10.2 Single-dish data in the “uv” domain

Once you have gotten your data into , a wide range of tasks become available to you. In addition to the single-dish specific tasks discussed below, these include data movement tasks (UVCOP, UVSRT, DBCON), data averaging (AVER, UVAVG, AVSPC), non-interactive editing (CLIP, UVFLG), interactive editing (SPFLG, EDITR, TVFLG), data backup and restore (FITTP, FITLD), and data display (PRTAN, PRTUV, UVPRT, UVPLT).

10.2.1 Using PRTSD, UVPLT, and POSSM to look at your data

In the process of calibrating, modeling, editing, and imaging of single-dish data, there are occasionally problems that seem to arise because users are not aware of the data that they actually have. PRTSD is the task for such users. It displays the data with or without calibration for selected portions of your data set. This will help you identify what pointing positions actually occur in your data, which channels are highly variable or bad, and the like. SPFLG, UVPLT, and others are good for looking at the data set as a whole, but PRTSD really shows you what you have.

To run it, type:

PRTSD will start and, after a pause to get through any data not included at the start of the file, will begin to display lines on your terminal showing the scan number, time, coordinates, and data for six spectral channels. After 20 or so lines, it will pause and ask if you want to continue. Hit  C R to continue or type Q  C R or q  C R to quit. If you decide to get hard copy, set DOCRT = -1 and the output will be printed. To save the display in a text file, without printing, set DOCRT = -1 and give the name of the file in the OUTPRINT adverb. See §3.2 and §3.10.1 for more information on printing.

There are a number of tasks which plot uv visibility data; see §6.3.1. The most basic of these is UVPLT, which can be useful for single-dish data sets. For example, to generate the plot of flux versus time in 12m OTF beam-switched continuum differenced data seen in the accompanying figure (Figure 10.1), the parameters given below were used:

After UVPLT is running, or better, after it has finished:

The second plot in Figure 10.1 was generated with BPARM = 6, 7 and shows where samples occur on the sky in a different data set.

With spectral-line data, POSSM will plot observed spectra averaged over selected “antennas,” time ranges, and the like. Thus,

LWPLA was then used to make a PostScript version of the plot seen in Figure 10.2.

10.2.2 Using UVFLG, SPFLG, and EDITR to edit your data

Editing is the process by which you mark data samples as “unreliable” or “bad.” In , there are two methods for doing this. The simplest is to have the editing software alter the weight of the sample to indicate that it is flagged. If the data are not compressed, this is a reversible operation. If the data are compressed, however, then the data themselves are marked as “indefinite” and the operation is not reversible. The second method is the use of a flag (FG) extension table attached to your uv data set. This method requires that the data be sorted into time order for large FG tables and is supported by most, but not all, tasks. Small (< 6000 row) FG tables may be used with data in any sort order. If the task does not have the FLAGVER adverb, then it does not support flag tables. However, since flag tables can be applied to the data by SPLIT, we use them in the recipes below.

To sort the data into “time-baseline” (TB) order,

The most direct flagging task is UVFLG, which puts commands into the flag table one at a time (or more than one when read from a disk text file). To use this task to flag channel 31 from 7 to 8 hours on the first day of observation from the second input (single-dish nomenclature) IF:

Multiple runs of UVFLG may be done to incorporate what you know about your data into the flagging table. Use PRTSD and the plot programs to help you find the bad data. If you have a long list of flagging commands, you may find it easier to use the INTEXT option of UVFLG to read in up to 100 flagging instructions at a time from a free-format text file.

The task CLIP is popular on interferometer data sets since it automatically flags all samples outside a specified flux range without interaction with the user. This blind flagging is often acceptable for interferometer data since each uv sample affects all image cells so that the damage done by a few remaining bad samples is attenuated by all the good samples. However, a bad sample in single-dish data affects only a few image cells and is hence not attenuated. Thus it is important to find and remove samples that are too small as well as those that are too large. For this reason, we do not recommend CLIP, but suggest that you look at your data and make more informed flagging decisions.

The best known of the interactive editing tasks is TVFLG (§O.1.6). This task is not suitable for single-dish data since it displays multiple baselines along the horizontal axis. The data on these baselines are related in interferometry, but, in single dish, they are from separate feeds or polarizations and hence neither numerous nor necessarily related. For spectral-line single-dish data, the task SPFLG is an ideal task to examine your data and to edit portions if needed. SPFLG is a menu-driven, TV display editing task in which spectral channel varies along the horizontal axis of the TV display and time along the vertical. (The spectral channels for each interferometer IF are displayed on the horizontal axis, but single-dish data in  has only 1 of this sort of IF.) The data may be displayed with as much or as little time averaging as desired and is very useful for examining your data even if you do not think that editing is needed.

