Observing multislit mask spectra with IMACS

These instructions are in addition to those "official" instructions provided by LCO.

Your masks should have been submitted 6 weeks prior and created (you would receive a response). You should have sent LCO your set-up instructions, which included the mask numbers for your run (that is how the staff knew to install those masks). Up to six masks can be installed at a time.

When you arrive at the Baade in the afternoon of your first night, check to make sure that IMACS is ready to go. The TO or Instrument Scientist should have set everything up for you. If it is set-up properly, you should be able to sit down at LLAMA (not BURRO) and see a number of windows open which you need while taking data. Make sure you create your directory using the "Options" button in the CamGUI (pink window) as DataPath. DO NOT type mkdir somewhere and hope for the best. Use CamGUI. Make sure you choose the correct file number to start you sequence of images (in the pink CamGUI window). For instance I like 1000 for the first night, 2000 for the second and so on.

Below, there is an attachment which shows what the instrument Desktop should look like. There are 8 things to note in this image and they are annotated. The IMACS ACQ controls what goes into the camera -- things like masks, filters, dispersers, etc. You won't be using the lamp buttons in this control window. Instead, you will use the Lamp options window at the bottom. Note that in this screen shot, I have the calibration arc lamps turned on (they are lit red) and the screen is IN. The CamGUI (pink for f/2, blue for f/4) controls the exposures -things like time, file names, object names, object types, etc. The "Hardhat" contains a lot of information I do not understand. The only important thing I can say is that if any of the green boxes turn yellow, something on the telescope is happening (mirror or focus adjustments). Wait for them to turn greenBEFORE you do anything. The MOSAIC readout is fairly obvious. The CCD #s are noted on a piece of paper above Llama. There is a little green box in this window that you can move by clicking the mouse at any location. It will show a zoomed image and also provide some count information (like min/max) within that zoomed region.

First, make sure your observing plan is ready. You will need to upload your observing plan to LLAMA. For instance, if you are logged in as obs2 on LLAMA, then put your observing plan in /Users/obs2. The password for log in should be posted on the wall behind LLAMA. If not, ask for the password. Make sure the telescope operator (TO) knows where this file is located. You will tell him/her where to go based on the numbers you provide in the first column of each row in this file. Note that the mask generation software has a routine called "obscat" which can be run once your masks files are ready. It will output a line per mask which you can paste into a nightly observing plan. I have attached an example plan below (miller_aug102010.cat) which has two imaging fields, followed by three multislit mask fields.

You will mostly likely be using the f/2 (for both imaging and spectroscopy). This is because the SDSS filters are there and not in the f/4 camera (which might only have BVRI). You will want to take your imaging in sdss g,r,i,z. So choose the camera accordingly (f/2 vs f/4). For the instructions that follow, I assume you will always use the f/2.

The following steps should be taken:

(1) In the afternoon, take direct images of the masks. The idea is to get enough light through the slits and onto the CCD that you can see the slits. You will want to specifically see the boxes which are your acquisition objects (6 or more stars that have magnitudes ~15-16). You will use this image to help define the ultimate alignment of the telescope in order to make sure that the photons from the galaxies go through at the right spot on the mask.                                                                          

• 1. Use the lamps INSIDE the dome. Ask the TO for help. Use ~1.5volts and a 10second exposure. But vary the exposure time until you have ~3k counts. Not too bright though, or the slits and acquisition boxes bleed. The lamps power supply is right up the stairs from the Control Room and on the base of the telescope. It is a blue HP power supply. Turn it on (black switch, bottom left). Crank the voltage to 1.5 (slowly) using the knob right next to the on switch. THIS TURNS ON THE QUARTZ LAMPS. YOU DO NOT (SHOULD NOT) ALSO TURN THEM ON USING THE QL/QH buttons in flatfield.tcl window.

  2. ExpType = Object

  3. Object = name direct mask image

  4. Make sure "Full" image is chosen (not subraster).

  5. Make sure the screen is IN (always tell the TO when you insert the screen).

  6. Make sure the guider probe is out (CF-guider = out).

  7. Make sure that the right mask is in.

  8. Make sure the disperser is set to f/2 imaging. You don't want dispersed light at this stage.

  9. Make sure the f/2 filter is set to R  (you should not care about or use the f/4 filter).

  10. Make sure the hatch is open and the mirror is visible.

  11. Take the images using START.

  12. Do this for all of the masks for the night.

Once you have an image with the right counts in an acquisition box (~3000), note the name somewhere. Also note which CCDs have acquisition boxes. Then, open a CCD image (each fits file is a CCD) containing an acquisition target via DS9 on the other (non-instrument) desktop (control -> gets you to the other desktop). You will need to visually examine the images and look for the BOXES which are the acquisition holes in the mask. You want to measure how wide they are (on my run, they were 30-35 pixels).

