Version 0.1, 15 Nov 2006
Get the MMTF out of storage. It should be in the ground floor storage room in the Baade telescope, either in the cabinets or in its big metal shipping box.
(Optional, recommended for the day before installation) Use the spare controller and cable to connect MMTF, balance, and verify operation with the etalon on a mercury light box. This is a good place to familiarize with the balancing procedure and practice the parallelism-by-eye tests. Note parallelism numbers for X and Y fine/coarse.
Note that the cables between CS-100 and MMTF should never be plugged/unplugged when the CS-100 power is on. To plug/unplug: release computer control if necessary (quit the IMACS mechGUI if it is on); switch the CS-100 from "operate" to "balance" mode; turn off the CS-100 power; then unplug the cables. (*** Instructions for balancing and parallelizing also needed here; see below. Also pictures.)
Balancing: You can use the X,Y,Z balance settings from previous runs as a starting point. (*** give settings here ***) The X, Y, and Z settings control the spacing between the etalon plates in X tilt, Y tilt, and Z piston. In balance mode, you are balancing out the capacitance bridges in the feedback loop. In operate mode, changes to the front panel knobs change the position of the etalon plates.
Follow the instructions in the CS-100 manual (*** summarize here ***). Turn on CS-100 in balance mode with the meter knob on offset, and adjust the X, Y, Z coarse and fine controls to null the meters. Then switch the meters to quadrature error and adjust the quad balance knobs to null the meters. Iterate: switch the meters back to offset and re-null with coarse and fine; then switch to quad error and re-null that. Finally switch the CS-100 to operate mode and make sure it operates (doesn't go out of range). Write down the knob settings.
For general operation at a relatively broad resolution, put the Z coarse to something around -1 (check?) Please do not go below Zcoarse = -2 without checking with an MMTF team person. You will often find that switching a full step in Zcoarse causes the MMTF to go "out of range." It is better to switch to balance mode, turn the Zcoarse knob, then switch back to operate.
Parallelizing: You can use the X,Y,Z parallelism settings from previous runs as a starting point. (*** give settings here ***)
Turn on the green mercury lightbox and set the MMTF on top of it. Set the CS-100 to operate mode. Look down through the MMTF at the lightbox. You should see a green ring. If you don't see it, try twiddling the Z fine knob to bring the ring in/out in radius.
If you move your head side-to-side while looking down through the etalon, you should see the ring change in size (easiest to see if the ring is fairly small). This is because the plates are out of parallel and further apart on one side. Your goal is to tweak the X and Y coarse/fine adjustments to even out the change in size. (The ring changes in size from center to edge of the plates because the coatings aren't quite flat. You can't take this out; you just want to make the ring the same size on opposite edges.)
To do this, you adjust X and Y independently. The Y axis is the direction along which the cables enter the etalon. Move your head along this direction to see the change in ring size and tweak the Y fine knob to attempt to take it out. Once you are close in Y, do the X direction.
(*** need some pictures of different size rings here.)
When done with parallelism, write down the X,Y,Z settings.
(*** Need to have a worksheet with blanks for stuff the user should record: various X,Y,Z settings, FP calibration, location of optical axis ... )
The plastic covers should be on whenever possible, to avoid touching glass surfaces. Mount on mounting plate, if not already done. This takes attaching 3 screws from mount plate to MMTF, and then attaching 3 brackets to the plate (You do not have to actually screw the brackets to the MMTF. The screw holes at the top of the brackets will be used later to mount a pupil mask.) Once you get the MMTF onto the plate, tape the covers back on; you should leave the covers on while it is being handled and mounted in IMACS.
Note that there are several possible orientations. With the MMTF on top of the mount plate, the MMTF should be mounted with the cable sockets 120 degress clockwise from the two plate "ears" that are close together, and it should be on the plate "upside-down" - with the MMTF on top of the plate, the writing on the cable labels should be upside-down. The MMTF optics are not sensitive to its orientation, but this way of mounting is needed to provide extra clearance for the largest cable connector. See pictures of the correct MMTF mounting orientation to mount plate.
