Observing at Gemini
In October 2017 we were privledged to be "Priority Visitors" at Gemini-N, and were trained on how to carry out queue observations with NIRI, GNIRS, and GMOS. The main reference is The Gemini North Visitor Observer's Guide. (A private copy is squirreled away here.) Although the Guide is very useful, there's a lot that was left out. This page is intended to serve as a re-introduction were we to ever get to go back!
On the wall there are a number of extremely large screens that allow you to judge the weather conditions. One of the most useful is a set showing the views in various directions as well as a "difference movie" of the sky, which allows you to judge the cloud cover. CC50 means it's photometric--no clouds! CC70 means there are some thin cirrus around. In addition, there's a plot showing what amount of water vapor is present, as some NIR programs place WV demands.
But most of what you're looking at are the 8 large computer monitor screens. We never actually ran the startup script "startup astro"; things were always up and running when we arrived. In addition to the 8 screens, there are multiple desktops for each of them. You can't move a window from one screen to another but you could kill it and restart it by going to the little "G" icon located at the (bottom-left?) on each of the monitors. Here's what we found to be the typical layout:
- Left top: This was the "guider" window, and had three plots.
- The top plot showed how well the astigmitism was being removed. When you slewed to a new object it would take a couple of minutes for the astigmitism to be tuned out; you need the red line and the green line to meet in the middle near 0 before starting an acquisition.
- The middle plot showed the seeing (green line) and not-very-useful focus value (red line). You could (should) autoscale the plots by clicking on the little squiggle button at the bottom of the plot twice.
- The bottom plot was something that I'm sure was very interesting, but which we never actually used.
There will be a menu window to the right of these plots which allow you to select which guider you're using. If you're using NIRI or GNIRs, you need to select the "P2" guider and then hit "BOTH". If you're using GMOS you need to select OIWFS and hit "BOTH." At the very start of the night when you're using ALTAIR you need to select (wait for it) ALTAIR and hit "BOTH."
- Left bottom: This was the Telescope Status Display (TSD), and oddly enough was probably the least useful window. In principle we were supposed to make sure that we weren't near a rotator wrap problem, and so on, but the operator did a great job of all of that. It's useful for seeing what the UT and LST are, although the layout is so busy that it was hard to pick these out easily.
- Top column 2: This was the home of the Seqexec, the thing that actually runs the observing sequence. More on this in a bit.
- Bottom column 2: This is where the Observing Tool (OT) would live. This is how we would select programs and observations to queue to the Seqexec. We would typically have two open, with different observing programs in each. If we had been really clever we would have the one about to finish in one, and the new one in the other, so that we could do the quality assessment "pass" (or not) without having to lose track of the new one.
- Top column 3: This was the cartoon display of what was happening with each instrument, the Instrument Status Display (ISD). I would occasionally glance at this to see what the hold up was in actually taking some data for an aquisition. Usually it was the slit moving in or out of the beam.
- Bottom column 3: This is where we would have Pyiraf running, along with ds9. This was used both for running the acquistion software (gacq) and for checking each frame (gdisplay for GMOS, ndisplay for the two NIR instruments).
On a second desktop we would keep the quality assement tool (QAP), seqplot, and so on.
- Top right: This was probably the window I spent the least amount of time in. I don't really remember what we had there--maybe the queue plan?
- Bottom right: Oh, boy. This is the night log. Gemini takes that night log VERY seriously. Details are given in the guide. This is also where we would keep the Firefox window open with its many useful tabs, as we would have to copy and paste the description of each program into the start of the log.
Typical Observing Sequence
Observing at Gemini was probably the most intense observing I've ever done. There's a ton of stuff going on.
- Some time before the current exposure ends, figure out what you're going to do next.
- What are the conditions?
- Check the screens for clouds (CC condition).
- What does the guider say in terms of the seeing (IQ condition). Note that there's this whole business that the IQ is actually based on the seeing corrected to the zenith. There's an IQTool (somewhere) that let's you convert to the zenith. There's also the folklore that the guider seeing can't be trusted except for ALTAIR; the actual IQ is
worse(?) by about 0.2".
- Is it "wet" or "dry"? Check the WV plot on the wall.
- The SB (sky brightness) was the least worrisome when we were observing as it was new moon.
- Consult appropriate queue plan. If you are thinking of moving things around, try them and see if they turn red.
- At this point you should have a program name (Q-20) and a science observation number . Find this in the text queue plan, which will be open on the bottom right-most window. Copy and paste the "----Q-20---B1--blah-blah-blah line into the night log.
- Find the program in the OT. Once you have the program open, you can do a Search (bottom left of the OT) for the sequence number . Select the science sequence () and the associated acquistion sequence (let's say ) and hit queue to send it to the Seqexec queue. One thing not clear to me is at what point we pick out the associated daytime calibrations to send over. Maybe it was now. But these will be one per program, typically, not one per observation. Note that queue will do an automatic sync.
- Select the science exposure and hit "Image" in the OT; this will give you an idea of what you're going to be dealing with.
- Wait for the current exposure to finish, and then confirm that the image is okay before going anywhere. See the Data Quality Assessment below.
