Imaging an EVLA OSRO HI data set: Difference between revisions
|Line 147:||Line 147:|
Revision as of 19:42, 12 March 2010
This tutorial explains how to image an HI dataset acquired with the WIDAR0 correlator. It assumes that you've already calibrated your data as described in the calibration tutorial, and that you now have a split dataset with a single source of interest in it. In this example, the source is called 'Leo-2'; see the calibration tutorial for more details on this data set.
Flag Your Split Data
For a spectral line dataset like this one, you'll probably want to average in various ways to spot bad data. Averaging channels together can make bad baselines pop up!
Presently, doppler tracking is not performed online by the EVLA, so we have to correct for any velocity shifts now, in post-processing, with cvel.
# cvel :: regrid an MS to a new spectral window / channel structure or frame vis = 'leo2.ms' # Name of input measurement set outputvis = 'leo2_cvel.ms' # Name of output measurement set passall = False # Pass through (write to output MS) non-selected data with no change field = '' # Select field using field id(s) or field name(s) spw = '0' # Select spectral window/channels selectdata = False # Other data selection parameters mode = 'channel' # Regridding mode nchan = -1 # Number of channels in output spw (-1=all) start = 0 # first input channel to use width = 1 # Number of input channels to average interpolation = 'linear' # Spectral interpolation method phasecenter = '' # Image phase center: position or field index restfreq = '1420405751.786Hz' # rest frequency (see help) outframe = 'BARY' # Output frame (''=keep input frame) veltype = 'radio' # velocity definition hanning = False # Turn on Hanning smoothing of spectral channels async = False # If true the taskname must be started using cvel(...)
Cvel creates a new measurement set (outputvis= 'leo2_cvel.ms') for which, at each time, the spectrum has been shifted and channels regridded to keep the spectral line centered in the frame of your choice (here, we choose barycentric; outframe= 'BARY'). You'll also want to give cvel the rest frequency of your spectral line (restfreq= '1420405751.786Hz' for HI).
In preparation for subtracting the continuum, let us plot up the combined spectrum on our science field and identify some line-free channels. Open up a plotms GUI window and load the doppler-tracked measurement set containing your science target. You'll want to average over both time and baselines to get as much signal-to-noise as possible, hopefully revealing a nice 21 cm profile. In the Data tab of plotms, set the below averaging options:
(See Averaging data in plotms for more details on averaging options).
You'll also want to click on the Axes tab in the plotms window, and change the axes settings to:
The figure to the right shows the resulting plot in plotms. There is clearly some line emission around channel 190, and possibly some faint emission around channel 88. Let's use channels 10--55 and 120--155 to fit the continuum.
Next, use uvcontsub to subtract the continuum from your data set. Here are the parameters we used:
# uvcontsub :: Continuum fitting and subtraction in the uv plane vis = 'leo2.ms' # Name of input visibility file field = '' # Select field using field id(s) or field name(s) fitspw = '0:10~55;120~155' # Spectral window/channel selection for fitting the continuum spw = '0' # Spectral window selection for subtraction/export solint = '60s' # Continuum fit timescale fitorder = 0 # Polynomial order for the fit fitmode = 'subtract' # Use of continuum fit (subtract,replace,model) splitdata = True # Split out continuum, continuum-subtracted data async = False # If true the taskname must be started using uvcontsub(...)
Here, we are fitting the continuum to channels 10--55 and 120--155 of spectral window 0 (the only spectral window). We're averaging over one minute intervals (solint= '60s') before fitting the continuum; note that the default will fit each integration (so, every one second for un-averaged EVLA data). We're fitting a simple mean to the continuum (fitorder= 0), although higher-order fits are certainly possible.
Note that the form of the output from uvcontsub can be a bit confusing, and depends on your choice of the fitmode and splitdata parameters. If splitdata= True, two new measurement sets will be created: one with the fitted continuum data ('leo2.ms.cont') and one with the continuum-subtracted data ('leo2.ms.contsub'). Uvcontsub will also alter your 'corrected' and 'model' data columns---exactly how depends on your choice of fitmode. We have set fitmode= 'subtract', which means the fitted continuum values are placed in the 'model' column, and the continuum-subtracted data are placed in the 'corrected' column.
If you have previously applied a calibration to the measurement set that you now want to continuum-subtract (with applycal, so that the 'data' and 'corrected' columns are different), note that uvcontsub will overwrite the 'corrected' data column. In this case, it is best to first create a new calibrated measurement set using split (datacolumn= 'corrected'), and then run uvcontsub on that newly-split data set.
Make an Image Cube
It's finally time to make an image! Pretty much all imaging in CASA should be done with clean, including mosaicing. Here, we discuss interactive imaging/deconvolution, but see the NGC 5921 tutorial for an example of non-interactive imaging.