Find MMTF cables at disperser wheel position number ?? - they may be tied off to the side when not in use - and free them. Mount the plate to pins on disperser wheel and secure. In disperser wheel service position, the MMTF cable sockets should be pointing out and down (about 4-5 o'clock as viewed from rear of instrument looking toward telescope). Insert cables into MMTF. See pictures of the correct cable orientation as installed in IMACS.
The blocking filter(s) you wish to use should also be installed now.
Once the MMTF is cabled, turn on the CS-100 in the IMACS electronics rack on the Nasmyth platform. Go through the balancing procedure described above, iterating offset/quad balance. Once balanced, write down the knob settings. The correct settings may be different from the ones done outside the instrument, because they can depend on the CS-100 and the cable.
In June 2006 the parallelism settings for the MMTF as installed in IMACS were:
X -1 7.25 4.21 Y 0 6.96 3.74 Z +1 1.59 4.27Switch to operate mode and set the X,Y,Z coarse/fine to the approximate positions desired - the parallelism from previous runs, and the desired Z coarse based on the filter passband width that the observer wants to use.
(Verify computer connectivity?) Once the cs-100 in the rack is powered on, you can start the IMACS software to make sure that the IMACS GUI is talking to the CS-100. However, you will then need to quit the software to release control to use the front panel knobs in the next step.
To record the basic CS-100 panel settings, someone with the instrument specialist password must run "imacs setuptool". Hit the MMTF button and enter the XYZ coarse/fine/quad bal settings, then exit. The user can restart the mechGUI from the modules menu in camGUI, or "imacs mechgui" on the command line.
Take green mercury lightbox and slip it behind the MMTF while it's in the disperser service position. (*** We need some way of hanging or clamping it there.) At this point, you should still have a clear cover on the MMTF for protection.
You'll need two people for this procedure. One person stands at the CS-100 in the electronics rack to adjust the X and Y coarse/fine dials while the other person looks through the MMTF, moving his/her head back and forth and judging how the size of the rings is changing. With the MMTF in the disperser service position, the X and Y axes are not horizontal/vertical; remember that the Y axis is always parallel to the direction that the cables enter the etalon. When done, record the X,Y,Z settings.
After the parallelism by eye, you can remove the lightbox and remove the plastic covers. Then install the quadrant pupil mask for further parallelism testing and close IMACS. The quadrant mask is screwed to the three mounting brackets. Make sure to note which orientation the open quadrant is in, relative to where the cables enter (noting whether it's on X or Y axis and which side).
Copy MMTF software onto observing computer if necessary. Set path to include the MMTF software directory (*** where?) Run (*** some script ???) to set the PGPLOT directory so it can find fonts.
The MMTF produces an image whose bandpass varies from center to edge of the field (edge=bluer). The gradient is about 100 A at H-alpha. Adjusting the plate spacing with the Z parameter tunes the wavelength. Because you only see a short section of the spectrum in any given image, and because a line can be transmitted in different orders, the challenge in setting up the instrument is to identify which line is which.
For this reason we want relatively few lines transmitted by the filter and use one of the available gas-discharge lamps (neon, krypton, xenon, argon, helium?) at a time. Use a lamp that has a few lines in your filter: neon for the 6600/260 filter, krypton or xenon for the 8150/150 filter.
(*** have a table of line wavelengths accessible here, and plots of the grism spectra of the lamps ***) We have rough linelists for the calibration lamps.
Insert the appropriate blocking filter into the IMACS beam. We will take a series of exposures, stepping the etalon through an entire free spectral range (order) and measuring flux at pixel X,Y as a function of the Z (spacing) axis of the data cube to build up a spectrum. To speed up this process, only read small sections of the CCD array. Define a subraster file (*** click on ? box in the IMACS gui ***). Make a subraster on chip 2 from x=1900-2000 and y=3900-4000, which is near the optical axis, and another on chip 1 from x=100-200 and y=3900-4000, far from the axis (and others to be at a range of radii, but only do one per chip).