Now, run epar icbox (in the IRAF window on the non-instrument desktop). Mask is the name of the file you just took without the cN (e.g., ift1010). Out is the name of the field (I use the mask name  -- e.g. 08134050). In the "sz" box, enter the size of the acquisition box that you had measured.

When running icbox, find the acquisition objects and mark them with the space bar. If a CCD does not have an acquisition object, hit q. Hitting "q" after the last CCD quits the routine. When you are done, two new files are created that you will use later: name.align and name.sub. Ignore objects that are in CCD gaps or are within 100 pixels of a CCD edge. You'd better have at least 2! 6 is best.

Don't forget to turn off the lamp when you are done. First, turn down the voltage knob (slowly) and then click the switch to the off position. You can leave the screen in if you go directly to the next step.

Note: The reason these directions state to use the quartz lamps in the dome is that their voltage can be controlled directly. The Quartz lamps from the GUI (see next step) have voltages set so high that the direct mask image will saturate. There is an image attached (see bottom) which shows this.  direct_saturated.jpg  left shows what happened when I tried to use the quartz lamp from the GUI to take a direct mask image. The pixels are saturated. On the right panel, you can see how the direct image should look. There are ~3000counts maximum in the slits. 

(2) Take flats. You will need these to map the slits on the CCDs for extraction in the reduction process. In this case, you want enough dispersed light on the CCD to see the dispersed slits clearly.

• 1. Use the Quartz lamps in the special lamps window (called flatfield.tcl--NOT the Cal-Lamp in the IMACS window). Use QL or QH (low or high, depending on your counts and exposure time. I used QL and 10s exposures to reach 10k counts).

  2. ExpType = Object

  3. Object = name qflat

  4. Filter = spectroscopic

  5. Or, use a nice blocking filter if you want (see below).

  6. Make sure that the right mask is in.

  7. Make sure the screen is in (always tell the TO when you insert the screen).

  8. Make sure the disperser is set to your grism (200l)  (previously it was set to f/2 imaging).

  9. Make sure the guider probe is out (CF-guider = out) and guiding is off.

  10. Take a 10s exposure using START. You want counts to 10k. Click on the QL-Tool image (in the instrument desktop) and a small Chip zoom will show the zoom and the QL-Tool should show the min/max count. Remember that the highest flux is at the center of each dispersed slit (center lengthwise).

  11. Take 3-5 of these using START. Don't  use the LOOP function though. Take them one-by-one.

  12. Do the above for all the masks in the night.

Note on the use of a blocking filter: I played with using WBXXXX-YYYY and also the SDSS filters (which have sharp edges in wavelength). They make a BIG difference and will hopefully help the data reduction process immensely. Note that the SDSS filters are only 70% efficient (and narrower) than the near 100% efficient blocking filters. So better to use the blocking filters. As an example, for the mask with a cluster at photoz = 0.75, the H&K break is at ~7000angstrom. So I used the WB5600-9200 filter. If the cluster is at z=0.5, I would use the 4800-7800 blocking filter.

When you are done, turn off the lamps and move the screen out. Then go to dinner. Come back when

the sun sets. Below, there is an image attached which shows the dispersed flats versus the direct image of the mask. You should be able to see how the slits + grism are dispersing light onto the CCD. direct_dispersed.jpg

(3) Take 10 bias frames. Zero exposure time. Set ExpType to BIAS. Use 1x1 binning.

(4) When you are back from dinner, take the first set of arc lamps for each mask. The idea is to have a solid set of 6 per mask at the beginning of the night. Then take one short arc after each science exposure. These first set contain 3 short (5s) and 3 long (20s) exposures. The problem is that the difference in brightness between Ne versus He and Ar is huge. So the short exposure gets you the Ne lines and the long exposure gets you the Ar and He lines. This of course requires that you stitch things together properly and make a super comp frame. The short comps after the science exposure are just to keep you honest.

• 1. Use the lamps in the special lamps window (called flatfield.tcl--NOT the Cal-Lamp in the IMACS window). Select one Ne, two Ar and two He. If you select two Ne a large number of arc lines will saturate the CCD, even with only one Ne lamp on some of the Ne lines will still saturate the CCD at 5s. But there will be enough to work with.

