Bench Mounted Echelle User's Manual

nb suntzeff, 26 april 1996
t. ingerson, 10 march 1995

1. INTRODUCTION

The Bench Mounted Echelle (bme) is a simple fiber-fed echelle available to users on the 1.5m telescope at CTIO. The instrument is described in the General Description of the Bench Mounted Echelle, written by Tom Ingerson.

The manual by Tom is a guide on the workings of the bme, and a guide to observers who require information on the instrument's capabilities. The information in Tom's guide should serve as a basic guide to those observers who are writing a CTIO Observing proposal for the bme. My manual is intended as a User's manual on how to observe with the bme. Some of the material is repeated (in different words) from the manual written by Tom.

2. THE BME SETUP

The bme is designed to be a simple user instrument with a small number of setup configurations that should satisfy most user needs. At least one month before your run, you will be asked to specify the exact setup for the bme. It is always used with a folded Schmidt camera with a focal length of 750mm.

Thus, you only need to specify the CCD (currently only a Tek 2K can be used), the cross disperser, the slit size (to set the spectrograph resolution) and a filter, if required. An order blocking filter is normally used in the red beyond 6000A to eliminate second order blue light. As the slit width is decreased, the resolution increases and the throughput declines, according to the following approximate table. Tek 2048; 750 mm camera

Slit width             200u    140u    100u     70u     45u     30u

Relative Throughput       1     .81     .61     .44     .28     .19

Resolution (approx.) 15,000  20,000  30,000  45,000  60,000  60,000

The maximum resolution of the system with the Tek2048 is thus about 60,000. There is no benefit from using the 30 micron slit with this chip. If the seeing is very good, in theory throughput with a small slit is better with the 100 micron fiber. In practice this almost never happens, so the 200u fiber is almost always used.

Normal grating setups with Tek 2048:


================================================================ 
Cross      Wavelength  lines/mm  cross-dispersion    Coverage
disperser    Range                                   
---------------------------------------------------------------- 

KPGL2      3200-6000A     316         40 A/mm         ~2000A 
Gr400      5000-9000A     158         80 A/mm         ~4000A

================================================================

We have a wide range of order blocking filters (generally to block second order blue when observing in the red). You can choose from the following list: Schott BG39, BG38, WG335, WG360, GG385, GG455, GG 420, GG495, OG515, OG550, OG570, OG590, RG610, RG665, RG695. All these filters have been anti-reflection coated to maximize transmission within the filter passband. See the CTIO filter list for further information and transmission curves.

3. THE OBSERVING ENVIRONMENT:

When you arrive at the 1.5m console room, you will be introduced to the various SUN workstations that you will be using. Someone from CTIO Observer Support (Ricardo Venegas, Daniel Maturana, Edgardo Cosgrove, or Arturo Gomez) will probably be there doing the spectrograph setup. Nelson Saavedra or Mauricio Navarrete will be on the mountain to help you with computer accounts and IRAF problems. Note that this staff goes to bed around midnight, so try to ask questions before then.

You will be using a Sparc10 Workstation called the "Arcon Data Acquisition Computer" to take data with the CCD and reduce it. There probably will be another SUN workstation connected to ctio60 that you can use for email and the like.

Arcon is the CTIO Array Controller connected to the CCD. The user interface is made to look like the ICE interface at KPNO and should be identical real soon now. If you do not have a CTIO visitors account, ask Nelson or Mauricio to get you one right away.

There are a number of manuals that you should have at your fingertips:

The Arcon-IRAF Interface: A Preliminary User's Guide for Spectroscopy with the 1.5m bme.". This guide tells you all you need to know about how to take data with the CCD.
Thorium Argon Atlas. For a setup to quickly see where you are, I like the atlas "ESO Sci. Rep. No. 6, An Atlas of the Thorium-Argon Spectrum for the ESO Echelle Spectrograph in the 3400-9000A Region" by D'Ororico, Ghigo, and Ponz. Pages 12-18 give pictures of the ThAr spectrum taken with basically the same setup as the bme. For wavelength id's when reducing the data, the atlas "A CCD Atlas of comparison Spectra: Thorium-Argon Hollow Cathode by Willmarth" from KPNO is better because the wavelengths are from the LLL tables and are the most accurate.

It is pretty easy to figure out where you are in the spectra. Remember that that spectra separate more as you go to the red (which is where you also see the very bright Argon lines). By knowing this, and by looking in adjacent echelle orders for the same spectral features, you can deduce the direction of increasing wavelength along the echelle order. This is simple with the BME since the whole free spectral range is covered in all the orders.