Take a test exposure (*** length? probably 1-5 sec for the IMACS internal lamps?) to check flux levels. Then take a loop over somewhat more than a full FSR. A full FSR is about 700 in Z at H-alpha (and scales with wavelength). Do a loop stepping by dz=20 with ~50 images (*** instructions for taking a loop and changing Z with an IMACS script here ***).
The spectrum through the Z axis is called a "sausage cube." To plot and fit the spectrum, use the program "fitsausage". First make a list of the input images with "makeimlist". If your cube started at image 30 and went to image 79 and you want to use the subraster from chip 2, type
makeimlist 30 79 2 list.cube1to put the list of the fits files in "list.cube1". Now run
fitsausage list.cube1 /xs [hit return to accept the default box to sum pixels in] [hit return to skip fitting the peak] [hit q to quit]This will plot the spectrum in flux as a function of Z (defaulting to averaging the flux in a box section of x=10-30 and y=10-30 out of the ~100 pixel square subraster that you set up). If you type /ps instead of /xs you can get a postscript file that can be printed out.
The spectrum over a full FSR should show one or more emission line peaks. Compare to example MMTF spectra to identify lines (see below). Currently examples are only available for the 6600/260 filter.
fitsausage also will fit data in an x-axis range that you specify and print the best fit parameters. The fit is a Voigt profile and the parameters are:
First, finish parallelizing the etalon. There are two approaches to this; both involve tweaking the X and Y settings and taking short scans across a spectral line to measure the effect.
6a. Parallelism by quadrant mask
If the etalon plates are not parallel, the gap on one side of the etalon is larger than the other. This means that it passes light at a slightly different wavelength; the wavelength-Z relation is different from one side to the other. The quadrant mask tests this by allowing light to pass through just one side of the etalon at a time; you will tweak X and Y and measure shifts in the Z-location of a spectral line. You will need a person on the platform and one in the control room.
Rotate the disperser wheel so that the etalon is accessible, reach into IMACS and set the quadrant mask to the Y-direction (open quadrant near the cable entry point). Close IMACS and rotate the MMTF into the beam. Take a short loop of (subrastered) images with the CS-100 dials at the nominal X and Y positions determined earlier. Run fitsausage to make sure you covered the line peak. Use the fit option of fitsausage to measure the Z of the peak.
Now have someone adjust the Y fine dial down by 50? units. Take another scan loop across the line. Adjust Y up to nominal+50? and take another loop. Rotate the disperser wheel to service position, open the panel, and rotate the quadrant mask by 180 degrees. Now do another set of nominal Y, -50, and +50 with a scan and fitsausage at each position. (For extra care you can do -100 and +100 also.)
For each of Y=nominal, -50 and +50, compare the peak Z values from the two quadrants. These will be near equal (*** to what tolerance?) if the etalon is parallel in Y and offset if not.
After determining the proper Y setting, repeat with the mask at the two X positions while adjusting the CS-100 X knob to nominal, -50, +50.
When you are satisfied with X and Y settings, note them, remove the quadrant mask, and close the instrument.
This method is faster since it doesn't require opening the instrument, but a little less precise. Remove the quadrant mask. Take a short scan across a line, stepping by dz=10? You need to sample the line well to measure the width. Pick a line that is not blended. Use the fit option in fitsausage to get the width.
Now try tweaking X down by 50 fine units, and take another scan. Then up by 50 and another scan. You may want to try more than three settings, e.g. X = -100,-50,0,+50,+100. Measure the line width in each scan. Wider widths mean the etalon is out of parallel. You will also see that the line profile becomes skewed when further out of parallel. Sometimes a skewed profile will return a formally narrow width, but beware - the skewed profile indicates non-parallelism.