  2. ExpType = Object

  3. Object = name comp

  4. Filter = spectroscopic

  5. Make sure that the right mask is in.

  6. Make sure the screen is in (always tell the TO when you insert the screen).

  7. Make sure the disperser is set to your grism (200l)

  8. Make sure the guider probe is out (CF-guider = out) and guiding is off.

  9. Take a 5s exposure using START. You will see the lines show up for each slit. The red means saturation and is okay.

  10. Take 3 of these using START. Don't  use the LOOP function though. Take them one-by-one.

  11. Now, increase the exposure time to 20s and take 3 more.

  12. Do the above for all the masks in the night.

When you are done, turn off all of the lamps and take the screen back out.

(5) When it is dark enough ask the TO to point the telescope. The ask him to go to your first spectroscopic target in your observing plan (you did send the file over, right?) You now want to take the first of many acquisition images. You will need to do these steps each time you move to a new field with a new mask. The first step in the process is to get an image of your field on the sky (a pre-image). You will use this image and the image you took in Step 1 to start the alignment.

• 1. Make sure that the mask is out and the lamps are off.

  2. Make sure you are guiding.

  3. ExpType = Object

  4. Object = name acq direct

  5. Mask = f/2 imaging

  6. Disperser -> f/2 imaging

  7. Filter = R

  8. Take a 10s exposure, use FULL (not the subraster).

When the image is read out, you should see some stars, hopefully you acquisition stars. If not, use a longer exposure. Now, switch to the other (non-instrument) desktop and run epar ialign in the IRAF window. The mask file is the image of your undispersed slitlets (step 1). Field should be the name of the image you just took (without the cN). Align is the root name of the out files you created in Step 1. SZ is the same size that you used in Step 1.

Follow the directions on the lco.cl IMACS page on how to use ialign. The idea is that you want to get the stars marked. You can use "space bar" to let the code centroid the star (but make sure the cursor is in the star itself). If for some reason you need to manually force a center, use the "m" where the cursor is. The "a" key will automatically use what is already marked as the circle (green). You can reset the z limits of the image (to separate double stars, for example) by hitting the "d" key (followed by say 500 for z1 and 40000 for z2).

Ideally, you have 6 or more stars to choose from and you can throw out any bad apples (i.e., doubles, high proper motion, non-existent, etc). You need at least 2 to make a solution. When you are done, type q. If a window pops up with a green grid in it, ignore it an press "q" again.

On the IRAF terminal, you should see a bunch of output. The things your care about are the Mask scale (it should be very close to one) and the three offset numbers. Which can be anything, but hopefully not far from 1 or 2.  Click "no", don't send the offsets to the TO.

If you made any decisions during the process of marking the stars, go back and do it again and make the other decision (if it is reasonable...e.g. the other star of the double). Compare you original values to you second (or third or whatever) values and choose which solution has the lowest scale and smallest offset (this is probably the right solution).

Once you know which solution you want, run ialign one last time with those decisions and send the offset to the TO (you will need to say yes multiple times and hit return at the end). TELL the TO that you are sending the offsets BEFORE you send them.

(6) Okay, almost there. Now you need to take another acquisition image. This time, make sure the mask is in. Otherwise, it is the same as before. The goal here is to look at the image of the mask and actually see the stars shining through the acquisition boxes. So you do not want dispersed light here.

• 1. ExpType = Object

  2. Object = name acq mask

  3. Disperser -> f/2 imaging

  4. Make sure the right mask is in

  5. Filter = R

  6. Make sure you are guiding

  7. Take a 10s exposure, use FULL (not the subraster).

Open up a CCD using DS9 (displ in the IRAF terminal). Make sure the CCD you are looking at in DS9 actually has an acquisition object in it. Hopefully, you can see the star through the acquisition opening and it is centered. You then run the "ifalign" script to make any new small offset changes. Supposedly, you can make big changes, like re-center yourself using the ifalign. But I couldn't figure out how. So I run ifaligh and just hit "q" a bunch of times until it kicks out a new set of offset parameters (and asks whether to send those to the TO. As with ialign, the full directions are on the LCO site.

At this point, your masks should be aligned properly and you can take data (i.e., put the disperser in). If your alignment stars are pretty close to being  centered, you should be okay. I don't know how perfect "perfect" alignment is. You can go back to the ialign stage, or even as far back as the icbox stage to make things more perfect.

Recall that you need to do the alignment for EVERY mask. You should also take a 5s comparison frame AFTER each exposure.