- Ask the operator to slew to the SCIENCE target (). If you've been keeping the Seqexec queue clean there'll be only one  but you might want to be really clear and say "Q20-45. If you are going to a blind offset (so that there's a USR1 and USR2) or a two star acquisition (again, a USR1 and USR2), tell the operator. For a two star they'll go to the designated middle position but for a blind offset they'll need to go to one or the other. I think the offset position is always USR1.
- While the telescope is slewing....
- Mark the final image (and any calibration images you haven't) as "pass" in the OT. [Arguably this can be done later.]
- LOAD the acquistion sequence into the seqexec by doing "File-->Load sequence from OCS" and pick the acquisition sequence .
- Click on the first step and hit "INST". This will start configuring the instrument.
- Examine the acquistion sequence and decide where to put pauses. For instance:
- GMOS long slit: Usually this starts with a field image, followed by an image of the slit. (The slit image will have a P=10 Q=0), followed by two through-the-slit images. You would want to put a pause on the the first through-the-slit image so that you could run gacq using the first pair, and a pause on the other through-the-images. In other wors, whenever you are going to need to run gacq, you need a pause.
- GNIRS. If the object is bright, then the typical sequence is slit image, field image, through-the-slit image, and the same idea applies as with GMOS. But if the object is wicked faint, then the sequence will be slit image, field image, sky image (with P and Qs), then through-the-slit object, through-the-slit sky, and you want to make sure that the pauses go after the object/sky pairs.
- Maybe about now the telescope is in position but you're still waiting for the astimatism to settle. This is a good time to update the observing log, trying to guess when you actually began the slew. ("Time Slew to ")
- Bored? You could try looking at your most recent image and indicating if it's a "pass" or not in the Data Analysis section of the OT.
- Okay, assume that the astigmatism is converged. Check that the instrument is done configuring by glancing at the cartoon and making sure nothing is happening. Now press "Confgiure from current TCS setup and then "Run."
- This should take the first few acqusition steps. You now need to run gacq in the Pyiraf window using the image number that corresponds to the field image, not the slit image. So for GMOS that would be the 1st image, and for GNIRS that would be the 2nd image.
- gacq may be the most amazing piece of software ever written on this planet, or any other. As soon as the first pair of acquisition images (or N, depending upon the intrument) run gacq nnn, where nnn is the frame number of the frield eposure. It will then do something highly appropriate, display the image in ds9, and drop you into imexamine. If you centroid the star, use "r"; if you want to use the cursor position use an x. If you are doing a two star on the slit acquisition do a "gacq nnn two+".
- The first pass will produce offsets. Send them over to the operator, and then announce, "I've just sent you offsets." The operator should acknowledge and tell you when they've been applied.
- While waiting, copy the offsets and paste them into "COMMENTS" tab on the observing log in the OT for the Acq, picking the appropriate line.
- Hit CONTINUE (not RUN). The reason for this is really just good hygene--if you skip some steps, RUN would go back and execute them. CONTINUE always continues.
- Repeat the gacq now on the through-the-slit or whatever, using "x" or "r" as appropriate, sending and copying coordinates, and so on until gacq says enough.
- If you still need to do another step, quickly add the additional one (by changing the number of "observe"), sync, queue [all on OT], and then "reload sequence" on the seqexec. Then continue.
- Once you're done, load the science exposure  from: file --> load from OCS.
- If this is a blind offset, please please please rememeber to tell the operator to go to the "other" (science) position.
- Record the time of start of science data in the log as "HH:MM Taking data." Log is always in HST.
- Clean up anything that needs to be cleaned up in the previous observation (the one you slewed from).
- Mark the acq images in the current sequence as "passed". This is under the data analysis tab in the observing log section.
- Whew! Breath!
Data Quality Assessment
A big part of the Observer's task is to assess the quality of the data. What this really means is that you look at every single frame and make sure it's not awful. For instance, GNIRS has this unfortuante tendency to develope a "bias flash." If you're very lucky two adjacent exposures will cancel (there's persistence), but you need to look at every frame.
After you're satisified give each image a "pass" rating under the analysis tab in the observing log in the OT.
If the image is okay but (say) of the wrong object, give it a "usable."
- GMOS: do a gdisplay nnn.
- NIRI: do a ndisplay nnn, or maybe a ndisplay nnn sub-
- NGIRS: After a pair have been taken, do a ndisplay nnn clean+. If it screws up identifying the right sky do a ndisplay nnn clean+ sky=yyy If you want to check out a single frame, do a ndisplay nnn sub- clean+
- If you have a series of something (say 10 twilight flats in the morning you need to check) you can run seqplot. This is an IDL task and you need to find the IDL window, maybe in a 2nd desktop on the bottom 3rd screen. The format IDL> seqplot,20171021,429,999 to display each image starting with 429 and going through the final one. I think a space bar moves between them.
At the end of the night
Maybe you need to do sky flats. I don't remember how you find those. When all is said and done you need to load the daytime calibrtions which hopefully you've queued up as you've gone along; you now need to add the "bias all" stuff. There's a "load a bunch of stuff" object in seqexec. You want the biases to be the last thing for reasons I never did understand.