If you see faint orders appearing between your reddest orders, you should consider putting in an order blocking filter to remove the contaminating second order blue light.

Navigating around in your echelle frame, you should use the following formulae for the bench echelle:

lambda = 563255/m (A) where lambda is the central wavelength in the echelle order and m is the order number.
FSR = lambda*lambda/563255 (A) is the free spectral range, or the separation in wavelength between orders. For the 2048, the FSR is always smaller than the chip up to about 1 micron, past where the fiber is transmitting.
R = 2*tan(theta)*f_cam/delta(pix) is the resolution, where lambda is in A, f_cam=750mm is the camera focal length, tan(theta)=2 is the echelle blaze angle, and delta(pix) is the FWHM resolution of the data in mm.
d(lambda)/d(x) = lambda/(2*f_cam*tan(theta)) (A/mm) where lambda is in A, f_cam=750mm is the camera focal length, and tan(theta)=2 is the echelle blaze angle.

The last formula is useful to get the dispersion or the number of A/pixel. For instance, h-beta at 4861A is in order 116 (115.9 actually), with a dispersion of 1.62 A/mm. A TEK2048 pixel is 0.024mm, so each pixel is about 0.05A. The FSR is 42A and the chip is 101A wide, so you spectrum fits on less than half the chip. With the 75 micron slit, I measured 2.2pix FWHM on the thorium-argon lamp, implying that with 0.024mm pixels, R=45000.

Following the instructions in the "Arcon-IRAF Interface Manual," login to the Arcon Data Acquisition computer. Choose SunView or OpenWindows. Let the SUN do its login procedure. When you see the phase "Array Controller ready for User Commands..." in the "ARCON CONSOLE" window and you hear the beep, you are ready to go. If you want, start up the Arcon Status and Countdown windows as described in the "ArCon-IRAF Interface Manual." Type "arcon" in the "IRAF Acquisition" window. This connects the SUN to the CCD. Type "bme" in the same window. This loads the parameter files for the bme.

You are now ready to observe! Just type "observe" and answer the obvious questions.

You should check the status of the disks by typing "df". Here is a sample df:

Filesystem            kbytes    used   avail capacity  Mounted on
/dev/sd0a              30807   15633   12094    56%    /
/dev/sd0g             870214  869459     755   100%    /usr
/dev/sd5d            1968782  468081 1500701    24%    /ua21
/dev/sd5e            1213227  478041  735186    39%    /ua22
/dev/sd6c            1923312  108112 1815200     6%    /ua23

You can see that there is lots of space on /ua23 or /ua21. If you are say v18, you would

set imdir = /ua21/iraf/v18/

Note however that this changes the imdir in IRAF, but the imdir that Arcon writes to does NOT change with the IRAF command. (yes, sad but true). The Arcon imdir is (confusingly) set in your .login file. If you need to change that imdir, edit the .login file. To activate the new imdir for Arcon, you must log all the way out of the computer, login again, restart Arcon, and click you heels three times while saying "There's no place like home."

At this point, you should consider choosing the CCD parameters that are right for your objects. If your objects are bright, you should consider using a CCD setup with a large number of e-/ADU and a fast read (which will have slightly higher read out noise). However, if you are observing very faint objects, you should consider a small number of e-/ADU and a slow read (with a small read out noise). Using the command "gainchange" in the "IRAF Acquisition" window, you can set the appropriate values. You probably don't want to chose a value for the gain (e-/ADU) which reaches pixel full well before the A-to-D saturates. For instance, the A-to-D clips at 65536 ADU, and the full well of the TEK is about 225000, so you don't want more than about 3.4 e-/ADU. Set the gain and leave it for the run. Typical values - bright objects or high S/N -- use 3 e-/ADU: faint objects or low S/N -- 1 e-/ADU.

Sample gainchange file for 2 amp TEK2048:


                           Arcon3.3 / Tek2048 
    dcsT Delay ____Read_Noise_____ _______1/Gain______  __Read_Noise___ Read
                      (ADU)              (e-/ADU)            (e-)       Time
    (us)        LL   LR   UL   UR   LL   LR   UL   UR   LL  LR  UL  UR   (s)
    ---  ----  ------------------- -------------------  --------------- ----
 1:    5    2    -    -   1.02 1.00  -    -  4.79 4.79   -   -  4.9 4.8   67 
 2:    7    2    -    -              -    -  3.21 3.21   -   -            75 
 3:   10    2    -    -              -    -  2.40 2.40   -   -            87
 4:   15    4    -    -              -    -  1.59 1.59   -   -           107
 5:   20    2    -    -   2.30 2.45  -    -  1.18 1.20   -   -  2.7 2.9  127
    Full well ~220,000 e- (non-MPP) which is only 45,000 ADU for dcsT=5 us! 
   