After settling on an X value, do the same procedure to determine Y. (*** Give a nominal FWHM for a given Z-coarse setting; depends on wavelength range, also.)
When happy with the parallelism values, you are done with adjusting the CS-100 front panel. Have an instrument specialist run "imacs setuptool," hit MMTF and enter the X,Y,Z settings so they are recorded in the image headers.
Wavelength is a nearly linear function of Z, BUT when the spectrum passes through an order jump, it repeats itself. Doing the line identification is subtle because lines can be interleaved from different orders. Both the resolution and the FSR (order spacing) depend on the Z-coarse CS-100 setting.
Adjusting the parallelism will have moved the zeropoint of the wavelength-Z relation, so start a new full-FSR scan (40-50 images at dz=20 or so). Meanwhile, plot the full-FSR scan you took back in step 5. If you are working in a filter and resolution that has been done before, compare the spectrum to a previously IDed one. Example spectral scans for the 6600 A filter are available. You are looking for a familiar repeating pattern of wavelengths (keep in mind that the relative intensity of the lines can vary from one lamp bulb to another and from year to year).
If you are working in an unfamiliar region, try to use linelists and the grism spectra of the lamps (*** need link) to identify lines. Keep in mind that lines from different orders can be interleaved if they are within the blocking filter's bandpass, and that very bright lines can leak through in the few-percent filter tails.
You can use the "specsim" program to make crude simulations of what the spectrum should look like for a given set of parameters: a list of lines and strengths, blocking filter range, etalon FSR and resolution, and zeropoint and slope of the wavelength-Z relation. This requires some amount of guessing, but is useful for predicting the interleaving of lines from different orders. The FSR and the slope (dw/dz) are roughly predictable for a given wavelength and Z-coarse setting (*** Give a table for H-alpha, where we know this already - see Appendix.)
Once you have identified several line peaks, you have several wavelengths each with a Z and order. If you measure lines in more than one subraster, you also have a range of different radii from the optical axis. These can be used to find the parameters in the Fabry-Perot equations:
W(z,R=0) = A + B*z W(z,R) = W(z,R=0) / sqrt(1 + R^2/F^2) A = zeropoint of wavelength-z relation B = dw/dz R = distance from optical axis (in pixels) F = focal length of camera (in pixels)As a quick solution to get on the sky the first evening, you can just find two lines (in the chip 2 subraster near the optical axis) of known wavelength and Z to get a crude estimate of A and B. For example, if Ne 6598.95 is at z=2674 and Ne 6678.3 is at z=2887, then B=dw/dz ~ 0.3725 and A=5602.9.
Generally dw/dz is predictable from the wavelength you are using and the Z-coarse setting (*** need table, see appendix), and F ~ 23670 pixels.
For a better long-term solution, you can use the program "fpsolve" which takes a number of wavelength and z pairs and solves for the parameters. fpsolve can also handle lines that are in different orders and at different radii, if these are specified in the input file. (*** need more information on fpsolve input and output)
(*** need to specify how to find the radius given chip#, x and y)
The wavelength gradient in the MMTF is due to the angle at which rays pass through the etalon. It is circularly symmetric aboout the optical axis (the projection of MMTF normal axis onto the CCD). To find the optical axis, use the fact that there is a faint ghost reflection between MMTF and the CCD. Insert a slitmask into the focal plane - any mask with star boxes will do. Take an ? second exposure of flat lamps. There will be faint reflections of the star boxes around the optical axis, which is near the center of the CCD array. Record the locations of the centers of star boxes and their ghosts in chip#, X and Y, and put these in a file, one box and ghost per line:
box_chip# box_x box_y ghost_chip# ghost_x ghost_y ...