To run SPFLG, type

The task will then read your data to determine which times occur in the included portions (you may set TIMERANG, restrict autocorrelations, etc.) and then construct a master grid file with spectral channel as the first axis, pseudo-regular times on the second axis (gaps are mostly suppressed), and, if needed, baseline number on the third axis. SPFLG tells you the size of the resulting file, e.g., SPFLG1: Basic UV image is 128 14079 pixels in X,Y (Ch,T).

At this point, SPFLG selects an initial display smoothing time long enough to fit all of the master grid onto your TV window. It then averages the data to this interval and creates a display not unlike that seen in Figure 10.3. Move the TV cursor to any menu item (it will change color to show which has been selected) and press button D for on-line help information or press buttons A, B, or C to select the operation. Normally, you will probably begin by reducing the smoothing time (ENTER SMOOTH TIME menu option followed by typing in the new smoothing multiple on your AIPS window). Note that the display does not change other than to add an asterisk after the smoothing time to indicate that that will change on the next image load. This behavior is to allow you to alter a number of choices before doing the potentially expensive TV display. In this typical example, you would either ENTER BLC and ENTER TRC by hand and finally LOAD the sub-image or you can do this interactively with SET WINDOW + LOAD. You may examine data values (like CURVAL) and flag data with the options in the fourth column. Flagged data are removed from the display. You may review the flags you have prepared, undo any that you dislike, re-apply the remaining ones to make sure the display is correct, and modify the appearance of the display with the options in the first column. The image may be shown in zoom only during editing in order to give you greater accuracy in examining the data values and locations. If you are doing some time smoothing within SPFLG, the DISPLAY RMS option allows you to view images of the rms rather than the value of the time average. Such a display allows you to find excessively noisy portions of the data quickly.

Figure 10.3: A display of a sample TV screen from SPFLG on single-dish data, made using the  task TVCPS to produce a negative black-and-white display. The SPFLG menu (in the boxes) and status lines (at the bottom) are displayed in a graphics plane which is normally colored light green. The data are grey scales in a TV memory and may be enhanced in black-and-white or pseudo-colored. The data actually displayed range in intensity from -1.7 to 5.2 Kelvins (as stated during the image loading) and have been averaged to 0.8 seconds. The entire master grid contains 14079 times, but the current window includes only times 5403 through 9538. Flag commands generated at the moment illustrated will flag all source names, all IFs (in the  sense), only the displayed baseline, and all Stokes. Note that the menu displayed is now out of date, more options are available in later versions.

Finally, when you are done, select EXIT. If you have prepared any flagging commands, SPFLG will ask you if you wish to enter them into your input data set. Answer yes unless you want to discard them or you have set DOCAT TRUE to catalog the master file in order to use it for multiple sessions. If you set OUTFGVER to zero, then the flag commands are put into a new flag table which can be deleted later if you wish.

SPFLG is not useful on continuum data; the interactive editor of choice for such data used to be the task IBLED, but is now EDITR. These tasks are also useful for spectral-line data in that they can display the average (and rms) of a selected range of channels. The spectral averaging should let you see more subtle level problems than can be seen on individual channels (i.e., in SPFLG). EDITR is a menu-driven, TV display editing task, but it does not use grey-scales to show data values. Instead, it plots time on the horizontal axis and data value on the vertical axis. The full data set for the chosen baseline is displayed initially in a potentially crowded area at the bottom of the TV window. This area is available for editing. If DOTWO is true, then it also displays above the edit area a second observable (initially the difference between the amplitude and a running mean of the amplitude) for the primary baseline. EDITR allows you to display up to ten other “baselines” (e.g., 12m-antenna IFs) in frames above the active editing frame. These should speed the process of editing and guide you in the choice of flagging one or all baselines at the time of the observation. A smaller time range or window into these full data sets may be selected interactively to enable more detailed editing. Be sure to set SOLINT to specify an appropriate averaging interval. Unlike SPFLG, no further time averaging is possible. The menu options allow you to work your way through all of your data, selecting time windows and baselines as desired. Consult §5.5.2 for more details about EDITR.

EDITR has the ability to display a second data set for reference in parallel with the one being edited. This option is likely to prove useful for beam-switched continuum observations. Select one of the beam throws for editing and the other for reference display. Then, if editing is required, reverse the roles. It may also be useful to look at your beam-switched data in its differenced form. The task DIFUV may be used to difference the plus and minus throws, followed by EDITR (or any other  uv-data task) to look at the differences. Be sure to tell DIFUV that the time difference between the plus and the minus beam throws should not be considered significant, i.e., SOLINT = 1 / 8 / 60 or a little bit more to avoid round-off effects.