        *** Select gain setting from the first column ***
        *** Current gain setting is 2

You should consider the CCD binning and read-out size. You have the ability to command the CCD to merge a certain number of physical pixels into a single data pixel. The advantage to this is that the read-out-noise is per data pixel, not physical pixel. You probably want to bin the CCD by a factor of 2 in the direction of the cross-dispersion. A 200 micron fiber de-projects by a factor of 1.3 to about 6 0.024mm pixels. You could reduce this to 3 pixels without really affecting the quality of the data, and you would therefore be reducing the read noise in the extracted data (3 pixels per extracted pixel rather than 6). You will save disk space and reduce the read time. The disadvantage (there always are, right?) is that a cosmic ray that previously fit in one physical pixel now affects the one data pixel, effectively smearing the cosmic ray over a larger chip area. The end effect is that the cr's are more of a problem.

You should also consider changing the read-out size. ArCon gives you control on what part of the chip to read out. Since much of the chip is not used because of vignetting, you can reduce the read-out size without losing data. In setdet :

        (gain = 3)              Gain setting 
    (preflash = 0.)             Preflash time (seconds) 
        (xsum = 2)              pixels summed in X direction 
        (ysum = 1)              pixels summed in Y direction 
      (xstart = 1)              Start of ROI in X 
      (ystart = 1)              Start of ROI in Y 
       (xsize = 2048)           Size  of ROI in X 
       (ysize = 2046)           Size  of ROI in Y

You can change xstart, ystart, xsize, ysize as you please. Note that these units are in (real) physical pixels, so you should take a full un-binned image to set these four parameters. IMPORTANT NOTE: As of today (26 Apr 1996), quadproc has a bug in it when it encounters CCD data that have read out with regions of interest. If you use the generic quadproc to reduce the data, with all the trim and sec stuff set to INDEF, the program will not process the data right. You must use a special form of the instrument file, and also set the trimsec explicitly. This is easy but ask TelOps, Heathcote, or Suntzeff what to do.

Things to note on the setup. No CCD is perfect, and if you have only one feature of interest, make sure it is not on a bad column or pixel. The CCD can be rotated, so you can ask to have the echelle dispersion roughly along an image line. Generally the Observer Support will align the central order to lie exactly along a line. If you are working REALLY faint, you may want to consider binning along the cross-dispersion direction. Read-noise is associated with the pixel as read, not the physical pixel. Binning by 2x along the cross-dispersion will generally not affect your data reduction.

4. CALIBRATIONS

Unless you are Bob Schommer, you will probably not get all the necessary calibrations done by the time you start observing on the first night. So you do them in the morning.

Here are the basic calibrations:

Zero frames (previously called biases). Take about 25 per night, usually done before dinner.
Milk flats. These are flats taken with a very bright light source and a diffusing filter in the spectrograph. You will flatten this image with a surface fit, and use it as a normal 2D flat field. Take about 5. This needs to be done only once per run. A special lamp with a focal reducer is attached to the fiber head assembly. Ask Observer Support to attach the lamp during the daytime of your second night. Your flats should have many 10000 of counts in the final combined image over the regions of interest in the FSR. It is not important to get any counts towards the ends of the FSR since there will be no spectral data there anyway. This flat is basically just a very washed out version of the quartz lamp data you will be taking. It roughly preserves the color along the direction of the cross-dispersion, and is very good (but not perfect) at taking out the pixel-to-pixel variations.
Quartz flats. This is a bit more complicated, and it depends on what you are trying to do. If you are doing the usual S/N=50 spectra, you can take a series of 10 quartz flats at the beginning and/or end of the night, and use these for your flats. If you are doing very high S/N spectra (200 or more), you should consider using a bright hot, rapidly rotating star near your object and/or quartz flats at every slew position. The hot star spectrum is the best, but you will have to remove any faint ISM lines.
Readers may be puzzled as to why TWO types of flats are taken: Milk and quartz. The milk flats will be used to flatten the data as if the data are typical 2D direct CCD data. In IRAF lingo, you will reduce your data through [OTZF], including your quartzes. This should get rid of the pixel-to-pixel variations to first order. To second order, there are all sorts of things that can happen: pixels have different color responses; there is a small amount of fringing; the milk flat is not the same f/ratio as in focus light so dust pupils will be different; etc. Once you extract all your [OTZF] data, you will divide all the extracted data by the extracted quartz/hot star spectrum to remove these second order effects. Your extracted quartz/hot star spectra also are good indications of how high you can push the S/N by just doing an [OTZF] reduction. I can usually get the S/N in the extracted quartz/hot star spectra to at least 150 with the simple [OTZF] reductions.
This "double flat field" is necessary because the spectra drift slightly in the direction of the cross dispersion: thus, dividing the extracted spectra by the extracted quartz that are only [OTZ] is not correct.
Thorium-Argon spectra. Typically you will take these spectra during the night, at a cadence of about 1 per 1.5 hours. The spectrograph drifts by up to 0.1pix per hour. Usually a 60 sec exposure is sufficient
Daytime flats. If you are doing radial velocities, or want a free spectrum of a G2V star (with EXTREMELY STRONG telluric lines!) you should take daytime spectra of the sky. DON'T FORGET THE TH-AR if you need these spectra for velocities.
Darks. You don't really need darks to reduce the data, but a word to the wise: Always take darks with any bench spectrograph. These spectrographs are wide open and very small light leaks (a partially covered LED in the room reflecting off black paper for instance) will blast the CCD. Every night you should take a dark and convince yourself that there is no stray light. Stray light usually looks like irregular, diffuse patches of light on the chip. IMPORTANT NOTE: The ArCon CCD controller uses fiber optic cables. These can leak light at the connectors. Since the electronics are inside the spectrograph room, the most common source of light is the CCD controller. You cannot see this light. If you have more than 2e-/minute dark, the problem is probably the controller. Have TelOps carefully wrap the Dewar electronics with black cloth. Since this cloth has to be moved in order to fill the Dewar, it is possible that the dark can change from night-to-night. CHECK IT EVERY NIGHT!

Spectrophotometric standards. Yes, you can get surprisingly good relative spectrophotometry through a 3.6 arcsec fiber, especially near the zenith. We have a small list of the Hayes/Latham standards that we have re-calibrated for absolute fluxes at 5A intervals. I can point you to the files with the flux information to be used in the "specphot.stand" IRAF routine.


==========================================================
star       RA           dec   epoch  V    B-V      name
----------------------------------------------------------
HR718    02:28:09.50 +08:27:36 2000 4.28  -0.06    chi2cet
HR1544   04:50:36.70 +08:54:01 2000 4.36   0.01    pi2 ori
HR3454   08:43:13.40 +03:23:55 2000 4.30  -0.20    eta hya
HR4468   11:36:40.80 -09:48:08 2000 4.70  -0.08    the crt
HR4963   13:09:56.90 -05:32:20 2000 4.38  -0.01    the vir
HR5501   14:45:30.10 +00:43:02 2000 5.69  -0.03    108 vir
HR7596   19:54:44.70 +00:16:25 2000 5.61   0.10    58 aql
HR7950   20:47:40.50 -09:29:45 2000 3.77   0.00    eps aqr
HR8634   22:41:27.60 +10:49:53 2000 3.40  -0.09    zet peg
HR9087   00:01:49.30 -03:01:39 2000 5.10  -0.12    29 psc
==========================================================

Calibration Scorecard:

Minimum set of calibration data:

1. Daily zero images (25).
2. Daytime sky + th-ar (if you are doing velocities or need a G2V spectrum).
3. Th-ar taken about once every 1.5 hours (60 sec).
4. Daily quartz (10) for flats -or-
5. a. Quartz (3) at every position during the night -or-
b.Hot star at every position.

5. THE NIGHTLY ROUTINE

Now that you have taken all your calibrations, and the sun has gone down, you are ready to observe.

Nighttime preliminaries: You should have Observer Support (x420) come and open your dome 1 hour before sunset, or after you have finished your calibrations. You should communicate to Observer Support when you will need your night assistant to start. Typically, he should show up around sunset.

When it is dark enough, the night assistant will do a zero point of the telescope. The 1.5m doesn't point too well (20 arcsec errors). You will now be able to see the star on the TV. The TV monitor sees a field 1 arcminute across, with S up and E to the right. The night assistant will know where the 200 micron fiber is and center the star on the fiber. What you will see on the TV is the stellar field and the fiber, which appears as a small image of two concentric circles. The inner circle is the fiber and the outer circle is the ferrule holding the fiber.