Then run "findghosts" on this file. findghosts has a model of the CCD mosaic geometry and will find the midpoint between each box and its ghost, and make a plot. If the midpoints don't agree to within a few pixels, you probably misidentified a ghost. Record the optical axis location (note that it is possible for it to be between chips).
understand relation between wavelength, radius, Z fpcalc program. You can use the web version of the Fabry-Perot calculator to solve the Fabry-Perot equation, e.g. to compute what Z you need to set to make the desired wavelength come in at a given radius, and so on.
will need a spectral cube and flats
*** May move description of observing to separate page.
focus procedure is normal for imaging.
tuning etalon to desired wavelength take short exposure to make sure you are getting flux from object (if it's a known emission line source)
use fpcalc or web page calculator to figure out wavelength, z, radius
taking periodic wavelength calibrations - single ring exposure (could bin it) or short subrastered sausage cube
flats, biases, wavelength scans
how to do flux calibration sources - both scanning an emission line, like a planetary nebula, and taking images of a spectrophotometric standard star
*** May put this on yet another page. Instructions for using the various scripts that write IRAF scripts to do overscan and bias subtraction, flatfielding, while handling all 8 ccd subimages. *** Need to have some software for wavelength calibration book-keeping. Describe operation of the ring subtractor program, skyring2mosaic.
Fabry-Perot equations:
On the optical axis (R = 0), for effective plate spacing d:
d = N * lambda / 2 FSR = lambda / N finesse = FSR/FWHM lambda = A + B*zN = order number, lambda = wavelength transmitted, finesse depends on the reflectivity, flatness and parallelism of the plates. A and B describe the wavelength-z calibration.
Off the optical axis at radius R:
lambda(R) = lambda(0) / sqrt( 1 + R^2/F^2 )where F is the focal length of the camera.
At a given plate spacing, the FSR varies as lambda^2, thus if the FSR is 300 A at 6600 A, it should be about 460 A at 8200 A. This is not exact because the coating depth varies with wavelength, changing the effective plate gap.
Note that z is really controlling the plate gap, i.e.
d = E + G*z, E and G are independent of wavelength, where A = 2E/N, B = 2G/N;A and B depend on the order you're working in.
Since B is what we also call dw/dz, this implies that dw/dz ~ 1/N, so at a given plate spacing dw/dz is proportional to wavelength, i.e. if dw/dz is 0.30 A/z-unit at 6600 A, it should be 0.37 A/z-unit at 8200 A.
CS-100 settings for balance and parallelism on previous runs.
(*** need to fill in here ***)
Measured MMTF FSR and dw/dz at 6600 A for a range of Z settings on the CS-100. These were measured in the lab at OCIW with a different CS-100 and cable and etalon illumination, so the FSR and dw/dz may be slightly different in operation and the FWHM may be significantly larger. The FSR and dw/dz should be a decent start for wavelength calibration.
For wavelengths other than 6600 A, the FSR in A scales very roughly as lambda^2 and dw/dz in A/z-unit scales proportional to lambda. So the FSR in z-units is proportional to lambda in A. (This is because moving up by one FSR is increasing the order number by 1, which means increasing the plate spacing by lambda/2).
# zc=zcoarse, all at 6600 A, wavelengths in A # #zc zfine dwdz fsr fwhm order-num # -3 3.00 0.536 375 18 17.6 -2 0.20 0.469 325 15 20.3 -1 0.20 0.413 285 13 23.2 0 0.20 0.361 248 9.7 26.6 1 0.20 0.33 226 8.9 28.9 2 0.20 0.294 202 9 32.7 3 0.20 0.268 185 7.9 35.7 4 0.20 0.251 173 6.7 38.2 4 9.99 0.236 163 5.9 40.5
As a rough example, translating from working at 6600 A to working at 8200 A:
# zc working-wl order-num dwdz(A/z) fsr(A) fsr(z-units) -1 6600 23.2 0.413 285 690 -1 8200 18.7 0.513 440 857 (approx)
Useful calibration lines. Now at calib_linelists.html.
Contact:
Benjamin Weiner, bjw at as.arizona.edu, 15 Nov 2006