# clean :: Invert and deconvolve images with selected algorithm vis = 'leo2.ms.contsub' # Name of input visibility file imagename = 'leo2_cube0' # Pre-name of output images outlierfile = '' # Text file with image names, sizes, centers for outliers field = '' # Field Name or id spw = '' # Spectral windows e.g. '0~3', '' is all selectdata = False # Other data selection parameters mode = 'channel' # Spectral gridding type (mfs, channel, velocity, frequency) nchan = -1 # Number of channels (planes) in output image; -1 = all start = 0 # First channel to use (0=first channel specified in spw) width = 1 # Number of input channels to average interpolation = 'nearest' # Spectral interpolation (nearest, linear, cubic) outframe = '' # velocity frame of output image gridmode = '' # Gridding kernel for FFT-based transforms, default='' None niter = 10000 # Maximum number of iterations gain = 0.1 # Loop gain for cleaning threshold = '0.0mJy' # Flux level to stop cleaning, must include units: '1.0mJy' psfmode = 'clark' # Method of PSF calculation to use during minor cycles imagermode = '' # Options: 'csclean' or 'mosaic', '', uses psfmode multiscale =  # Deconvolution scales (pixels);  = standard clean interactive = True # Use interactive clean (with GUI viewer) npercycle = 500 # Clean iterations before interactive prompt (can be changed) chaniter = False # Clean each channel to completion (True), or all channels each cycle (False) mask =  # Cleanbox(es), mask image(s), and/or mask region(s) imsize = [512, 512] # x and y image size in pixels. Single value: same for both cell = ['12.0arcsec', '12.0arcsec'] # x & y cell size(s). Default unit arcsec. phasecenter = '' # Image center: direction or field index restfreq = '' # Rest frequency to assign to image (see help) stokes = 'I' # Stokes params to image (eg I,IV, QU,IQUV) weighting = 'briggs' # Weighting of uv (natural, uniform, briggs, ...) robust = 0.0 # Briggs robustness parameter npixels = 0 # number of pixels to determine uv-cell size 0=> field of view uvtaper = False # Apply additional uv tapering of visibilities modelimage = '' # Name of model image(s) to initialize cleaning restoringbeam = [''] # Output Gaussian restoring beam for CLEAN image pbcor = False # Output primary beam-corrected image minpb = 0.2 # Minimum PB level to use calready = True # True required for self-calibration async = False # If true the taskname must be started using clean(...)
The viewer which pops up during interactive clean
This will create an image cube called 'leo2_cube0.image', along with some other relevant files (all with the prefix 'leo2_cube0'). We are making a cube (mode= 'channel') which includes all channels in our measurement set (nchan= -1, start= 0). Each image plane will be 512x512 pixels (imsize) and each pixel is 6x6 arcseconds (cell). The images will be made with the robust = 0 weighting (weighting= 'briggs', robust= 0.0) that was the default in AIPS IMAGR.
Wait while CASA processes your measurement set, producing output that looks like this on the command line:
CASA <25>: go clean ---------> go(clean) Executing: clean() 0%....10....20....30....40....50....60....70....80....90....100% 0%....10....20....30....40....50....60....70....80....90....100%
After the second '100%', a viewer window will pop up with your dirty image cube (like the figure above and to the right). You can make the image itself bigger in the viewer by clicking on the dashed line above the panel with the DVD-like control buttons and dragging it onto your desktop. Similarly, you can also drag the bottom-most panel which tracks your mouse and displays basic image information out of the viewer and on to the desktop (see the figure to the right).
To view different channels, use the DVD-like control panel. The button will "play" through the channels at the speed set by the Rate slide bar. To advance by just one channel, click the icon. To zoom in on a portion of the image, click the (near the top of the main viewer window) with the mouse button you'd like to use for zooming (note the three little squares beneath the magnifying glass in the icon. Here, the left square is black, which means the left mouse button is activated for zooming). Click on the image with the zoom-activated mouse button, and then hold and drag to draw a square around the region of the image you'd like to zoom in. A green square will appear---double click in this square to zoom into it. To zoom out, click the (zooms partially out) or the button (zooms all the way out to show the whole image). If you are unhappy with the color map of the image, fiddle with it by clicking on the or button and then, holding down the colormap-activated mouse button, move your mouse around the image. To change the range of data values displayed, click on the button to call up the Data Display Options window. In this window, under the Basic Settings menu, change the minimum and maximum Data Range.
Drag panels out of the viewer and on to your desktop to enlarge the image
After getting a good look at the image cube in viewer, it's time to start setting clean boxes. For this, we'll use the and buttons. Make sure Add (and not Erase) is selected near the top of the viewer window in the menu that looks like this: . You'll probably want to mark a box in just one channel, so select Displayed Plane from the menu that looks like this: . Then, find some flux that you would like to clean, and click the icon with the mouse button you would like to use for setting clean boxes. Click and drag with this mouse button to create a green square, which will represents a potential clean box. However, it is not a clean box until you double click in it, and the square turns pink. Alternatively, the button allows you to set irregularly-shaped clean boxes by clicking to mark the vertices of a polygon. Double click to complete the polygon, and then double click again to turn the polygon into a pink clean region. In the occasion that you would like to delete a clean box, select the Erase option from the green Add/Erase menu, and then make a green square or polygon around the clean box. In this case, when you double click, any clean boxes inside the green region will disappear. Feel free to set clean boxes in multiple channels. Once you are happy with the clean boxes, click the button to perform a deconvolution iteration.
Clean will then clean npercycle components (here set to 500) from each channel with a clean box in it. When it has completed this, the viewer will become active again and you can place more clean boxes if necessary. When ready, click the button again for another deconvolution cycle---or you can continue cleaning non-interactively (up to the limit set by the niter or threshold parameter) by clicking . If you'd like to stop deconvolution all together, click the button.