Focusing the tv is a bit tricky. The Observer Support people will focus the tv on the fiber by focusing on the cladding (the inner circle). This focus will not change during the night. The night assistant will now focus the object on the fiber. The ideal way to do this is to put the star on the fiber and focus, rather than the field, since the field may be slightly out of the focal plane. If the object is bright enough (V<11 or so) this can be done easily. Fainter than this, you may want to go to a nearby star and peak up.

If you are really paranoid, you can also check the focus with the following two sure-fire techniques:

1. Put a bright star on the fiber, and take a series of exposures changing the focus by 5 units each time. Make sure that the object is not moving. Check a given column using the IRAF task "implot."

2. Use the photometer on a bright star. We have a small, cheap, brick of a photometer which takes light from the secondary shadow in the spectrograph (the secondary shadow is pretty well preserved by the fiber). If you open the shutter (or take a really long exposure), you can quickly figure out where the "sweet spot" is on the fiber for guiding, and also estimate the best focus by maximizing the counts. You can also get fooled by this method if you are away from the zenith. A wide range of colors is being imaged in the spectrograph. The photometer is sensing mostly red light so if there is any atmospheric refraction, maximizing the counts in the photometer may mean you are locating the red image of the star and not the blue one.

The control of the quartz and th-ar lamps can be found in a elegantly hand-made aluminum box labeled "Fiber Opt. Feed." Two switches are on the box, one to move a comparison lens in, and the other to turn on the comparison light.

6. WRITING THE DATA TO TAPE

Many of you will leave the mtn with [OTZ] or [OTZF] processed data. The former is in "ushort" (2byte integer unsigned) and the later in "real" format (4 byte).

These are BBIIGG images and you must use an exabyte or DAT. The 1.5m has both. To write to the drive, first go to the drive and copy down the name, which is usually something easy to remember like /dev/nrst1 (or is it /dev/nrts1 ?) ["/dev/nrst1" is actually very easy to remember if you use the mnemonic: it sounds just like "/dev/nrst1"]

Next, within IRAF, load the "ctio" package and print out the help pages on "exuse." (just type "help exuse") This will tell you how to write a tar tape. Generally, we recommend that you write FITS images into a directory and tar the directory to tape, copying the directory twice. It is safer to copy the data to two tapes, but sometimes there is not enough time, so one merely copies the data twice with two "tar" commands. If you are going to copy unreduced data which is in "ushort" format, see the note in the "The ArCon-IRAF Interface Manual" on how to write the data to fits. Basically, you write the data as

wfits @list bscale=1 bzero=32768 bitpix=16 autoscale- scale+

You can also copy the data directly to tape in FITS format, using the usual IRAF wfits command. More and more observers are just writing simple FITS tapes, rather than tarring FITS images. The advantage of a FITS is that you can easily access the data later on. That is, with a FITS tape you can dump the FITS headers with IRAF "rfits" command and then pull off the appropriate image. With a tar file, you will have all your FITS images in a single tar file and you will not know the name of the object in the header, unless you astutely dumped the headers to a file before writing the tar tape.

The only disadvantage of a FITS tape is that there are lots more EOF's on tape which means more ways that tapescan get confused and fail. I am still uncomfortable with this option, so I try to write the data twice: one on a FITS and the other on a tar tape.

I strongly recommend that you check your data by trying to read at least one file from the tar tape back. To do this, try.

Skip to a file:

mt -f /dev/nrst1 fsf ? (replace ? by number of eofs you want to skip)
mt -f /dev/nrst1 status (to see where you are)

then

tar tvf /dev/nrst1

If the tar command can access the tape via this list, chances are the data will be able to be read.

8. FINALLY

I would appreciate any comments on how to improve this manual (yeah, I know, "try learning how to write well, stupid") or new observing tips that can be included. Email me at nsuntzeff@noao.edu. I'm especially interested in problems you have encountered in data reduction, so I can plan new calibration strategies for future observers.

APPENDIX A

The 4m dome will occult the 1.5m if you point the telescope too far SE. During the day, however, the 4m dome provides an excellent source for the solar spectrum. In the following table, if you exceed the given HA at the listed declination, the 1.5m telescope will be occulted by the 4m dome.


          =========================
          Declination    Hour Angle   
          -------------------------
          -50             5 44E
          -55             5 30
          -60             5 05
          -65             4 55
          -70             4 50
          -75             5 10
          =========================