10.2.3 Using CSCOR and SDCAL to calibrate your data

The current calibration routines for single-dish data in  are fairly rudimentary. The concept is similar to that used for interferometers. Corrections are developed in an extension table (called CS in single-dish, CL in interferometry) which can be applied to the data by some tasks. In particular, the single-dish tasks PRTSD and SDGRD are able to apply the CS table to the data without modifying the data as stored on disk. They do this using the DOCAL and GAINUSE adverbs. Other uv tasks, designed primarily for interferometry, also use these adverbs, but do not understand or apply CS tables. For such tasks, you should carefully turn off the calibration option. If you do not, such tasks will fail.

There are two tasks which can create CS tables: SDTUV discussed above and INDXR. To use the latter, enter

Note that this task requires the data to be in time order and expects an antenna (AN) table. You may set CPARM(5) to the maximum antenna number (beam number) in your data set and, with a few grumbles, INDXR will still create and initialize a CS table when you do not have an antenna table.

At this writing, the CS table may be used to correct the recorded right ascension and declination (i.e., the pointing) and to correct the amplitudes for atmospheric opacity and other gain as a function of zenith angle effects. To add an atmospheric opacity correction to the CS table produced by INDXR, type:

Note that CSCOR only writes those records in the output file that you have selected via TIMERANG, ANTENNAS, etc. To make a new CS table to work for the full data set, you should first use TACOP to write the new table and then set GAINVER and GAINUSE to both point at the new table. CSCOR needs to compute the zenith angle and therefore needs to have an antennas file. If your data set does not have one, you may give the antenna longitude and latitude in the CPARM adverb. The other operations offered by CSCOR are GAIN, PTRA, and PTDC which apply as second-order polynomial functions of zenith angle corrections to the gain, right ascension, and declination, respectively. The format of the CS table allows for an additive flux correction as well. There are no tasks at this time to determine such a correction.

The basic single-dish tasks PRTSD and SDGRD can apply the CS table to the data as they read them in. Other uv tasks which are more directed toward interferometry data cannot do this. If you need to use such tasks with corrected data, then you must apply the corrections with SDCAL and write a new “calibrated” data set. To do this:

The output file from SDCAL can then be fed to UVPLT, SPFLG, or any other uv-data task including of course PRTSD and SDGRD.

10.2.4 Using SDLSF and SDVEL to correct your spectral-line data

It may be convenient to remove a spectral baseline from each sample before the imaging step. Doing so may allow you to skip the removal of a spectral baseline from the image cubes (as described in §10.4.1). To do this, type:

You may, and probably should, use FLUX and CUTOFF to flag those data having excessive noise or excessive signals in individual channels. These “excesses” are measured only in the channels selected by CHANSEL for fitting the baseline.

If you have observed a wide field with relatively narrow spectral channels, there is an effect which you should consider. The “velocity” corresponding to a particular frequency of observation depends on the velocity definition (e.g., LSR or heliocentric), the direction at which the telescope pointed, the time of year, the time of day, and the location of the telescope. Most telescopes adjust the observing frequency to achieve the desired velocity for some reference time and position and many adjust the frequency periodically to account for time changes. However, few, if any, can adjust the observing frequency for every pointing direction and time in a rapidly scanned on-the-fly observing mode. The 12m telescope now sets the frequency once per image with respect to the reference coordinate (usually the image center). In this mode, the maximum velocity error in a 2 degree by 2 degree image is about 1.16 km/s (in LSR velocities) and 0.79 km/s (heliocentric). Since mm lines are often narrow, this can be a significant effect. Fortunately, single-dish OTF data may be fully corrected for this effect so long as your spectra are fully sampled in frequency. The task SDVEL shifts each spectrum so that the reference channel has the reference velocity for its pointing position. The DPARM adverb array is used to tell the task how the telescope set reference velocities and to ask the task to report any excessive shifts and even flag data having really excessive shifts. The latter are to detect and/or remove times in which the telescope pointing was significantly in error (i.e., high winds). DPARM(1) should be set to 0 for 12m data taken after 5 May 1997 and to 2 for data taken before that date. The task VTEST was written to help you evaluate the magnitude of this effect.

10.2.5 Using SDMOD and BSMOD to model your data

It is sometimes useful to replace your actual data with a source model or, if your continuum levels are well calibrated, to add or subtract a model from your data. The task to do this is called SDMOD and allows up to four spatially elliptical Gaussians (or an image) to replace the data, or to be added to the data, with either a Gaussian or no frequency dependence. When the data are replaced, a random noise may also be added. SDMOD has options for modeling beam-switched continuum data (set BPARM(1) = 1) as well as for spectral-line data. For example, to see what a modestly noisy point source at the origin would look like after all of the imaging steps:

The output file from SDMOD can then be fed to SDGRD, BSGRD, or any other appropriate task as if it were regular data. The input model is convolved with the single-dish beamwidth given in the uv data header before being used to replace or add to the input data. The history file will show in detail what was done.

Beam-switched observations may be modeled with task BSMOD. No input data set is needed. Instead two regular grids of switched data are constructed from a specified model plus noise and a variety of instrumental defects.