset, with the sense of the offset being the offset in RA and DEC to take the telescope FROM the reference star TO the science object. Different users have had varied success with this procedure; our instrument team's tests indicate that the telescopes' blind pointing accuracy is 0.2-0.3 arcseconds RMS. We therefore advise against performing such offsets with a slit width less than 1.0 arcseconds, as the object may miss the slit. 

"Trust But Verify" Offsets: For blind offsets smaller than ~20 arcseconds, the setup star will remain visible in the acquisition camera's field of view. In these instances, the fidelity of the blind offset can be verified (and adjusted if necessary) by predicting what slit-viewer pixel the setup star should land on. For larger offsets the star falls off the acquisition FoV and there is no recourse. The FIRE team has measured optical distortions on the slit viewer and written routines to assist in "Trust but Verify" offsets, as detailed below.

1. First, slew to the setup star and center it on the slit crosshairs in fatv.
2. Then, on the fatv command line, issue the command "fatv_refoffset, arcsec_east, arcsec_north", where the second two arguments are the offset distances in arcsec E and N to move FROM the setup star TO the science target (i.e. the direction the telescope should move). West and South offsets are negative. This will send a coordinated offset to the TCS, moving the telescope and the guide probes, and resuming guiding and shack-hartmann corrections. You will then see a small crosshair appear on the slit viewer image depicting where the setup star should appear after the offset.
3. Now take another image with the slit viewer. Your science target should be on the slit, and the setup star should appear where the crosshair was. The software deletes the crosshair when it refreshes the image, but you can re-plot it by typing fatv_marksetupstar. If you are happy that the setup star is centered in the crosshairs, then the offset was successful, and you can begin integrating.
4. If the setup star is not centered on the crosshairs, it means that your science target is probably not in the slit. One approach is to center from there manually using the handpaddle. If you would like a more automated approach, you can type fatv_tuneoffset. This attempts to centroid on the star, and calculate the correction required to center it on the crosshairs. Do check the suggested offsets before clicking "OK," sometimes there is a large error in the centroid calculated by the software, but most of the time it is OK. When wrong, it suggests large offsets. Strangely, this can sometimes be solved by going into ImExam Mouse Mode and centroiding on the same star, and then re-running fatv_tuneoffset. We have no idea at present why this works, but it often does.

In preliminary tests we found that this method was able to locate sources to within ~1 pixel on the slit viewing camera, or ~0.15 arcseconds for offsets of ~20 arcseconds. Smaller offsets were a bit more accurate. This new mode (as of 2018) has been tested somewhat but not in a completely exhaustive fashion.